Giving it all Away

 Cognitive Modules
› Background
 Wason Selection Task
› Purpose
› Puzzles vs Social Contract problems
 Fiddick & Erlich’s Paper
› Introduction
› Methods
› Results
› Discussion

Our minds consist
primarily of “a
constellation of
specialized mechanisms
that have domain-specific
procedures, operate over
domain-specific
representations, or both”
- Cosmides and Tooby
(1994), p. 94

People struggle to identify what information is
necessary in order to test the truth of a logicalreasoning problem.
› Wason Selection Task is used to examine this issue.

Typical experiment: presents a rule and asks
subjects how to find out if the rule is violated.
› Abstract problems: difficult to answer correctly
› Social contract problems: more likely to be answered
correctly

If a card has a D on one side, it has a 3 on the other side.
What card(s) should you flip over to determine if
the rule is true?

Correct answer: D and 7.
 Seeing reverse of 3 can confirm rule but won’t disprove it.

If you borrow my car, you must fill up the gas tank.
What card(s) should you flip over to determine if
the rule is true?

Correct answer: borrowed car and empty gas tank.
 People reason correctly when confronted with social
contract problem.

Cosmides’ study showed elevated levels of
performance on cheater detection tasks (1989)
›

Suggests humans have cheater-detector
mechanisms
Detecting altruism ≠ tracking cooperation
›
›
›
Cooperator accepts benefit and pays cost
Altruist pays cost without accepting benefits
Cheater accepts benefits without paying cost
 Different ways of maintaining
cooperation
with cheaters and cooperators depending on
if rewards or punishment used
› Punishing lack of cooperation more effective
› Generous behavior usually unrewarded
› Supports idea that mechanisms to detect
cheaters will be more useful in maintaining
cooperation



Studies seem to support that people are better at
detecting cheaters
Some researchers challenge idea that people are
better at cheater-detection; believe people should
also have mechanisms to detect altruists too.
Other studies have shown people have ability to
detect altruists (Brown & Moore, 2000).
› Enhanced altruism detection may be a way to detect
people who are “fake” altruists.

Altruism-detection tasks in multiple studies
contain embedded answers.
› Ex. “You suspect that Big Kiku will be altruistic and
give food even if the man does not get a tattoo.
(Evans & Chang, 1998)

Interested in whether enhanced altruism
detection is a way to detect “fake” altruists.
› If true, altruism detection would be govern by same
mechanism as cheater detection.
› Compared altruist-detection to cheater-detection
tasks to see if there was an association.
Subjects performed better on altruist-detection
tasks despite absence embedded answers.
 Cheater-detection task confounded with
embedded answers.
 Wording of cheater-detection scenarios may
have affected subjects’ answers.


Wanted to address confounds of previous
studies
› Are embedded cues why subjects performance
better on some altruist-detection task?

Questioned existence of altruist-detection
mechanism.

Experiment #1: Answers embedded in questions
presented potential confound
› Used (non-)embedded answers to test whether
embedded answers were a confound, which would
undermine support for cognitive modules for cheater
detection

Experiment #2: revised published altruistdetection problems to remove embedded
answers
› Results indicated embedded answers are a confound
for altruism detection

Experiment #3: based on findings by Oda et al.
› Tested whether altruism detection is a form of
cheater detection or independent of cheating module
› Methodological issues present possible confounds
 May not be a special altruism detection module
Participants
 Materials

› Booklet with 4 selection tasks
 Weather, Hare Mantra, abstract, social contract
› 2 versions: embedded & non-embedded answer

Procedure
“The results suggest that embedding the answer
within the selection task scenario can
significantly alter performance on the task, at
least when the scenario does not involve cheater
detection.”
Embedding answer  improves performance on
tasks that do not try to detect cheaters
Researchers removed embedded text to see
effect on altruism detection ability
 Participants
 Materials

› Booklet with 3 altruism detection tasks
 Blood donation, altruist cassava root, generous uncle

Procedure
NSS
NSS
SS
“As predicted, removing the embedded solutions
from these altruist-detection problems did have a
significant influence on performance.”
Fiddick & Erlich argue that removing embedded
solutions prevented subjects from identifying
altruists

Did removing embedded solutions prevent altruist
detection?
 Results were statistically significant after pooling data

Results of Oda et al.
› Tested whether altruism detection is a form of
cheater detection or independent of cheaterdetection module
› Argued for separate cheater/altruist detection
mechanisms

Fiddick & Erlich: attempted to replicate results
with a non-confounded cheater-detection
scenario

Cheater-detection booklet
› Sticker task

Altruist-detection booklet
› Volunteer task

Two groups of participants; one received
cheater-detection booklet first and the other
received the altruist-detection booklet first

Participants performed significantly better on
the cheater-detection task (58.5% correct) than
on the altruist-detection task (20.0% correct)
› No correlation between performance (r = -0.047)
› When cheater detection task was first, r = +0.472
› When altruist detection task was first, r = -0.472

Why should cheater detection prime altruist
detection?
 Embedded solutions do confound results (Exp
1 & 2).
 Elimination of confounds in exp 2 did not
completely reduce altruist-detection levels.
› Non-standard instructions may affect subject
performance.
› Categorization task (altruist-detection) vs. rule
violations (cheater-detection)
 Exp 3 also suggests that altruist-detection
may prime cheater-detection
› Challenges findings of Oda et al. study
› Rule-following methodology of Oda et al. study
may reduce performance on cheater-detection
tasks.
 Conclude lack of evidence supporting
existence of an altruist detection mechanism.
 Many social contract theory (SCT) studies
confounded by having embedded answers.