Protected Values and Heuristics and Biases Approach

1
Protected Values and Heuristics and Biases
Approach Rumen Iliev and Douglas
Medin Northwestern University
Hypothesis Stronger focus of attention, or more
cognitive resources dedicated to a concept
related to PV, could lead to less attention to
peripheral information, thus to less anchoring
affect for people who hold PVs on a particular
issue. Method The same participants from Study
1.a completed the anchoring task after they had
finished with the previous tasks. The five
possible anchors (5, 10, 20, 50, 100) had equal
chance for appearance. Results The mean target
answer was 26.7, SD 17.6, whereas the true
answer to the question is about 43. The Pearson
correlation between the target and the anchor
across all subjects was r .29, plt.05. There were
no difference in the means or standard deviations
between the two groups.
Abstract Protected values, which are values
resisting any trade-offs, have recently entered
the field of decision making and have challenged
some of the main assumptions and theoretical
models in the area. So far, the main efforts in
research of protected values have been directed
toward establishing adequate definition,
identifying their main properties and describing
some of their psychological correlates. In this
paper we suggest an extension of this approach,
focusing on the cognitive interpretation of the
problem. We hypothesize that holding protected
values will be related to the representation of
information and thus will impact performance in
some well known decision making tasks. We
compared people who hold protected values on
abortions with people who do not, and found
different patterns of performance in conjunction
fallacy and anchoring tasks. Participants with
protected values showed higher rates of
conjunction fallacy, but essentially no anchoring
effect. Protected Values For long time the
dominant paradigm in decision making research has
been the utility theory, which relies on
mechanisms of trade-offs to determine the value
of an outcome. However, in the last decade there
was increasing interest in morally motivated
decisions, which proved to be challenging for the
existing utility models. B aron and Spranca
(1997) defined protected values (PVs) as
those that resist trade-offs with other values,
particularly with economic values. Similarly,
Tetlock et al. (1996) raised the question of
existence of sacred values, which later (Tetlock
et al., 2000) were defined as any value that a
moral community implicitly or explicitly treats
as possessing infinite and transcendental
significance that precludes comparisons,
tradeoffs, or indeed any other mingling with
bounded or secular values. Features like
infinite utility and restrictions of any trade
offs could raise reasonable concerns if PVs exist
at all. However, acts of environmental activists,
anti-abortion groups, suicide bombers and
self-immolating monks clearly demonstrates that
PVs may exist in their strongest form. In the
present research we explore the idea that
protected values, being highly important, could
have an impact to the information representation,
thus people holding PVs could perform differently
on some well established decision making tasks,
depending on the content of the scenario. In the
next part we compare the performance of PVs and
no PVs subjects on anchoring and conjunction
fallacy tasks. Our hypothesis is that PVs make
part of the information more salient for the
subject, and thus concentrate cognitive resources
more on some cues of the situation, relatively
ignoring others, and as such people with and
without PVs will have different performance in
relevant cognitive tasks.
Hypothesis PVs could lead to higher salience of
particular common features, thus to higher
representativeness of the target to a particular
group. If so, we could expect higher rates of
conjunction fallacy among people holding PVs, but
only for scenarios relevant to particular
PVs. Method 128 Northwestern University
undergraduates participated for course credit.
First they were given a questionnaire on PVs,
where on of the questions asked about their
opinion on abortions. If they checked this is
unacceptable under any circumstances they were
considered to hold PVs on abortion. In addition
we had one more abortion relevant scenario,
implying that the person described is pro-life,
and one neutral scenario, implying that the
person is an accountant. If a subject judged that
a conjunction is more likely than one of its
part, it was counted as conjunction fallacy.
The responses of the PVa subjects did not
correlate with the anchor, r.06, while the noPVa
responses were significantly correlated, r.37,
plt.001. To test for a difference in the slopes,
we ran regression analyses, regressing target
answer on the anchor and PVa as well as their
interaction. The first model showed three
influential cases, further than 3 standard
deviations from the regression line, and these
cases turned out to be the most extreme responses
to the target, 70(noPVa), 80(noPVa) and 80(PVa).
After deleting these influential cases, we ran
the model again, and found a significant
interaction between PVa and the anchor b-.13,
plt.05). The two correlation coefficients now were
.42 (plt.05) for the noPVa group and -.02 for the
PVa group, suggesting that an anchoring effect
occurred only for subjects without PVa,.
Results 39 of all subjects (31) showed PVs on
abortion (PVa). For the consequent analysis we
combined the two abortion relevant scenarios in
one measure, combined abortion scenarios. The
neutral scenario fallacy showed no relation with
the PVa measure, while the combined fallacy was
higher for people with PVs (F(1,127)3.64,
p.06). We also explored if the combined abortion
fallacies were related to the neutral scenario
and rune 2x2 (PVs, neutral fallacy) ANOVA. The
results showed main effect of the neutral
scenario (F(1,124)43.52, plt.05) , as well as
interaction between PVs and neutral fallacy
(F(1,124)4.68, plt.05). In other words, perhaps
for more normative thinkers or for people who
knew and might have applied probability theory,
having a PVa did not affect whether a combined
abortion conjunction fallacy was made, which is a
plausible expectation. However, less normative
readers revealed a relationship between holding
PVa and occurrence of the fallacy, but only for
the related scenarios.
Fig. 3 Two separate regression slopes for the two
separate groups of subjects. After deleting
three outliers (answers higher than 70) the
correlation between the anchor and the target was
significant only for the no PV group. A
regression analusis s showed significant
interaction between PVs and the anchor (b-.13,
plt.05)
Fig.1 Conjunction fallacies for each of the three
scenarios.
Summary and conclusion The goal of this research
was to explore the hypothesis that strong moral
positions, in this case PVs, could influence the
representation of information and the performance
in decision making tasks. Our findings show that
PV did influence performance, depending on the
relevance of the task information to persons PVs.
However, there is no unequivocal conclusions if
PVs heart or help normative decision making. In
the case with conjunction fallacy, PVs were
related with less normative answers for the
relevant scenarios, while in the anchoring task,
people with PVs showed no anchoring. In short,
this experiment demonstrates the possibility for
exploration of the link between moral stands and
cognitive processes and application of heuristics
and biases paradigm to ethical research.
Study 1.a Conjunction fallacy Conjunction fallacy
is a logical error which occurs when the
probability of an event happening ( is judged to
be lower than the probability that a single
property that is part of the event happens. A
conjunction fallacy would be to judge a person to
be French as more likely than the probability
that he is European. From a logical point of view
it is an error since if AB, P(B) P(A) in
other words, the highest possible value that P(B)
could have is the probability of the inclusive
set A. For this task we adjusted two of the
original TverskyKahneman (1983) scenarios. The
abortion relevant one was Linda is 31 years old,
single, outspoken, and very bright. She majored
in philosophy. As a student, she was deeply
concerned with issues of discrimination and
social justice, and also participated in
anti-nuclear demonstrations. And the subjects
had to estimate the probability of two statements
about Linda. A. Linda is a bank teller. B. Linda
is a bank teller and pro-choice.
Fig. 2 Combined conjunction fallacy as function
of PVs and neutral fallacy. PV did matter only
for people who failed the neutral scenario.
Study 1.b Anchoring For this task we used the
basic anchoring paradigm (Wilson et al., 1996),
according which the answer to a target question
could be influenced by an anchor of non-related
information. For example, we used the following
scenario, where the number of years is the anchor
and the percentage of abortions is the target
In recent research, published in Netherlands, it
was estimated that the percentage of women who
have had an abortion in Europe almost doubled in
the last 20 years. What is your best guess for
this percentage in the USA nowadays? Where we
varied the number of years from 5 to 100.
Acknowledgements We are very grateful to Dan
Bartels, Will Bennis, Sonya Suchdeva and Sara
Unsworth for their valuable comments, as well as
Brittany Koscher for part of the data collection.