Guidelines for Best Practice in Unsupervised Online Assessment Introduction CEB recognises that on-demand online solutions are now the reality in the assessment of people at work. During 2005 alone, We administered over one million assessments online. Organisations large and small are alert to the massive cost benefits of testing candidates unsupervised online for selection, particularly when screening large numbers of applicants during a recruitment programme. In this context, ‘unsupervised’ means that there is no person present with the test taker to authenticate their identity and to oversee them taking the test. This unsupervised online environment has created a new landscape in which new issues must be addressed to preserve the integrity of the test and the accuracy of the outcome. Part of the benefit for test takers is that they take tests in their chosen environment – at home, in an Internet café, or at work. Part of the benefit for organisations is the saving on costs associated with traditional supervised assessment. However, it is difficult to confirm an applicant’s true identity, or determine whether they have had assistance when tests are completed without supervision. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 1 Organisations large and small are alert to the massive cost benefits of testing candidates unsupervised online for selection. The emergence of online testing has led to a debate about what is good practice (i.e. what are the minimum standards and safeguards one should adopt) and what is ‘best practice’ (i.e. what should one aim for) when dealing with unsupervised assessment in high stakes settings. The challenge faced by unsupervised testing online is achieving the correct balance between minimising risk and maximising value. The use of objective assessments must be properly managed in order to reap the anticipated benefits. This becomes more important when tests are delivered unsupervised over the Internet. If inappropriate tests are used, if the information from tests is misunderstood, or if the process is not secure, the benefits can be quickly eroded. Feedback from SHL clients and test users has highlighted the need to help practitioners to do things better - help them to move in the right direction, to assess the risks and mitigate them where possible. These guidelines provide a pragmatic, stepby-step set of recommendations that fit into the reality of the world in which today’s organisations operate. These guidelines are not built on desk or academic research alone. Interviews with users of assessments around the globe, including professional psychologists, consultants in the recruitment and coaching markets, HR and line managers in both large and medium-size organisations in the private and public sectors all contributed to the development of this Guide. Who Are These Guidelines For? These guidelines are designed to assist HR practitioners who are developing unsupervised online assessment programmes for the first time and practitioners who already use online tests but are expanding or evaluating their use. In addition, these guidelines are a useful reference for IT professionals who need to understand the broader context around the introduction of an online testing programme. Appropriate Use and More Information This Guide highlights the areas that should be considered when designing online assessment processes. This document, alongside the White Paper on this subject, should be considered as a framework rather than an absolute reference containing all the answers. A reference list of additional sources on the subject of online testing is in the Appendix. More information can also be found at www.ceb.shl.com. CEB can supply expert consultancy services if further guidance is needed. In all cases, readers should ensure that the use of these guidelines and the development of any procedures or policies comply with local legislation and/or professional practices. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 2 Structure of the Guide The document takes a step-by-step approach to the development of unsupervised online assessment programmes: Lay the Foundations Step 1 Evaluate if unsupervised testing online is right for your organisation Step 2 Determine the objectives, stakes and scope of your programme Step 3 Identify where your process needs to change and the impact of the change Develop Policy and Procedures for Unsupervised Testing Online Step 4 The four Better Practice principles for unsupervised testing online: ■■ Manage the security of the assessments ■■ Defend against cheating ■■ Build in verification procedures ■■ Establish and communicate an online assessment contract with the candidate Step 5 Define new policies & procedures for unsupervised test usage in your organisation Implement the New Process Step 6 How to apply the Better Practice principles: a. Choosing a test provider b. Things to look out for and avoid c. Developing a new policy for unsupervised testing online d. Responsibility towards test takers before, during and after the online test session e. Guidelines on using cut scores f. Cross checking results with other information g. Retesting and verification testing h. Supervising the verification test i. Handling verification discrepancies j. Monitoring test usage and results k. Confidentiality, storage of results and access to test materials www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 3 Lay the Foundations – Companies can realise many benefits from online testing but they must be carefully weighed against the potential risks. Step 1: Evaluate if unsupervised testing online is right for your organisation The foundation of an effective unsupervised testing online policy is being clear on whether or not it is appropriate for your organisation. Companies can realise many benefits from online testing but they must be carefully weighed against the potential risks. Benefits Earlier Decision-Making Unsupervised online processes mean tests can be delivered to applicants during preliminary stages, which means that information can be used for screening and decision-making much earlier. Saves Time By identifying the best talent earlier and progressing them faster, the length of hiring processes can be reduced. Online testing processes also save time by removing the need for staff to physically administer tests until a later stage, and then for a smaller pool of candidates. Saves Money Unsupervised testing online enables HR teams to do more hiring with the same HR staff. There is no need to purchase, store and transport materials to assessment centres. Costs are also saved through better hiring decisions, which reduce staff turnover and improve performance. Administrative & Practical Efficiency The benefits of unsupervised testing online are particularly strong in high volume recruitment scenarios. The process of accessing candidates in different locations and in different languages is far easier. It requires little administrative effort and results can be accessed faster. A 3rd party workflow or Software-on-Demand provider can provide the online platform to enable organisations to adopt this approach. (In an online testing context, a ‘Software-on-Demand’ provider owns and operates a software application that facilitates the delivery and scoring of an assessment. The provider maintains and services the application and makes the online assessments available to users via a browser, or within a client application.) Positive Candidate Experience Applicants complete tests when it is convenient for them, in their preferred language, with 24 hour accessibility 7 days a week. As a highly convenient, speedy and modern process, online testing delivers a positive candidate experience, which can reinforce an employer brand. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 4 International Consistency and Consistency Across Sites Unsupervised online testing offers organisations, especially those with multiple locations, the opportunity to standardise their approach to assessment. The availability of the same tests in multiple languages enables multinational organisations to create a consistent hiring process globally. Centralised Data Storage and Reporting A significant benefit of online testing, especially through a ‘Software-on-Demand’ model, is that all recruitment data is held in a single, globally accessible system (subject of course to access authorisations and data protection regulations). Risks Psychometric Integrity When objective assessments (and in particular ability tests) are delivered online, the impact on the psychometric properties behind the test is significant. Organisations need to ensure that the underlying science of their online test is robust. The test provider should ensure that security and protection methods are in place to protect the test content, and also that candidates cannot preview and practice the test covertly. Technology Issues Ensuring that an online testing system performs reliably and provides a smooth experience to users is a key requirement. There are steps that can be taken to manage the risks. Putting the IT department in contact with the test provider at the outset will ensure that any possible issues can be foreseen and overcome before an online testing programme is launched. Reputational Issues Moving to unsupervised online test delivery reduces the level of control over the test environment and reduces the opportunity to engage in a dialogue with the candidate at the time of the test. It is important to ensure that the candidate can obtain help and support if they need it, and that reassurance is provided to candidates that the testing process is fair to all and not open to abuse. Legal Implications It is critical that a clear assessment policy is in place that outlines how the assessment process is appropriate for the business need and meets the requirements of employment law. The policy should also address assessment integrity, security and cheating issues, which are specific to the unsupervised online testing context www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 5 Fairness and Quality Data presented at the Association of Test Publishers conference in February 2006 showed that the number of distractions experienced with online tests was no greater than those experienced in traditional proctored or classroom testing. There is mounting evidence to remove the concerns around online testing in areas such as gender and age bias. Examples of such evidence include the following: In early 2006, the Irish Civil Service reported that online job applicants had a broader age range and reflected the same proportion of women compared to offline applicants. In addition, online applicants outperformed offline applicants on selection tests (source: www.management-issues.com). Data presented at the Association of Test Publishers conference in February 2006 showed that the number of distractions experienced with online tests was no greater than those experienced in traditional proctored or classroom testing. In 2005, we surveyed 1,300 people participating in test development trials. This sample covered potential job applicants with an equal gender split, age range from 18 to 70 and a wide educational range. 63% of respondents preferred to complete assessments online. Lay the Foundations – Step 2: Determine the objectives, stakes and scope of your programme Regardless of the size or scope of an assessment programme, most value is gained when organisations adopt a strategic approach and link online assessment programmes into their overall talent management strategy. Key steps can then be followed to define the objectives, stakes and scope of a specific project. Case study: Online testing achieves business needs for an investment bank With thousands of graduate applicants across several countries, this organisation looked to unsupervised ability testing online to meet three objectives: 1. Improve the quality of candidates attending the assessment centre. 2. Speed up the process of identifying those who should be called forward. 3. Reduce the overall cost-to-hire of graduate entrants to the organisation. Introducing unsupervised online ability tests online resulted in a 40% improvement in the success rate of those attending the assessment centre stage, enabled the organisation to identify talent at a much earlier stage in the process and enabled the organisation to reduce its cost by 37% from US$3,200 to US$2,000 per graduate hire. 1. Specify the business need Articulate the business need that the assessment programme will address, including how unsupervised online testing will meet that need and what metrics will be used to measure success. For example, online assessment may enable faster and wider access for candidates and may improve the organisation’s ability to collect and link assessment data with performance data. 2. Define the key stakeholders For small-scale assessment tasks, it is critical that the internal client is clearly identified and requirements are agreed as early as possible. For larger programmes, the range of stakeholders can be fairly large and involve a number of disciplines and backgrounds. For example, where complex technical requirements such as integration with other applications are involved, then an IT contact engaged from the start is critical. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 6 3. Define the needs of the intended users Users of online assessments include those responsible for managing the administration of assessments, the candidates, and those who will make decisions based on assessment results. Reports summarising the results of an assessment may also need to support different users, such as the HR function and the candidate’s future manager. 4. Clarify the project scope and investment choices to be made Scope will dictate complexity, time and cost of a project and will also highlight who needs to be involved in requirements and decision-making, how suppliers will be selected and managed, and how internal ambitions and expectations will be managed once the assessment is deployed. Key questions to ask that can help clarify the scope of an unsupervised online testing project include: What Levels of Volume and Scalability Will be Needed? ■■ ■■ ■■ How many roles and candidates does the system need to cater for? Are there peak periods - if so, what volume must the system handle at any one time? What levels of security are required? What kind of ‘future proofing’ is required to ensure the system can evolve? How Much Functionality is Required? ■■ ■■ ■■ ■■ ■■ ■■ ■■ Is there one stage of assessment or several? What types of assessment tools are required - a sifting tool to screen candidates, or a richer assessment against core competencies? How are the assessment stages organised – sequentially, or in parallel? What navigation does the user need to access assessments and reports? Does the assessment need to integrate with other applications or offline processes? What types of reports are required for who and when? What level of training is required for users of the assessments? What Other On-Going Costs are Likely to be Involved? ■■ In addition to the costs associated with functional requirements (system, training etc.), other costs should also be considered, such as staffing and infrastructure to support on-going use of the assessments. Buy, Build or Lease? ■■ Will the assessment and delivery technology require a custom-built system? Will this be undertaken internally or through an external contractor (such as a Software on Demand provider)? www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 7 Case study: New legislation leads to the introduction of online testing What is score aberrance?Aberrance is observed when a subject answers the test questions in a manner that is inconsistent with demonstrated knowledge and behaviour. Examples are inconsistencies in the amount of time taken to respond to test items, and answer selections that are inconsistent with a student’s demonstrated ability on other test items. Source: Caveon Test Security New European Union legislation which required a demonstration of competence in technical sales staff, led this multinational organisation to undertake a review of the key competencies set out in the legislation against those used in its hiring, training and promotion processes for key sales roles, including front-line sales, account management and customer care. They identified that a number of softer skills relating to communication and customer care were not sufficiently covered at any of these key employment stages, and the links between recruitment, training and development needed to be strengthened and more clearly documented. As such, the organisation felt it was exposed to possible regulatory challenge. These competencies were translated into a suite of assessment tools that were delivered online across all of its European geographies and in local languages. This was further strengthened by documented evidence showing how the online assessments, which included screening tools for hiring as well as 360-degree tools for development and performance management, were related to business metrics ranging from sales revenues, through to customer satisfaction. Lay the Foundations – Step 3: Identify where your process needs to change and the impact of the change For most organisations, the integration of an online testing programme represents more than a straightforward addition of ‘psychometric testing’ to an existing recruitment or development process. In reality, it often provides an opportunity to re-engineer the entire process. Embarking on the journey with this expectation up front, and by approaching it as a change programme, will lead to greater success. Even if an organisation is already testing candidates online under supervised conditions, introducing unsupervised online testing brings with it significant challenges. Along with the obvious introduction of a new technology, it will also create different demands on staff, which leads to training and potentially different skill-set requirements. It is critical to outline both the current process and the new process to identify what needs to change, and also the impact the changes will have. Map out current process – who is involved, at what stage, and how? Map out the new process – who is involved, at what stage, and how? Put a plan in place to manage the changes Compare the existing processes with the new, and identify areas that will change www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 8 When comparing the old and new process, consider each of the following areas and how the requirements around each will change with the introduction of the new process: ■■ Confidentiality and storage of results. ■■ Test taker responsibilities. ■■ System administrator responsibilities. ■■ Communications to candidates and users of the system. ■■ Delivery of tests at the test session and troubleshooting. ■■ Scoring and reporting. ■■ Using results for decision-making. ■■ Feedback and communication after the test session. ■■ Communicating follow-up tests if applicable. ■■ Handling verification result discrepancies. ■■ Monitoring and identity authentication. ■■ Training, access and usage requirements. ■■ Who will access and use the tests and results? ■■ What are the new training requirements for users of the system/reports? Reading the application notes in Step 6 (‘How to Apply the Better Practice Principles’) in this document will provide valuable guidance on these areas. Develop Policy and Procedures – Step 4: The four Better Practice principles for unsupervised testing online As already outlined, the delivery of tests in an unsupervised online environment is not without its risks. As a globally-recognised, independent advisor in the field of objective assessment, the International Test Commission (ITC) has developed a set of guidelines for promoting good practice when testing on computer and online in their International Guidelines on Computer-based and Internet-delivered Testing (2006). We strongly recommends the ITC guidelines as a key reference for any organisation about to implement online testing. Consistent with the ITC’s work, CEB has provided additional Better Practice recommendations to preserve the integrity of unsupervised testing online. These take the shape of four principles that can be followed to overcome the risks and realise the benefits from unsupervised testing online: ■■ Better Practice principle 1 – Manage the security of assessments ■■ Better Practice Principle 2 – Defend against cheating ■■ Better Practice Principle 3 – Build in verification procedures ■■ Better Practice Principle 4 – Establish and communicate an online assessment contract with the candidate These four Better Practice principles have been taken from the SHL Better Practice for Unsupervised Online Assessment White Paper, in which can be found more detail supporting each of these principles. By following these principles, an organisation should gain a higher degree of confidence that the tests and test results are fair, objective and accurate measures of performance potential. Following the ITC guidelines for unsupervised assessment online, the basic requirements that we recommend: 1. The candidate is registered for the assessment. 2. The candidate is issued with an ID and password. 3. The candidate is required to accept terms and conditions for the assessment, which includes an ‘honesty contract’ and also their consent to a verification of their test performance at a later date (see Principles 3 and 4). www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 9 What is collusion? “Collusion occurs when examinees share answers for the test items either during or before the test. It also manifests itself when an educator provides the same answers to test items to multiple students. Statistically, collusion indicates that the tests are not being taken independently.” Caveon Test Security talking about educational testing ITC Guidelines (2006) cover four areas relating to good practice in the delivery of tests via computer or Internet: 1. Give due regard to technological issues This area advises on hardware, software considerations, the robustness of the test, human factor issues around presentation of tests on-screen and adjustments for testing candidates with disabilities. 2. Attend to quality issues This area covers issues to do with psychometric qualities of online tests, evidence of equivalence, scoring, result interpretation, feedback and equality of access. 3. Provide appropriate levels of control This area outlines different levels of control over test conditions, supervision, controlling prior practice and item exposure, test taker authenticity and cheating. 4. Provisions for security and safeguarding privacy This area guides on security of test materials, security of test taker data and confidentiality of results. ITC guidelines can be downloaded from www.intestcom.org Better Practice Principle 1: Manage the Security of the Assessments 1. Create and communicate an assessment policy. The policy must set out the objectives and scope of the assessment, stating the reasons for such a policy and the people, materials and systems that are affected by the policy. The objectives and scope will normally make reference to other policies developed by the organisation, such as employment contracts and employee handbooks that set out employee responsibilities for intellectual property. There will also usually be reference to trademarked/copyright information and products, as well as information technology security policies and plans. ■■ Security and protection of intellectual property and data. The policy should encompass assessment content, any preparation materials provided to candidates, the security of the technologies that will be used to deliver and score the assessments, as well as the technologies that will be used to store assessments, scores and reports. ■■ Individuals responsible for implementing and maintaining assessment security. The policy should identify those who have operational responsibility for the deployment and use of assessments and those who advise on assessment use, as well as those who will have responsibility for data security. The assessment policy should be communicated to all relevant parties and the consequences of violating the policy should also be stated. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 10 2. Implement active processes for security audits and specify how security breaches will be dealt with. Better Practice around assessment security should include active processes for auditing security and procedures for dealing with suspected or confirmed security breaches. This can be done by the organisation itself, or can be formally delegated to a supplier. Key audits that should be applied in securing assessments are: ■■ Data forensic audit. This involves analysing sets of assessment data for the occurrence of unusual or “aberrant” patterns of scores. Essentially, a data forensics audit uses a number of sophisticated indices to analyse if response patterns to an assessment are statistically different to normal response patterns that would be expected if the assessment was taken honestly. ■■ Web monitoring. This involves searching the Internet to identify web sites and discussion forums where information on assessment is being exchanged. This includes advertising or requests for proxy assessment takers, meeting boards of people seeking to collude and where content such as questions and answers may be offered by pirates. For example, among the sites that CEB monitors in order to protect tests are : www.doctorjob.com/testingzone/test.asp www.ebay.com www.vault.com www.assessmentday.co.uk (The fact that CEB monitors these sites does not mean or imply that these sites are sources of pirated information, or that they actively promote cheating.) ■■ Security audits. These involve regular checks of the security policy to confirm that the list of those responsible for security is current, that recent threats to security have been logged and the policy amended accordingly. Checks should also be made to ensure that a plan is in place to respond to security breaches – for example the use of alternate versions of the assessment, re-assessment of candidates and replacement of compromised content. Data forensic auditing and web patrolling is routinely conducted on behalf of clients to safeguard CEB online assessments. Presently, these audits are carried out on aggregated data from assessments taking place all over the world and individual organisations may not have access to data in relation to their specific organisation. However, we believe the trend is moving towards a more granular review of assessment data and it is anticipated that, in the future, organisations will be able to access data forensics for their own online candidates. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 11 Better Practice Principle 2: “Any Internet test that administers the same set of items to all examinees is asking to be compromised. At the very least, the items should be administered in a randomised order. It would be better yet to sample items from a reasonably large item pool.” Source: Unproctored Internet testing in employment settings. Personnel Psychology, 59, 189-225 (2006) Defend Against Cheating Cheats can be very innovative, as our partners (Caveon Test Security) illustrate on their website www.caveon.com. Caveon’s work with clients internationally in the fields of educational and skills testing shows that 5% - 20% of people attending a supervised, paper-based assessment are expected to engage in some form of cheating. These figures suggest that it is debatable whether supervised assessments are more secure than unsupervised ones, especially when the right safeguards are built into an unsupervised online assessment. Evidence suggests that the majority of people are in fact honest. The key issue is how to minimise the risk of cheating to ensure that an assessment is as fair as possible to honest candidates and supports valid decision-making. Organisations and practitioners in the field of objective assessment can act to manage the risks of cheating and the good news is that both the science and practice of assessment security are now very well advanced. These advances have been incorporated in these Better Practice principles, which act to prevent cheating by tackling the motivation, means and opportunity to cheat. Actions to Defend Against Cheating 1. Use cheat-resistant assessment tools. To qualify as cheat-resistant, an assessment must: ■■ minimise the potential for faking answers if the assessment is a self-report questionnaire. ■■ minimise the opportunity to obtain questions and/or answers in advance if the assessment is an ability test. 2. Adopt checks and rules in the assessment process. ■■ Design the assessment process to incorporate process checks on the validity of scores and the authenticity of the candidates (see Principle 3). ■■ Set out clear rules for the candidate about what is expected of them (see Principle 4). ■■ Communicate the process and the checks clearly to candidates, which act as tangible incentives for candidates not to cheat (see Principle 4). 3. Implement security procedures to defend the integrity of the assessment. In partnership with your assessment provider: ■■ conduct on-going forensic audits of assessment data (see Principle 1). ■■ check on the activities of online pirates and other information sources to ensure that assessment content has not been compromised (see Principle 1). Knowledge that these cheat-resistant features of assessments exist, and are backed up by well managed test security policies, will act to challenge the motivation for candidates to cheat. It will also communicate to potential cheats that the means and opportunities to cheat are being minimised and monitored. A key part of that communication is the use of verification procedures. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 12 Better Practice Principle 3: Building in Verification Procedures A verification procedure is a check on the integrity of a score, or other candidate information, that may be used to make a decision. For example, background checks are often used to check the integrity of information in resumés and CVs. Verification procedures in an unsupervised online testing context follow a similar principle by using other sources of data to check the integrity of a test result that was gained from an unsupervised online test. Depending on your process, organisational policies and the stakes surrounding the assessment, a number of verification methods could be employed, including: ■■ The use of psychometric verifications, such as in-built consistency checks within questionnaires like the SHL Occupational Personality Questionnaire (OPQ), or the use of specific verification tests, such as those in the SHL Verify Range of Ability Tests. In the case of assessing ability, this method is the strongest and most reliable form of verification. ■■ Using other assessments to check the candidate’s potential and fit to the job. Information from ability tests, personality questionnaires, assessment centre exercises and a well-structured interview can be used to contribute to an overall assessment of potential and fit. Ability and personality assessments administered unsupervised online are often used to provide data on whether the candidate should be progressed to a second stage of the process, and if there are areas that should be probed in more depth before an appointment decision can be made. Crossreferencing information from several assessments to focus on areas of inconsistency is another form of verification. ■■ The use of other verifiable information related to the assessment. For example, if the assessment is numerical reasoning, then information on the candidate’s educational performance related to numerate subjects would be relevant and could be verified. As well as deciding what types of verification procedures will be used, it is also important to decide how a verification procedure will be carried out, and at which point in the process it is undertaken. A number of possible models exist: ■■ No verification at all ■■ Random verification – where some candidates are randomly selected for verification procedures ■■ Verification is always conducted – across all scenarios and for all candidates ■■ Mixed models – where verification might always be conducted in some scenarios (for example in high stakes recruitment contexts) and not in others www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 13 SHL continues to recommend that if unsupervised ability test results are being relied upon when making ‘yes’ or ‘no’ decisions, then some form of verification should always be carried out. Providing prescriptive guidance on the process that should be implemented is not straightforward, as much depends on the context of the organisation, including: organisational policies; an organisation’s acceptance of risk; stakes of the assessment; legal considerations and regulations in that country; and presence of other checks in the process. We continue to recommend that if unsupervised ability test results are being relied upon when making ‘yes’ or ‘no’ decisions, then some form of verification should always be carried out. The importance of verification increases when the stakes of the assessment increase (i.e. in high stakes recruitment scenarios). Verification procedures need to be handled sensitively. It should be noted that if a score from an unsupervised online test appears to be inconsistent with a verification check, this does not automatically mean that the candidate has cheated. What it does mean is that the score from the unsupervised assessment may not be a true measure of the candidate’s ability and competencies, and needs to be followed up. The effectiveness of verification procedures also relies on how well the process has been communicated to the candidate and whether they have bought into an online assessment ‘contract’ with you. The use of an honesty policy and statements around verification processes are covered in Principle #4. ITC Guideline 45 addresses the need for test users to confirm the results from unsupervised tests and to ensure that test takers know this will be required. For moderate and high stakes assessment (e.g. job recruitment and selection), where individuals are permitted to take a test in controlled mode (i.e. in which, although no direct human supervision is required, the test is made available only to known test takers), those obtaining qualifying scores should be required to take a supervised test to confirm their scores. Procedures should be used to check whether the test taker’s original responses are consistent with the responses from the confirmation test. SHL continues to recommend that if unsupervised ability test results are being relied upon when making ‘yes’ or ‘no’ decisions, then some form of verification should always be carried out. Test takers should be informed in advance of these procedures and asked to confirm that they will complete the tests according to instructions given (e.g. not seek assistance, not collude with others etc). This agreement may be represented in the form of an explicit honesty policy, which the test taker is required to accept. Provide test takers with a list of expectations and consequences for fraudulent test taking practices, and require test takers to accept or sign the agreement form indicating their commitment. Source: ITC 2006 www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 14 Clearly this kind of process fBetter Practice Principle 4: Establish and Communicate an Online Assessment ‘Contract’ with the Candidate Having a technically valid online assessment supported by legitimate and defensible processes is one part of ensuring the relevance of an online assessment. Another significant part depends on how the objectives and decision rules governing an assessment are communicated to candidates. Managing expectations is critical to motivating candidates to engage positively with the assessment and to obtaining accurate and valid scores. Far from providing a distant and clinical means of assessing candidates, online assessment can provide an active means of managing perceptions of fairness and relevance. In effect, the stronger the social contract established with the candidate, the greater control the organisation will have over managing the assessment process and dealing with issues such as faking and cheating. Whether it is through a web site, web pages that precede and follow the assessment, or through information provided by email prior to an assessment, the key points to be covered in the communications supporting an assessment are outlined below. These points build on the American Psychological Association’s (APA) guidelines on the rights and responsibilities of test takers, which can be found at http://www.apa.org/science/ttrr.html. ■■ Describe what the objectives of the assessment are. This should be an explanation and not a justification or defence, and should cover the relevance of the assessment to the job, role or position. ■■ Explain why these assessments have been selected. Reiterate the relevance of the assessments to the job, role or position. ■■ Explain the process. Communicate what will happen once the candidate has registered, completed the assessment, and what will happen once their results have been received. Also explain what the ‘authentication policy’ is. While verification is about the validity of scores, authentication is about validating the identity of the candidate. Many organisations have now adopted authentication as an essential part of their selection processes, for example requiring the candidate to provide some form of identification, preferably a legal document containing photographic ID, such as a passport or driver’s licence. ■■ Explain the rules for the assessment. This should include confidentiality of assessment results and reports, when and in what form feedback or decisions will be communicated to them, and what they are signing up to, including what the consequences will be of fraudulent test taking or cheating. ■■ Include an honesty agreement. A key step in ensuring that candidates understand the rules for the assessment is requiring them to agree to a set of conditions before they proceed with the assessment. This process can be described as signing up to an ‘honesty agreement’. The conditions should include: ■■ The expectation that the candidate will take the assessment honestly and without the help of others ■■ The expectation that the candidate will respect the confidentiality of the assessment and not share its contents with others ■■ The expectation that the candidate will respect copyright, trademarks and other legal rights over the content of the assessment ■■ That the candidate accepts that verification will be undertaken and that they may be required to sit a verification assessment at some point in the process www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 15 Managing expectations is critical to motivating candidates to engage positively with the assessment and to obtaining accurate and valid scores. Please see the box below for an example ‘honesty agreement’ from the SHL Talent Measurement online system, which is used for all our online unsupervised tests. “You are about to begin a critical reasoning assessment. You will receive detailed instructions and a practice test before starting the real test. You can exit this process at any time before the real test begins. Please read and accept these conditions before proceeding: ■■ I will take this assessment honestly and without assistance from others. ■■ I will not copy or distribute information from this assessment in any form. ■■ If requested, I agree to complete a further supervised assessment to verify my results of this assessment.” Source: Taken from the SHL Online Assessment Platform Also, explain to the candidate where he or she can get further information on the assessments and on where he or she can try out practice assessments. Tell the candidate what to do and who to notify if they experience problems undertaking the assessment. Having followed the four step-by-step Better Practice principles set out above, organisations will safeguard the integrity of the assessment process and be best positioned to overcome the risks and realise the benefits from online unsupervised testing. Develop Policy and Procedures – Step 5: Develop new policies and procedures for unsupervised test usage in your organisation Having worked through Steps 1 to 4 and assessed the stakes, policies, culture and current processes, organisations should now be better placed to decide how the risks of unsupervised testing can be best managed. As a result, the next step will be to map out and document an updated set of policy and procedures for the organisation, specifically dealing with unsupervised testing online. Probably the most important changes in policy will involve decisions around: ■■ Will a verification process be introduced, or amended if such a process is already in place? If so, what will it look like? ■■ How will verification results be managed? How will discrepancies be handled? ■■ What is the impact on disability and adverse impact policies / procedures? ■■ What is the impact on data storage, confidentiality and access of results? The updated/new policies and procedures should be documented and reviewed to ensure alignment with local employment legislations and legal requirements. Practical guidance on the development of policies based on the Better Practice principles, is provided in Step 6 – ‘How to apply the Better Practice principles’. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 16 Implement the New Process – Step 6: How to apply the Better Practice principles In summary, these guidelines have provided the framework to assist you to: ■■ balance the pros and cons of unsupervised online testing. ■■ evaluate the objectives, stakes and scope as they relate specifically to your organisation and assessment programme(s). ■■ identify where your current processes need to change and the impact of the changes. ■■ update relevant policy and procedures. SHL has also introduced four Better Practice principles to ensure any new process retains integrity and delivers maximum value back to the organisation. The purpose of the Better Practice Application Notes is to provide guidance on the following key areas, which must be considered when implementing an unsupervised online testing programme. They have particular relevance to unsupervised ability testing online: A. Choosing a test provider B. Things to look out for and avoid C. Developing a new policy for unsupervised testing online D.Responsibility towards test takers before, during and after the online test session E. Guidelines on using cut scores F. Cross checking results with other information G.Retesting and verification testing H. Supervising the verification test I. Handling verification discrepancies J. Monitoring test usage and results K. Confidentiality, storage of results and access to test materials These key subject areas will continually evolve and, as such, each step will be updated regularly and be available for individual download from: www.shl.com/. It is also important to note that these guidelines have been updated to reflect the latest knowledge around the management and delivery of psychometric tests based on established best practice. These updates deal only with areas where online testing requires the adaptation or expansion of best practice. Readers unfamiliar with standard best practices for the management and delivery of psychometric tests should also refer to existing best practice guidelines, which can be found in www.shl.com. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 17 Breached tests rarely impact the test provider, but can often impact the clients that use them, for example resulting in mis-hires, fairness claims, etc. How to apply the Better Practice principles A: Choosing a Test Provider When objective assessments (and in particular, ability tests) are delivered online, the impact on the psychometric science behind the test is significant. A new set of requirements emerge and, in effect, the science behind an online test needs to be even better. While the important questions around test reliability and validity remain, organisations need to ask new questions to ensure that the underlying science of the online test is robust. A thorough evaluation of the test provider’s capabilities will be critical, paying particular attention to: ■■ the appropriateness and adequacy of tests and test content. ■■ the availability of tests appropriate for unsupervised online usage that defend against cheating and piracy. ■■ the presence of security measures to protect the integrity of the testing process (e.g. item banks, secure technology, response pattern monitoring, etc.). The checklist below provides some tips and questions to guide you when evaluating a prospective test provider. What should you Look for When Selecting an Unsupervised Online Test Provider? Tools that are cheat and fake-resistant and specifically designed for unsupervised online testing scenarios. Look for: Different test versions generated for each candidate. DO NOT use fixed item tests unsupervised online, or even tests with small numbers of parallel forms as these are highly susceptible to piracy and cheating. Evidence of equating the scores candidates get if they are all getting different tests. The psychometric technology used to equate different tests is Item Response Theory (IRT). Check that the tests for each candidate have been created from a bank of items using IRT to ensure that each candidate gets an equivalent tests and that the scores produced are comparable. Security and data forensic monitoring ideally by an independent agency. This provides employers with reassurance about the security and validity of their testing process. Built-in verification processes which ensure the integrity of unsupervised test results can be confirmed accurately. This is particularly important in high stakes recruitment contexts. Technological reliability. Ensure the test provider can provide evidence that their infrastructure is resilient, has high availability and has the connectivity, concurrency and scalability to handle assessment test loads. Test timing mechanisms that are independent from Internet connections. This ensures candidates using slower machines are not disadvantaged and also guards against the loss of data caused by connection dropouts. Technical information about test reliability, evidence for validity and suitability for online delivery. Psychometric properties for online tests should be evaluated in the same manner as traditional tests. Mechanisms for assessing people with a disability. Ensure that when required, adjustments can be made to the online testing system when testing people with a disability. Also ensure that the on-screen presentation of test items will not disadvantage candidates with colour blindness. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 18 Choosing a Workflow or Software-on-Demand Provider In an online testing context, a Software-on-Demand provider owns and operates a software application that facilitates the delivery and scoring of an assessment. The provider maintains and services the application, and makes the online assessments available to users via a browser or within a client application. If the decision is made to partner with either a Workflow or Software-on-Demand provider, essentially a trade-off decision has been made between a high level of direct control at a high price and a lower level of direct control at a lower price. The need to have skilled IT and application support staff in the organisation is removed, as is the cost associated with the required hardware platforms. However, it is critical that the partner chosen can deliver the level of service that is required. In addition, since the users in the organisation and also the candidates will be accessing the application across the Internet, there are a number of connectivity requirements that must be checked. With a desktop application such as Microsoft Office, as long as the PC is running, the application will be available for use. With a Software-on-Demand provider, there are several things that determine ability to use the application at will. For example, is the application always running and can the application be accessed when needed? Tips when selecting a Software-on-Demand provider… Check the access settings in your organisation Many IT departments in organisations block access to certain Internet sites, block certain types of content and also block certain types of common Internet application such as Macromedia Flash. Speak with your IT department and the Software-onDemand provider prior to signing any agreement, to ensure that connection and use of the necessary application/s is possible. Check the bandwidth settings in your organisation It is important to understand what bandwidth requirements are necessary to use the application and ensure that these are in line with the bandwidth that an organisation has available. Failure to do this is likely to result in poor user experience and a failed implementation. Consider the candidate’s accessibility Remember that many candidates will be accessing an application from what might be considered to be an unusual connection or configuration. It is important to both minimise the bandwidth required and the level of technology required to run the candidate part of the application suitably. Evaluate data protection and back-ups Partnering with a third party Software-on-Demand also means that the data is hosted by someone else and is not under direct control. Determine whether the provider has backup procedures that will protect data to ensure compliance with data protection and privacy regulations around the world. Ensure a high degree of system security With the Internet becoming an increasingly risky environment, ensure that the Software-on-Demand provider has appropriate safeguards in place to stop hackers getting access to data or tampering with an application. This is increasingly important as the stakes of the application and testing rise. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 19 B: Things to Look Out For and Avoid Unsupervised testing online offers rewards for both candidates and organisations. However, because it is a new approach to testing, guidance around, for example, test provider selection and test use is not readily available. The checklist that follows below can help organisations avoid potentially costly mistakes in relation to unsupervised ability testing online. Look out for and avoid: ■■ Fixed item tests online that are promoted as safe or secure forms of testing. Test providers who promote such tests for unsupervised use offer little or no protection over their test content. ■■ Test providers who place fixed item tests online for unsupervised use and claim that no evidence exists that indicates their test content has been breached. If problems were discovered indicating that items or tests were breached, the damage has already been done at the expense of clients. Breached tests rarely impact the test provider, but can often impact the clients that use them, for example resulting in mishires, fairness claims, etc. ■■ Parallel tests that are promoted under the description ‘randomised testing’. This is certainly an improvement on fixed item testing online, but parallel forms of testing still offer limited protection from integrity breaches. ■■ The promotion of ‘unlimited’ testing volumes. This is a good indication that the test provider has no experience of real-world, high volume testing situations. Credible providers understand that the Internet has limits and will provide appropriate guidance on volume testing. ■■ Test providers who claim that no training is required to use their tests. Specialised or technical training on test use and interpretation may be unnecessary, but professional organisations will suggest or provide training to ensure users do not inappropriately use or interpret the results, no matter how simple the outputs are. ■■ Tests that have no associated documentation detailing psychometric properties or effectiveness. A lack of documentation or research proving its effectiveness in workplace settings should raise serious questions about the validity of that psychometric test. ■■ Online unsupervised tests that use another full test as an independent retest. Using two separate tests that provide two separate scores raises many questions about how to equate those scores appropriately, and how to determine which score is more valid for use. Instead, look for tests that provide built-in verification tools that determine the confidence of the original ability score. ■■ Test providers who rely on the prosecution of individuals who steal test content as their primary protection of test integrity. This is a protection for the test provider, not the client using the tests. Once a test has been breached, the integrity of the test has been compromised, regardless of whether someone is prosecuted or not. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 20 C: Developing a New Policy for Unsupervised Testing Online Psychometric tests are powerful tools used by organisations for the selection, development and management of people. Tests can be used to enhance the decisionmaking process throughout the employee lifecycle process. Ability tests are a particularly useful tool for screening and selection because they are the single most effective predictor of potential. In general, occupational tests have consistently been shown to be better predictors of job success than interviews. Although highly predictive, many of the tests in use today were developed for traditional paper and pencil administration. The rapid growth of the Internet for recruitment and selection purposes has placed significant pressure on organisations and test providers to offer their tests online. Deploying tests online opens up a host of new questions and concerns around transferring existing best practice in the use of employment tests. Who Should use Tests? Knowledge and experience are required to use psychometric tests effectively. It is recognised throughout the world that psychometric instruments are potentially dangerous in the wrong hands. Indeed in many countries, only qualified psychologists are allowed to use them. For example, existing best practice in the UK requires that reputable test providers maintain a register of qualified people to whom test material may be supplied. To access such material requires both general training in test use and often familiarisation training for the specific instruments in question. Publishers usually give some recognition to training from other providers. Current guidelines suggest several training levels for users of tests, including Test Administration training and BPS Level A Occupational Testing Guidelines for the use of online tests. Better Practice still recommends that whenever possible, all test users follow existing best practice guidelines. However, new technologies and capabilities of test publishers have opened up possibilities to provide self-guided interpretation tools and outputs that can be safely used by others within an organisation. This becomes particularly relevant with online testing, as it allows tests to be tightly embedded in organisation-wide systems that enable greater access to decision makers. In all cases where traditional tests and psychologically rich outputs are used online, users of these outputs should be fully trained in their use and interpretation. If the selection of online test has been made by a qualified user and there is sufficient guidance and interpretation within the output, it is feasible to allow minimally trained users to safely use psychometric test outputs. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 21 Ability tests are a particularly useful tool for screening and selection because they are the single most effective predictor of potential. Provided below is a checklist to help determine if a test is appropriate for minimally trained users: If you can answer “yes” to all of these questions, the test is likely to be appropriate for minimally trained users Is test selection governed by the supplier and/or a trained user within the organisation? Is the bulk of the test administration automated and managed by the system? Tests for minimally trained users should be automated so that users do not need to set up and manage test administration sessions. If the test is timed, does the system automatically enforce timing? If not, a qualified test user should administer the test. Have test cut-scores already been set by a trained user? If cut-scores are easily changed or set, the test should be restricted to users trained in this area. Do the test outputs provide interpreted narratives and supporting guidance? If the test outputs require psychometric knowledge, interpretation, or other specialised knowledge about ability tests to interpret the results, proper training in the use of the tests should be enforced. If tests are designed for use with minimal training, does the provider offer basic training for use? Regardless of the simplicity of the test and outputs, some training is strongly advised for all users about appropriate use of the results. Organisations seeking to expand the use of tests within the organisation should also refer to existing best practices and other guidelines, such as those provided by the International Test Commission, that provide guidance on the training required for test users of online tests. Case study: Understanding new training requirements for test administrators 1. It could substantially reduce costs to both candidates and the organisation by eliminating the need for candidates to be screened and tested centrally; 2. It improved and expanded the applicant pool by making it easier and faster for more graduates to apply for positions; 3. It provided more substantive and direct involvement of the local branch HR personnel sooner in the selection process, a proven critical step in making the right hiring decisions. To realise the benefits of decentralised testing, the organisation had to involve their branch managers in some parts of the testing process. Candidates would complete the tests online and then go straight into the closest branch for next steps and interviews rather than have central testing sessions at the corporate facility. This presented a new challenge – how to train the staff on the interpretation and use of the test results. Rather than focusing on technical training around the interpretation of traditional test scores, the organisation chose test outputs that provided interpreted results. They also provided structured guidance for users who did not have specialised training. They selected an online test that did not require branch managers to reconcile two test scores when they retested candidates. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 22 The organisation felt that the controls and the structure of the new online testing process eliminated much of the training normally required of test administrators. Branch managers did not need to take over the central test administrator’s role of coordinating testing sessions because the testing process was automated. Instead of training branch managers on how to administer tests, the organisation planned to use the time saved to help managers with: ■■ structured interviewing techniques. ■■ appropriate use and interpretation of test results. ■■ how to verify and confirm online unsupervised test results. The organisation decided that unsupervised, online testing supported at the branch The organisation decided that unsupervised, online testing supported at the branch level represented a good balance between risks and rewards without requiring advanced training for branch managers or HR staff. D: Responsibility Towards Online Test Takers Current best practice specifies that whenever tests are used, testers should be honest and open with candidates about why the test is being used and what will happen to the results in some countries (e.g. the UK), guidance also recommends that candidates should be offered feedback of their results. In general, these practices still apply, but adaptations to these guidelines may be appropriate for online testing due to several factors. Online testing can substantially increase the volume and scale of test use in the recruitment process. It may no longer be reasonable to provide feedback to all candidates who complete the test. Better Practice recommends that organisations should still make feedback available to candidates who proceed to a shortlist, or final testing stage. Organisations may choose not to offer personal feedback to candidates in the screening, or sifting stage. However, it is recommended that individual requests for candidate feedback are honoured whenever reasonable. The automation of online tests often allows test user organisations to explain the purpose, intent and process of testing via standardised online communications. Better Practice for online testing allows for this to replace the traditional test administrator activities and scripts. Depending on the online process, users may need to be trained on specific areas of test administration, such as retest procedures and communications. (a) Before the online test session. Since the test takers are now responsible for their own testing environment, it is important that the online test screens provide suggestions on creating a suitable environment before the test begins. In supervised test sessions, test administrators are normally responsible for communicating information about the tests and the process. They are also responsible for establishing a serious but sympathetic atmosphere. As with current guidelines, providing this structure can help candidates relax, feel confident that the process is fair, and reduce the need for support or complaints as a result of an inappropriate testing environment. Guidelines for establishing the proper test setting are below. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 23 ■■ Online tests often change the feedback process for candidates from facetoface experience to an automated process. ■■ ■■ ■■ ■■ ■■ Online tests should provide for practice tests and example items. The use of practice tests and items should be encouraged, and should not only allow test takers to explore both the test content and style, but also become familiar with the format and technology required to successfully complete the tests. It is also important to consider how supplemental information is communicated online, such as: ■■ how candidates can get help ■■ what to do if there is a technology problem ■■ what to do if disabilities may prevent successful completion of the test ■■ tips and techniques on when/where to take the test Online tests should establish the seriousness of the session. Candidates must be informed about their responsibilities and the stakes involved with the test. Honesty agreements, statements of test taker responsibilities or similar instructions are important to ensure test takers understand the seriousness of the test. Instruction text and test guidance must sufficiently guide test takers to avoid interruptions and distractions. For example, candidates should be reminded to: ■■ turn off mobile phones or other devices that might interrupt them ■■ avoid working in a busy or noisy environment Options should be provided for the candidate to leave the process and return later. Before the test begins, candidates should have the option to delay starting the real test until they are in a comfortable environment. Information should be made available on the use and protection of data and results. Online tests should provide links to policies that explain who will have access to test data and what the test results will be used for. Use of calculators and paper for notes and calculations. The online test instructions should advise candidates on the use of calculators and having paper available for rough calculations. b) After the online test session. The use of online tests may result in changes to how candidates are managed after the testing session is complete. Online tests often change the feedback process for candidates from face-to-face experience to an automated process. This requires special consideration to ensure that the candidate experience remains positive and productive. In general, the better the candidate experience is throughout the entire testing and feedback process, the more likely it is that candidates will perceive the process as fair and effective. Below are some guidelines for managing the post-test process and communications: ■■ Provide a method for candidates to contact the organisation if they wish to request feedback. If tests are used within an automated process, individual feedback may not be available. If feedback is not provided to candidates, or if feedback is automated, candidates should still have the option to request it. This transparency helps to establish confidence in candidates that the entire process is fair and considerate. ■■ Ensure that feedback and responses to candidates are timely. Results of an online test should be provided as soon as is reasonable. In many situations, candidates will perceive the fairness and legitimacy of a recruitment process as much by the overall experience as by the results of a test. Timeliness and appropriateness of feedback on tests is an important part of the candidate experience. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 24 ■■ ■■ Consider how soon automated “rejections” are sent to candidates when tests are used for sifting. In situations where automated decision rules are used based on test scores, consider timing of the candidate feedback, particularly for candidates who are not successful in reaching a cut-off minimum level of ability. If “rejection” notices are sent in real time, candidates may react negatively to the automated decision. Delaying the results 12 to 24 hours can improve candidate reaction to the process and provide time for hiring managers to review decisions, if this is part of the process. Ensure that all test users in the organisation are trained on next steps. The advent of online tests often means other people within the organisation, besides the traditional test administrator, will interface with candidates who have taken tests. In this case, such people must have the basic training and knowledge to answer candidate questions about where and how to get feedback on their results. This may be as simple as ensuring hiring managers know how to contact a qualified user within their organisation to request feedback. E: Guidelines on Using Cut Scores bility tests are a useful method of screening large volumes of applicants fairly, efficiently, and consistently. The use of ability tests often involves a cut score. Cut scores are minimum thresholds of performance on a test that are based on the minimum requirements of a job/position. In situations where an ability test is used as a sole criterion for sifting (screening) purposes, we recommend using a cut score no greater than the 30th percentile (however, due diligence is required to ensure cut scores are set at an appropriate level). Using a single criterion to screen out greater numbers increases the risk of screening out candidates who are suitable for the job, as well as potentially increasing the risks that there will be negative impacts on protected demographic groups. If an organisation desires to screen out a greater number of applicants using a cut score approach, a multi-faceted method is recommended. This involves either setting cut scores based on multiple criteria, such as gross negative disqualifiers, ability, competency questionnaires and/or background and skills qualifications, or creating composite scores from these sources. The former approach is complex because it involves setting a number of cut-scores that together have to produce the requisite overall pass rate. The latter approach is less complex but assumes that one can let high scores on one attribute compensate for low scores on another. Where it is possible to define a criterion-related minimum qualification level for one attribute, a cut-score should be used. In other situations, compensatory composite score models work well and are simpler to manage. If using unsupervised ability tests with a cut score, it is important that the cut score is not adjusted to try to compensate for cheating, unless technical advice is obtained from your test provider. For example, if a cut score is set at the 70th percentile, which results in 45% of applicants passing (instead of the expected 30%), this may be caused by frequent incidents of cheating (assuming the norm is appropriate for the applicants). Instead of increasing the cut score upwards, suitable verification techniques should be used to identify individuals who cheated from the short list of candidates. There may be an inclination to push the cut up (e.g. to the 80th percentile) to attempt to get back to the intended pass rate (i.e. 30% instead of 45% passing). Using a higher cut score has two adverse effects. First, it may increase the incentive to cheat (the higher the barrier, the higher the incentive, other things being equal). This will increase the proportion of people in the shortlist who have cheated, ultimately being counterproductive for the organisation. Second, it is unfair because it excludes candidates who would legitimately meet the original cut score. Cut scores should be set on the basis of job requirements, not just to reduce numbers. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 25 F: Cross Checking Results with Other Information Cut scores should be set on the basis of job requirements, not just to reduce numbers. Consider information from other competency assessments Use results from other assessments that relate to the competencies and/or skills important for performance in the job to evaluate the person’s actual ability level. (E.g. work simulations, assessment centres, etc.) This technique is useful to confirm that an individual has the required level of ability for a specific role independent of a test result without having to administer a follow-up test. Advantages of this method include low cost, flexibility, no requirement to have testing materials/equipment, and an independent evaluation of the ability level using means relevant to the job. Disadvantages include the lack of consistency, more subjectivity in evaluating the actual ability level, increased documentation requirements and the potential for variability in decision-making. This can be a useful alternative to evaluate an individual’s approximate ability level. It can be used in situations where trained test administrators are not available to administer and proctor follow-up tests, or in situations where test materials and equipment are not readily available. In addition, it often fits in well with a normal interviewing process and can reduce negative impacts on the candidate experience. Competencies (as defined in the SHL Universal Competency Framework) related to cognitive ability include: ■■ Presenting & Communicating Information. ■■ Writing & Reporting. ■■ Applying Expertise & Technology. ■■ Analysing. ■■ Learning & Researching. ■■ Creating & Innovating. ■■ Formulating Strategies & Concepts. There are known relationships between ability and measures of competency potential based on personality data in the above areas. Competency ratings from Assessment Centres in the above UCF areas are also known to have relationships with ability. Guidelines for evaluating an individual’s level of ability using other competency assessments: Using simulations that mirror the challenges presented within cognitive ability tests, i.e. providing significant amounts of data and information (balancing strategic and operational information as well as important and distractor information) which need to be analysed within a set timeframe. For example, to simulate numerical and verbal reasoning analysis, presentations and in trays will assess: ■■ The breadth and depth of an individual’s analysis of a complex business situation. ■■ Their ability to challenge current thinking and identify new and innovative ways of approaching business challenges. ■■ Their ability to balance operational with strategic issues and think broadly about business issues and/or set strategic direction for the organisation. ■■ The effectiveness of their communicate style (written or oral) and the strength of ■■ their arguments. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 26 Use information from other sources Results from examinations, qualifications, grades and other attainment tests that are appropriate measures of a person’s verbal and/or numerical ability may help to evaluate the person’s actual ability level. This technique uses other information available about an individual to independently confirm the approximate ability level. It is recommended when a follow-up test is not possible and/or when additional competency assessments are not used. Advantages to this approach include the ease of obtaining information, low cost and convenience. Disadvantages include inconsistency in the measures evaluated, the subjectivity of the evaluations, increased documentation requirements and the difficulty in corroborating the information. Guidelines for evaluating information from other sources include the following: ■■ Do not attempt to directly equate information, such as grades or skills qualifications to ability level. Use them as a general evaluation of potential. ■■ Use the information to help determine the individual’s approximate level of ability in comparison with other candidates. Avoid trying to approximate a specific ability score or level using this information. ■■ Grades or educational qualifications are usually a broad-based measure of capabilities. There are typically moderate positive correlations between ability tests and school or university grades, but they should not be used as a direct replacement. ■■ Other sources should be viewed as corroborating evidence supporting (or denying) a specific ability score. ■■ Use sources of information that can be independently verified. For example, grades or qualifications from institutions can usually be verified using official transcripts. Other information may be harder to confirm. ■■ Look for evidence of training success or job performance from previous employment. This can be corroborative of ability level, because ability levels are often used to predict training success or job performance. Again, try to use information that can be independently verified. ■■ Avoid using certifications or awards where little is known about the purpose or content used to achieve success. These can be difficult to verify, and may not be an appropriate corroboration of ability. Use structured interviewing techniques to probe related competencies As noted above, competencies (as defined in the SHL Universal Competency Framework) related to cognitive ability include: ■■ Presenting & Communicating Information. ■■ Writing & Reporting. ■■ Applying Expertise & Technology. ■■ Analysing. ■■ Learning & Researching. ■■ Creating & Innovating. ■■ Formulating Strategies & Concepts. Structured interviewing around competencies that are known to relate to ability level can provide useful supportive evidence to corroborate ability levels. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 27 Structured interviewing around competencies that are known to relate to ability level can provide useful supportive evidence to corroborate ability levels. Advantages of this approach include: ■■ Practical and easy to integrate into most existing hiring processes, since interviews are a ubiquitous hiring tool. ■■ Structured interviewing is accepted in many organisations, and there is often adequate training amongst existing staff on how to do effective structured interviews. ■■ No additional hurdles for candidates (retests) means minimal disadvantage to the candidate experience. Disadvantages include: ■■ Subjectivity in evaluating approximate ability levels. ■■ Requires interviewers to be trained and listen for specific cues from the interviews that can corroborate ability levels. Guidelines for using structured interview techniques relating these specific competencies are available from SHL. G: Retesting and Verification Testing Current best practice addresses the issue of retesting and whether an applicant may reapply or (be allowed to) retake a test. Online unsupervised ability testing adds another retesting element to the mix – the option for organisations to retest candidates at a future point in the process to verify unsupervised online test results. In the context of retesting an applicant who is reapplying for a position, there is no hard and fast rule on the policy an organisation should adopt. However the organisation should ensure that it applies a consistent policy on retesting for applicants who are reapplying. As with existing best practice guidelines for ability tests, it is still not desirable to allow candidates to be retested regularly unless alternate forms of the test are available. Whenever possible, candidates should be retested with a parallel or alternate form of the test, or ideally with a uniquely generated form such as that available from the SHL Verify ability tests. It is normally reasonable to allow an applicant to be retested where there is evidence that he or she might have under-performed the first time (e.g. due to illness). Personality questionnaires with in-built consistency checks do not require verification testing to confirm results. In the context of retesting for ability verification purposes, it becomes more critical that alternate test forms are available. Verification testing is used to confirm the result of an unsupervised test and typically is administered within a relatively short time-frame compared to situations where candidates are reapplying. Candidates are therefore more likely to remember questions. Organisations that use verification testing should document the procedures around verification testing to ensure that consistency and fairness is maintained. Verification testing policies should include: ■■ Frequency – Consistency of the verification testing model is important to retain fairness and credibility in the testing process. Is everyone retested, a random sample, or only candidates where a specific concern is raised? Depending upon the stakes, local regulations and organisation policy, any of these models may be appropriate. ■■ Process – If only a subset of candidates is retested, how are the candidates selected? Criteria should be clear and consistent. ■■ Supervision – Verification tests should be supervised using existing best practice governing supervised testing. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 28 ■■ ■■ ■■ Score comparisons – How will the organisation compare test scores? If a different test is used for verification, the scores should be equivalent to the original test to ensure comparisons can be made. Acceptance criteria – What are the acceptance criteria for verification test scores? How do test users determine significant differences between scores? If a test does not automatically equate test/retest scores for the user, clear and specific guidance must be provided to inform users when a retest score is statistically different from the original test score. The organisation may wish to seek advice from their test provider on this complex area. Negative verification results – how should the organisation manage a candidate who yields a verification test result that is significantly different to the original test score? If a verification test yields a score that is significantly (statistically) lower than an original test, organisations should provide specific guidance to test users on how to investigate this difference and determine what to do next. Choosing the appropriate Verification Testing Model There are different methods organisations may use to verify the results of an unsupervised test. Four models are presented here that encompass many of the options: 1. No verification testing. 2. Random verification testing – where some candidates are randomly selected for verification procedures. 3. Verification testing is always conducted – across all scenarios and for all candidates. 4. Mixed models – where verification might always be conducted in some scenarios (for example in high stakes recruitment contexts) and not in others. In situations where high candidate volumes result in a large number of short-listed candidates, companies may decide to use the verification test at a later stage, verifying only the results of candidates who are being seriously considered for a final selection decision. Choosing the method most appropriate for an organisation can be a complex task. Examples of each scenario are presented below. This information should be used only as guidance. Organisations should carefully evaluate their needs before making a final decision. 1. No Verification Testing There are two common “no verification testing” models: a) No explicit retest used to verify an unsupervised test score, but other techniques are used to confirm the results; b) No verification step at all. Option b), no verification at all, should only be considered in low stakes situations (e.g. career guidance) where there are no other alternatives available. Typically there will always be some low-cost, simple methods available to verify the result of an unsupervised test even if another test cannot be used. These methods include competency interviews, using other information to rationally evaluate an individual’s ability level, or other similar techniques. Using no verification at all is a risky proposition and may ultimately be counterproductive. Option a) is typically what companies choose as a “no verification testing” model. This option uses some other technique to verify the results of an unsupervised ability test other than a retest. This approach has the advantages of low cost and great flexibility while still providing some protection against the damaging effects of cheating. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 29 “No verification testing” models may be more attractive in certain organisational situations, such as: ■■ No verification testing environment is available for a retest, such as manufacturing environments, field locations, mobile recruiting, or mass hire situations. These types of environments often have logistical challenges that make it difficult or impossible to administer a retest in a controlled environment necessary to provide an effective retest. ■■ Remote or multiple location recruiting that is geographically dispersed. Companies often process candidate applications and screen them centrally through an online process, but send candidates to local offices for the next steps. Often, these local offices are not equipped or trained to process verification tests. Therefore, alternative verification methods are common due to the convenience and flexibility they offer. ■■ Remote screening and candidate qualification followed by a group interview or assessment centre process. If candidates convene in a central location for group assessment centres or interviews, group testing can be cost prohibitive as well as time consuming. Clients may find it more convenient to use alternative verification methods in the interviews or assessment centres that can verify ability test results. 2. Random Verification Testing Random verification testing uses sampling techniques to ‘spot check’ the candidate pool for potential discrepancies in unsupervised testing. The premise is one similar to random security spot checks in airport screening lines; some individuals are selected for a thorough screening while the remaining individuals are screened using minimum techniques. SHL recommends between 25% and 50% of candidates be retested if this is to be a meaningful disincentive to cheating. The advantages of this model are threefold: ■■ The random spot checks help to discourage cheating because candidates know there is a chance they will be selected for verification testing. ■■ Verification test data can be examined periodically to spot cheating trends and address potential issues related to these trends. ■■ It provides a method to verify some short-listed candidate scores directly without substantial time or cost. Disadvantages include: ■■ Cost and time to randomly retest may not be substantially less than other retest models. ■■ Random retest models can be harder to manage for consistency and compliance. ■■ It may impact the candidate experience as some candidates will be retested and others not. ■■ It requires regular review of group data to monitor for potential problems. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 30 Random check verification testing models may be more appropriate for: ■■ ■■ ■■ Companies that use a multi-faceted approach to screening (e.g. use multiple criteria to short-list candidates). This reduces the emphasis on a single ability test criterion as the screening tool. The combined criteria can a) reduce the stakes of the ability test, reducing the desire to cheat among the candidate pool, and b) mitigate the effects of cheating by considering other forms of qualifications. Companies that use assessment centres to qualify short listed candidates for next steps. Assessment centres often include work simulations, competency-based interviews or other structured methods to determine candidate qualifications. The assessment centres can help to verify unsupervised ability results without a verification test for all candidates. Group or mass hire situations where a limited number of computer or testing facilities are available. Random testing can act as a disincentive to candidates to cheat and provide adequate checks in the testing process to monitor for cheating trends or aberrations. 3. Verification Testing Always Conducted (Retest) Administering a verification test to all individuals considered for next steps is the most secure method of verifying ability results obtained from unsupervised tests. This is an ideal approach to testing, as it represents the most secure method of verifying ability scores. It is more important in high stakes situations or situations where ability scores are a key selection criteria. However, where there are a large number of candidates and several subsequent selection stages, companies may decide that it is more efficient to conduct verification testing only for candidates that are being seriously considered in a final selection decision. The advantages of this method are: ■■ Highest level of confidence that the testing process is providing accurate test results for all candidates. ■■ Greatest likelihood of catching potential cheats. ■■ Consistent and objective process for all candidates, establishing fairness and removing potential points of subjectivity from decisions based on test results. ■■ Greater defensibility in situations where results may be challenged. Disadvantages include: ■■ Not always achievable or appropriate due to cost, time, and resource limitations. ■■ Can reduce flexibility in the interviewing and hiring process. ■■ May be difficult to manage and control in remote or distributed organisational models. ■■ Increases candidate testing time and can have disadvantages for the overall candidate experience. This verification model is appropriate for: ■■ Any organisation that requires strict and consistent use of tests and results. ■■ Organisations that use testing for high-stakes situations. The higher the stakes, the more likely bad hiring decisions will have negative impacts on an organisation. ■■ Companies that rely on ability test results as a single selection criterion, or as one of few key criteria. If ability is substantially weighed in a selection decision, the importance of an accurate and secure measure is paramount. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 31 ■■ ■■ Companies that have adequate facilities and capabilities to administer a verification test. If companies have these capabilities, and if there are no obvious barriers such as substantially increased costs, inconvenience, or other barriers, there is little reason not to administer a verification test. The level of confidence and integrity the verification test can provide will often outweigh the disadvantages and risks associated with limited or no verification testing. Companies that do only limited interviews, assessment centres, or other additional assessments on candidates. If there are few other in-depth evaluations of candidates before making hiring decisions, the potential for unqualified candidates getting through the process is greater. It may be easier and more cost-effective for companies in this situation to administer a verification test (and other structured assessments) to mitigate risks caused by the absence of other interviewing activities. 4. Mixed Model Verification Mixed model verification may be appropriate for companies where a single model is not appropriate or cannot be applied consistently. Mixed models blend different verification methods based on situational factors, or differences between the types (and stakes) of selecting for different roles. For example, mixed models can vary the methods of verification used for different roles, or they may vary the points at which verification tests are used, and therefore how many candidates are verification tested. (Please also refer to Application Note F: ‘Cross checking results with other information’.) Advantages of mixed models include: ■■ Highly adaptable to different organisational structures and different recruitment/ hiring models. ■■ Greater flexibility to adapt to changing needs and conditions. ■■ Useful approach to use when balancing costs and resources is important. ■■ Adaptable to changing staff requirements and internal testing expertise/resources. Disadvantages include: ■■ Less consistency and therefore more challenging to monitor effectiveness of the verification procedures. ■■ More requirements for documentation and training to ensure individuals involved with the testing and verification process understand their roles and the process that should be used. ■■ Potentially more difficult to defend to fairness challenges. If companies use different approaches in different situations, it can place a greater burden on the organisation to defend their approach. Mixed models may be appropriate for: ■■ Companies that hire for many different levels and positions. There are different stakes involved for different positions that may warrant different verification models. For example, in hiring key management staff (high stakes) verification may be used in all cases. In contrast, hiring for a high volume call centre may mean lower stakes and a more cost-effective or practical approach might be appropriate. ■■ Companies that use ability testing differently in their evaluation of candidates for positions. For example, some positions may rely on ability as a single criterion or one of few key criteria for decision-making, whereas others may use ability more generally. This can influence the level of verification used for different positions. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 32 Case study: How one company determined the ideal verification method This case study explores how one company decided to use a no retest verification model. The company used a thorough review and analysis of the many different steps in the selection process to help determine the potential risk caused by eliminating the retest. Ultimately the company decided that, due to the many other checks used throughout the interviewing process, a retest would add very limited additional security to an already strict process. A large multinational organisation was using critical reasoning tests as a key part of their corporate leadership programme. The organisation screened thousands of applications down to a few hundred candidates using a consistent but varied mix of screening and sifting criteria. They used the results of a critical reasoning test to determine the smaller group of finalists to proceed to in-depth and lengthy assessment centres and on-site interviews. The organisation had strict and documented criteria that established a minimum threshold of performance on the critical reasoning test. The HR team ultimately determined that using an online, unsupervised test without a mandated retest was in the best interests of the candidates, the programme and the organisation. The decision not to retest was supported by a number of specific criteria in three areas: test design and security, pre-test activities and post-test activities. Test design and security criteria included: ■■ The decision to use a fully randomised test rather than a fixed item or parallel form test to mitigate cheating risks. ■■ The use of test(s) that were as short as possible while still providing appropriate accuracy (reliability). The team felt that the shorter tests mitigated a candidate’s attempt to use assistance (e.g. get someone to help them). The team felt that most candidates who would be tempted to take the test with assistance would think twice about it as it would likely take as much or more time to coordinate the assistance and could easily result in a lower score. ■■ Selection of a testing vendor that monitored content for security breaches, such as the sale or exchange of test items and answers, or other inappropriate or illegal activities. Pre-test criteria included: ■■ The use of the critical reasoning test which was one part of a multi-faceted screening and selection process. The team used a wide range of other criteria, at both the screening and selection stages, to determine eligibility. ■■ The organisation carefully screened the candidates invited to complete the tests well in advance of the invitations. The screenings included a thorough review of qualifications, experience and references. ■■ The candidates were invited to complete the tests online only after an in-depth telephone interview. These interviews were rarely less than 30 minutes and firmly established the organisation’s legitimate interest in the candidate as part of the selection process. ■■ The telephone conversations provided HR staff with the opportunity to explain the purpose and expectations of the online test to each candidate personally. They clearly communicated the candidate’s responsibilities and the organisation’s expectations in advance. ■■ The invitations sent to the candidates emphasised the requirement to complete the test unaided/unassisted. ■■ The tests required candidates to complete an online ‘honesty contract’. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 33 Test administrators are responsible for setting up the verification test and providing instructions to candidates about the process. Post-test criteria included: ■■ Consistency of cut-off scores. As long as candidates scored above a specific cut-off score, they were eligible for further consideration. ■■ Use of the result as one small component of the overall decision. The organisation decided that the ability test result was only a small part of the overall decision criteria. ■■ Candidates were subjected to an in-depth, 3-day assessment centre process that included simulation and scenario-based workshops designed to evaluate reasoning and decision-making capabilities. These workshops provided an effective picture of each candidate’s reasoning abilities and were used to confirm candidates’ test scores. ■■ Trained interviewers that understood the links between the results of the critical reasoning tests and candidate performance in structured competency-based interviews. This allowed them to pick up on inconsistencies between test scores and actual candidate performance in the interviews. ■■ The hiring decisions were not made until candidates were referred to the specific division managers and had completed several other on-site interviews within the division. These managers did not see or use the test scores in their decision-making process. Finally, the organisation clearly communicated the option to retest candidates at any point in the process. The team decided that, if other data consistently indicated the test scores did not represent other criteria used to verify the scores, candidates could be asked to sit another test. Combined, all of these measures built the confidence of the corporate leadership team, the stakeholders within the organisation and the candidates themselves, that the process was extraordinarily thorough and secure, and that an integrated retest process was not warranted. H: Supervising the Verification Test Test administrators are responsible for setting up the verification test and providing instructions to candidates about the process. Well-designed online tests should provide most of the guidance and instructions to candidates on screen. Test administrators are also responsible for candidate authentication, controlling, monitoring, and communicating certain elements of a verification test, including: ■■ Ensuring the testing environment is quiet and free from interruption. ■■ Monitoring test takers for inappropriate behaviours, such as collusion or use of disallowed assistance (e.g. mobile phones). ■■ Knowing how to activate or start the test for candidates. ■■ Providing sufficient guidance to candidates if they have questions about the instructions, example questions, or testing process. ■■ Knowing what to do in the case of test interruption or technical problem with the computer system. ■■ Providing calculators and paper for notes/calculations when appropriate. Some tests allow test takers to use calculators and/or paper for calculations or notes. Provide ready access to these materials for test takers or clearly inform them who to contact to get these materials before they begin the test. ■■ Collecting paper used for notes/calculations after each test taker finishes. It is important to collect and destroy paper used by test takers for notes and calculations to prevent other candidates from using these materials. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 34 Even though verification testing is used to confirm the results of an unsupervised online test completed earlier, companies still hold the responsibility of maintaining the appropriate test environment for the verification tests. Test administrators responsible for verification tests should follow most existing guidelines for test administration. Depending upon the form of verification testing, test administrators may be required to explain details of the test itself and how the results will be used. SHL also recommends that test administrators who are coordinating verification tests, understand the organisation’s policies on the frequency of verification testing and the selection process used. For example, if the organisation uses random selection to test only a subset of candidates using a verification test, candidates selected should be told that their selection is random. As with current best practice, it is important that test administrators have access to a trained and/or qualified professional who understands the technical details and interpretation of the test(s). Depending upon corporate guidelines or local legislation, this qualified user may need to be within the organisation. In other situations, the test publisher or agent may be able to provide this expertise. I: Handling Verification Discrepancies Verification procedures should always be handled sensitively. If the results of any supervised online test are not verified by a verification test or other measure, this does not automatically mean that a candidate has cheated or acted dishonestly. A discrepancy highlighted by the verification procedure should not by itself cause an individual to be eliminated from consideration, or removed from the process. The cause of the discrepancy should be investigated with the individual. Communicating the result of a verification discrepancy to the individual. Before the feedback session begins: ■■ Do not assume the discrepancy is a result of a candidate’s dishonesty (cheating). There can be other reasons for discrepancies between test results. ■■ If the feedback provider did not supervise the test, he or she should ask the test supervisor if any technical or environmental factors during the verification test might have resulted in a person getting an abnormally low score. ■■ Ensure that the feedback provider has access to other candidate information that may be helpful to provide independent confirmation of a candidate’s ability level, such as CV/resumé information, other assessment results, etc. ■■ Ensure that the company’s verification procedures were appropriately followed. ■■ Verify that the feedback session with the candidate can occur in a private location, away from other candidates or distractions. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 35 A discrepancy highlighted by the verification procedure should not by itself cause an individual to be eliminated from consideration, or removed from the process. Initiating the feedback session: ■■ Feedback providers should set the candidate at ease. ■■ Explain to the candidate that there is a significant discrepancy between the candidate’s original ability test results and the results of the verification test. ■■ Explain that the verification tests are a standard part of the testing process and that it is not uncommon for discrepancies to occur. ■■ Explain that the verification tests are used to ensure that the process is fair and transparent and it is not directed at specific individuals. ■■ Explain that there maybe legitimate reasons for large discrepancies and that this feedback session is intended to resolve any questions about this discrepancy. ■■ Explain that the discrepancy, by itself, does not mean the candidate is being eliminated for consideration. ■■ Explain that the feedback session will help the company (test administrator) determine if there were any specific issues that could have influenced the candidate’s poorer performance in the verification test. ■■ Ask the candidate if they had any questions or concerns about the verification process at this time. Once the introduction to the session is conducted, explore the results with the candidate. Explore the result with the individual. A discussion with the individual can provide insight into why the two results may be different. There are several factors that may have influenced an individual’s performance and caused inconsistent test results, including: ■■ the individual’s emotional or physical well being during the tests. ■■ test anxiety, unease, or other conditions. Explore these factors with the individual to determine if they influenced their performance. Specific prompt questions (such as the ones below) can help structure a discussion with the individual. ■■ What was his or her strategy or approach to the verification test regarding speed and accuracy? ■■ Did the candidate understand the test was timed? ■■ Did the candidate focus on answering all the questions, or was he or she more focused on accuracy of the items answered? ■■ How did he or she feel about completing the second (supervised) test compared to the first test? ■■ Was the test taker significantly more nervous during the verification test? ■■ Did something happen during the verification test, such as technical problems, disruptions, etc. that may have impacted the candidate’s performance? ■■ How seriously did he or she take the verification test? ■■ Is there any explanation they can suggest as to why their performance on the verification test was significantly worse than on the original unsupervised test? ■■ Any emotional, physical, environmental or other condition that could have affected their performance? (Note: appropriate caution should be taken when questioning candidates that may involve disclosure of disabilities. Not all jurisdictions allow for this form of questioning. Consult with a local qualified test administrator.) ■■ Did the candidate complete the verification test using appropriate assistance devices, e.g. screen reader for vision disabilities? ■■ Did the candidate experience any illness, condition, or other physical distress while completing the verification test that may have impacted the results? Other techniques for managing feedback with candidates can be found in many occupational testing manuals and guides. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 36 Decide on the most appropriate course of action. If there is sufficient evidence that suggests that legitimate factors may have been involved in a discrepancy, the following two options are open: ■■ If the discussion indicates the unsupervised test scores are likely to be an accurate measure of the person’s true ability, the existing scores can be used as normal. ■■ If the individual can offer no good reason for the discrepancy, administering another ability test in a supervised setting is recommended. The result can be used in place of the unsupervised ability test to determine the person’s ability level. The advantages of using a follow-up test to resolve any discrepancies include consistency of measurement, candidate perception of fairness and objectivity of the measure. Disadvantages include additional time and cost to administer another test, difficulty in comparing multiple test results and the disadvantages to the candidate experience as a result of taking another test. Guidelines for using an additional ability test to resolve any discrepancies include: ■■ Administer the follow-up test in a supervised setting. ■■ This helps to resolve any concerns around environmental factors interfering with a candidate’s performance. ■■ Ideally use a test that is administered in the same format (e.g. computer-based test) as the unsupervised test. ■■ This reduces the risk of test results caused by the administration format. This can make comparisons between results more consistent. ■■ If a different ability test is used for the follow-up test, it is important to understand how to equate the results for comparison purposes. It is essential that any norms used be as close to the original unsupervised ability test’s comparison group as possible. ■■ In all cases, the results of the follow-up test should be used as a replacement for the unsupervised test result. We recommend this method as the ideal way to resolve discrepancies for companies that have the ability to conveniently administer a test. This method is often the most consistent and objective method to resolve discrepancies. The SHL Verify ability tests allow for the administration of another randomised test in a supervised setting. This simplifies the comparisons between test scores because the tests used in both instances are directly comparable and use the same comparison group data. The final decision on how to confirm and use the person’s test results should follow internal policies and guidelines. Organisations should evaluate the risks involved, their own corporate policy/governance, the use of other screening and selection tools, time, cost and other factors. All of these may be important when deciding how to proceed with discrepant results. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 37 J: Monitoring Test Usage and Results Existing best practice requires that the use of tests and other psychometric instruments be continually monitored to ensure continued appropriateness and effectiveness. Online unsupervised testing makes test monitoring even more critical. Since online tests often have content exposed publicly, the risks that the test content will become compromised, and therefore compromise the results seen from the test, are greater. Monitoring of online unsupervised tests should include the normal checks, including: ■■ ensuring that the tests remain relevant to the job. ■■ ensuring that test versions and norms remain up to date. ■■ monitoring of scores at regular intervals. ■■ monitoring by gender, ethnic group and other protected groups to identify any adverse impact. Online testing should additionally include: ■■ monitoring of test response patterns for isolated or wide-scale aberrant or deviant patterns. ■■ monitoring sources and websites for possible selling or posting of test content. Greater vigilance of online test content is required to ensure that negative impacts of a compromised test are caught as early as possible. Test providers should provide monitoring services, or other methods, to protect content from inappropriate use, theft, or distribution. K: Confidentiality, Storage of Results and Access to Test Materials Data and information protection has become more complex with the introduction of Internet-based systems and new regulations governing the storage and use of personal data. However, best practices that currently govern the storage of results and access to testing materials still apply to online tests. Generally, test results should be treated as personal data and stored with due regard to confidentiality. Access should be restricted to those with a need to know and in accordance with what has been agreed with the respondent during administration and feedback. Persons who are untrained should not be allowed access to raw data from tests, but only to clearly described interpretations. Special considerations that organisations may need to consider for online tests include: ■■ Control or restrict access to test scores in online recruitment or talent management systems. ■■ Ensure that data storage systems (internal and vendor systems) allow for the proper deletion of data as governed by local legislation governing personal data. ■■ Ensure that data storage systems provide adequate protection from intentional, or unintentional disclosure, or theft. ■■ Ensure that any data transferred across legislative borders are governed by Safe ■■ Harbour or similar Data Protection Acts. Best practices governing the security and storage of test materials for online tests are different than those governing fixed or paper/pencil tests. Policies should still exist that govern who can buy materials, where they are stored and who has access, but the availability of the tests and results online may expand access to other users inside the organisation that would not normally have had access to test materials in the past. For most online tests today, scoring keys or other sensitive scoring information are no longer available in a distributable form. Therefore, organisations are unlikely to have scoring keys on site that must be protected and kept under lock and key. If scoring keys are available, all existing best practice should apply to the protection of these materials. www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 38 Because of the potential for wider use and access to tests and test materials from online distribution, it is critical that organisations select tests that are developed and designed for online use. If online tests are fixed tests, appropriate due diligence and policy should be developed internally to ensure that test items and content are not inappropriately copied and distributed for unauthorised use. Whenever possible, randomised tests should be used to manage the potential risk of fixed test items being inappropriately copied or disclosed. Most reputable test providers have updated their guidance to allow for the access and use of online tests and materials by minimally trained users, as long as the tests are designed for online use and have been selected and implemented under the advice of a qualified test user. Because of the potential for wider use and access to tests and test materials from online distribution, it is critical that organisations select tests that are developed and designed for online use. This includes selecting tests that provide results or outputs that would be appropriate for minimally trained users to use safely. Conclusion and Summary We believe that unsupervised testing online with the safeguards described does represent a step change in the science and practice of objective assessment. The steps outlined in the Guide mean that unsupervised online assessment is backed up by sound science and good practice. By following the steps in this guide, organisations will have created a roadmap for using unsupervised testing. Our commitment to Better Practice This Guide is not the only way in which we demonstrates its commitment to promoting Better Practice. The purpose of this Guide has been to look at the current state-of-play in terms of what is happening to organisations and their employment processes, the challenges that organisational requirements are presenting to the existing science and practice of assessment, and to describe how practical steps can be taken by organisations to secure the validity of their assessments as they move them increasingly online and unsupervised. But, Better Practice will develop through learning and the sharing of experiences. we therefore provided an email address: [email protected] through which comments on the Guide can be fed back so that improvements can continually be made. Views, case histories as well as suggestions, are encouraged, so that learning and experience can be shared across the assessment community. This Guide can be downloaded from www.ceb.shl.com/betterpractice The application notes in step 6 will continually evolve and be updated regularly - these will also be available for individual download from the same site www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 39 Appendix References i. Bartram, D. & Hambleton, R.K. (Eds) (2006): ‘Computer-Based Testing and the Internet: Issues and Advances’. Chichester, UK: Wiley. ii. Tippins, N.T., Beaty, J.,Drasgow, F., Gibson, W.M., Pearlman, K., Segall, D.O. & Shepherd, W. (2006) ‘Unproctored Internet testing in employment settings’ Personnel Psychology, 59, 189-225. iii.Weiner, J. & Reynolds, D. (2006): ‘Issues in Unproctored Online Testing’ (paper presented at the Association of Test Publishers Conference, Orlando, Florida, February 2006). iv.Burke, E. (2006): ‘Better Practice for Unsupervised Online Assessment’ White Paper. Thames Ditton, UK. SHL Group. v. International Test Commission (2006): ‘International Guidelines on Computerbased and Internet-delivered Testing’. International Journal of Testing, 6, 143-172 Resources and Additional Information The American Psychological Association (APA) - based in Washington, DC, the APA is a scientific and professional organisation that represents psychology in the United States. With 150,000 members, APA is the largest association of psychologists worldwide. For further information, go to www.apa.org The British Psychology Society (BPS) is the representative body for psychologists and psychology in the UK. The BPS aims to encourage the development of psychology as a scientific discipline and an applied profession, to raise standards of training and practice in the application of psychology, to raise public awareness of psychology and increase the influence of psychological practice in the society. For further information, go to www. bps.org.uk. For information on testing go to www.psychtesting.org.uk The Society for Industrial and Organizational Psychology (SIOP) aims to enhance human well-being and performance in organisational and work settings by promoting the science, practice, and teaching of industrial-organisational psychology. For further information, go to: www.siop.org The International Test Commission (ITC) is an association of national psychological associations, test commissions, publishers and other organisations committed to promoting effective testing and assessment policies and to the proper development, evaluation and uses of educational and psychological instruments. Its membership covers most of the Western and Eastern European Countries and North America, as well as some countries in the Middle and Far East, South America and Africa. For further information and access to ITC Guidelines, go to www.intestcom.org Caveon Test Security were the first end-to-end test security services firm to offer protection against cheating and test fraud with detection services to identify breaches, remediation to leverage legal process to halt abuses of sensitive test information, and prevention to secure tests from compromise. For further information, go to: www.caveon.com Contact Us to Learn More [email protected] www.ceb.shl.com www.ceb.shl.com UNSUPERVISED ONLINE ASSESSMENT BEST PRACTICE GUIDE © 2013 SHL, a part of CEB. All rights reserved. UOABPG-102013-UKeng-GLO 40 Registered office: The Pavilion, 1 Atwell Place, Thames Ditton, Surrey, KT7 0NE, UK
© Copyright 2026 Paperzz