QuizMASter –
A Multi-Agent Game-Style
Learning Activity
Mark Dutchuk
Vancouver Island University,
Canada
Khalid Aziz Muhammadi
Fuhua Lin
Presented by:
Edutainment 2009
Government of Alberta, Canada
Athabasca University, Canada
Dr. Fuhua (Oscar) Lin
Banff, Canada
August 9-11, 2009
Some Problems in Game-based Learning
 Okonkwo & Vassileva (2001) showed that while
educational games can be successful at
motivating students to play them, their effect on
actual learning might be small.
 Kirriemuir et al. (2004): among other reasons,
educational games often fail because they are too
simplistic when compared with commercial video
games, too repetitive, and the tasks do not
support progressive understanding.
Athabasca University
EduTainment 2009
2
Our Research Goal
 To study how to to construct effective
educational games with TV-show style for elearning through using the multi-agent
systems (MAS) approach.
Athabasca University
EduTainment 2009
3
Related Work
 MCOE (Multiagent Co-operative Environment) (Giraffa
et al., 1998)
 simulates a lake environment.
 learners learn the effects of pollution & try to control
it.
 Includes
 reactive agents such as fish and plants, and
 cognitive agents including a tutoring agent & a
virtual ‘ecologist’.
Athabasca University
EduTainment 2009
4
Related Work (Contd..)
 REAL (Reflective Agent Learning Environment)
(Bai & Black, 2006) is an agent-based learning
environment provides a framework for simulation
game scenarios, providing
 an Expert agent that contains the knowledge about
the system being simulated
 a Reflective agent to model what the learner
‘knows’ about his or her environment
 a Pedagogical agent compares the Reflective
agent’s knowledge with that of the Expert agent’s
and adjusts it teaching strategy.
Athabasca University
EduTainment 2009
5
Our Approach in QuizMASter
 Multi-Agent System-based educational game that
would help students learn their course material
through friendly competition
 Use Pedagogical Agents to provide feedback &
motivation
 to assess learners’ emotional states through
examining learner’s standing, response timing,
and history, and banter; and
 to provide appropriate feedback to students in
order to motivate them
Athabasca University
EduTainment 2009
6
QuizMASter’s Functional
Requirements
 Similar to a TV game show, where a small group of
contestants compete by answering questions
presented by the game show host.
 Contestants score points by correctly answering
questions before their opponents do.
 Questions are drawn from a learning management
system database and presented to player’s one
question at a time.
 The answer given, the length of time taken to
respond are transmitted back to a central agent.
 Scores will be tallied, and the feedback on a player’s
standing will be provided to motivate the player.
Athabasca University
EduTainment 2009
7
The Architectural Design
 Player Agent
 Host Agent
 Banter Session
 Bonding
 Scoring
Subsystem
Athabasca University
EduTainment 2009
8
Tasks of Agents
 Calculating Contestant Standing:
 Assumption:
 a contestant that is winning is sufficiently engaged, and
 that those in last place are somewhat unhappy with, or
uninterested in their performance in the game.
 Based on such assumptions we use the contestant’s current
standing as a factor when calculating their attitude
 Recording Response Timing and History that then be
analyzed and compared with the other players.
 Examining the interaction between the host and the player
during what we call ‘banter’ sessions to infer a user’s state
from their conversations.
Athabasca University
EduTainment 2009
9
Player Agent
 The responsibilities of the Player agent:
 Assess and maintain the contestant’s emotional
state
 Receive and display questions
 Calculate response timing
 Send Response objects back to the Host agent
Athabasca University
EduTainment 2009
10
The Player Agent
Athabasca University
EduTainment 2009
11
The Host Agent -- Responsibilities
 Present questions
 Provide feedback to the contestants
 Engage in banter sessions with contestants
 Attempt to ‘bond’ with the contestant, by displaying
appropriate emotion.
 Maintain an high level of interest/excitement
 Scorekeeping
Athabasca University
EduTainment 2009
12
The Host Agent --- Banter Session
 Periodically throughout the game (typically after
every 2-3 questions) the Host agent will engage in
‘banter’ with a contestant.
 The contestant is chosen based on:
 The contestant with the lowest ‘emotion’ factor
 The winner after the current ‘round’ of play
 Random selection when no clear data is
available
Athabasca University
EduTainment 2009
13
The Host Agent -- Bonding
 Each Host agent attempts to bond with its
contestant by
 celebrating their success
 ‘sharing the pain’ of an incorrect response.
 This is accomplished by presenting an appropriate
face to the contestant. In an attempt to maintain
a positive attitude, the Host agent will normally
display a ‘happy’ face.
Athabasca University
EduTainment 2009
14
The Host Agent – Scoring
 is responsible for
 processing response objects:
 the answers
 timing
 emotional state information.
 calculating scores
 standings for current game
Athabasca University
EduTainment 2009
15
Host Agent Interface
Athabasca University
EduTainment 2009
16
Conclusions
 Agent autonomy is the principle reason for
choosing MAS as the game control system (Aylett
and Luck, 2000)
 The current version of QuizMASter provides basic
perception and feedback systems to assess a
player’s attitude during game-play and provide
appropriate responses. It has allowed us to
identify and study several implementation issues.
Athabasca University
EduTainment 2009
17
Future Work
 Voice recognition:
 Currently our prototype Host agent provides textonly interactions.
 FreeTTS3 speech synthesis software and Sphinx-44
voice recognition software are being considered to
provide natural language communications between
contestants and the QuizMASter host.
 This would provide a less distracting interface for
contestants, while permitting QuizMASter to
identify keywords directly from the contestant’s
speech.
Athabasca University
EduTainment 2009
18
Future Work (Contd..)
 Emotion: Software is currently available that is
able to detect emotions based on facial
recognition. Software such as this could
supplement the conversation-based perception
subsystem used by the current version
 Implement QuizMASter in Sun’s Wonderland 3D
virtual environment. The Wonderland
environment will improve our system
substantially by adding graphics, animation, and
sound elements.
 Incorporating Adaptive Features into QuizMASter.
Athabasca University
EduTainment 2009
19
MAS for adaptive immersive GBL
Athabasca University
EduTainment 2009
20
Intelligent Educational Systems Research Group (IESRG)
Athabasca University, Canada
http://oscar.athabascau.ca
Thank You!
Athabasca University
EduTainment 2009
21