You are on page 1of 16

On the reality of Cognitive

Illusions
• Cognitive illusions arise from interaction of perceived
reality with assumptions about the world (prior
knowledge), leading to “unconscious
inferences”. Cognitive illusions rely on stored
knowledge about the world (depth, rabbits, women)
and are also under some degree of conscious control
(we can generally reverse the perception at will).
• The way you look at an object can affect how you see
it. Sometimes there are two images in the same
picture, but you can only see one at a time so your
brain chooses one (when it deals with too much
information).
Cognitive illusions

• Instead of demonstrating a physiological base


they interact with different levels of perceptual
processing, in-built assumptions or ‘knowledge’
are misdirected. Cognitive illusions are commonly
divided into ambiguous illusions, distorting
illusions, paradox illusions, or fiction illusions.
They often exploit the predictive hypotheses of
early visual processing. Stereograms are based on
a cognitive visual illusion.
• Ambiguous illusions are pictures or objects that offer
significant changes in appearance. Perception will ‘switch’
between the alternates as they are considered in turn as
available data does not confirm a single view. The Necker
cube is a well known example, the motion parallax due to
movement is being misinterpreted, even in the face of
other sensory data. Another popular is the Rubin vase.
• Paradox illusions offer objects that are paradoxical or
impossible, such as the Penrose triangle or impossible
staircases seen, for example, in the work of M. C. Escher.
The impossible triangle is an illusion dependent on a
cognitive misunderstanding that adjacent edges must join.
They occur as a byproduct of perceptual learning.
• Distorting illusions are the most common, these illusions
offer distortions of size, length, or curvature. They were
simple to discover and are easily repeatable. Many are
physiological illusions, such as the Café wall illusion which
exploits the early visual system encoding for edges.
Other distortions, such as converging line illusions, are
more difficult to place as physiological or cognitive as the
depth-cue challenges they offer are not easily placed. All
pictures that have perspective cues are in effect illusions.
Visual judgments as to size are controlled by perspective or
other depth-cues and can easily be wrongly set.
• Fiction illusions are the perception of objects that are
genuinely not there to all but a single observer, such as
those induced by schizophrenia or hallucinogenic drugs.
Base-rate neglect
Gigerenzer’s critique Authors’ response
1. Focuses on a particular design in which 1. The base-rate neglect describes
base –rate information is explicitly situations in which a base rate that is
provided and experimentally known to the subject, at least
manipulated. approximately , is ignored or significantly
underweighted. We tested the
hypothesis in several experimental
paradigms.
2. If one lets the subjects do the random 2. Even in Gigerenzer’s own study, all six
drawing base-rate neglect disappears. informative descriptions deviated from
the Bayesian solution in the direction
predicted by representativeness; the
deviations ranged from 6.6% to 15.5%
(observed substantially larger deviations
in the same design). It appears that
Gigerenzer’s announcement about the
disappearance of base-rate neglect is
premature.
Base-rate neglect
Gigerenzer’s critique Authors’ response
3. In many natural environments 3. Contrary to Gigerenzer's unqualified
frequencies must be sequentially learned claim, the replacement of subjective
through experience and suggests that this probability judgments by estimates of
process allows people to adopt a more relative frequency and the introduction of
effective algorithm for assessing posterior sequential random sampling do not
probability. provide a panacea against base-rate
neglect.

4. Gigerenzer‘dismisses the results of one 4. Unaccountably, he fails to mention that


study involving a particular case (Tom W.) identical results were obtained in a more
on the grounds that our subjects were not extensive study, reported in the same
given reason to believe that the target article, in which the instructions explicitly
vignette had been randomly sampled referred to random sampling.
(Gigerenzer, 1991, p. 96).
Base-rate neglect
Gigerenzer’s critique Authors’ response
5. The outcome-ranking paradigm is 5. First, it is evident that
especially relevant to Gigerenzer's subjects sometimes use explicitly
complaint that we have not provided mentioned base-rate information to a
formal definitions of representativeness much greater extent than they did in our
or availability and that these heuristics original engineer-lawyer study, though
are "largely undefined concepts and can generally less than required by Bayes'
post hoc be used to explain almost rule.
everything" (1991, p. 102). Second, the use of repeated random
sampling is not sufficient to eliminate
base-rate neglect, contrary to
Gigerenzer's claim.
Finally, the most direct evidence for the
role of representativeness in intuitive
prediction, obtained in the outcome-
ranking paradigm, has not been
challenged.
Conjunction Errors-a) The normative issue
Gigerenzer’s critique Authors’ response
1. Imagine a young woman, named Linda, 1. Although Gigerenzer invokes the
who resembles a feminist, but not a bank meaninglessness argument with great
teller. You are asked to consider which conviction, his position on the issue is
of two hypotheses is more likely: (a) Linda problematic.
is a bank teller or (b) Linda is a bank teller On the one hand, he surely does not
who is active in the feminist movement. regard statements of subjective
Gigerenzer insists that there is nothing probability as meaningless; he has even
wrong with the statement that (b) is more collected such judgments from subjects.
probable than (a). He defends this view On the other hand, he invokes the
on the ground that for a frequentist this argument that subjective probabilities are
proposition is meaningless and argues meaningless to deny that these
that "it would be foolish to label these judgments are subject to any normative
judgments 'fallacies'" (1991, p. 95). standards.
Conjunction Errors-a) The normative issue
Gigerenzer’s critique Authors’ response
2. In support of his agnostic position, 2. Normative agnosticism is particularly
Gigerenzer cites von Mises's (1928/1957) inappropriate in the case of the
statement that We can say nothing about conjunction rule.
the probability of death of an individual First, the application of this rule does not
even if we know his condition of life and require numerical estimates, only an
health in detail. The phrase "probability of ordinal judgment of which of two events
death," when it refers to a single person, is more probable.
has no meaning at all for us (p. II). Second, the normative basis for the
conjunction rule is essentially logical: If
the conjunction A & B is true then A must
also be true, but the converse
does not hold.
Conjunction Errors-b) The descriptive issue
Gigerenzer’s critique Authors’ response
1. Gigerenzer's major empirical claim is 1. This finding is a counter-example to
that violations of the conjunction rule are Gigerenzer's often repeated claim that
confined to subjective probabilities and conjunction errors disappear in
that they do not arise in judgments of judgments of frequency, but we have
frequencies. found no mention of it in his writings.
2. Gigerenzer has essentially ignored our 2. Contrary to Gigerenzer's position, the
discovery of the effect of frequency and results demonstrate a violation of the
our analysis of extensional cues. conjunction rule in a frequency
Gigerenzer concludes that "the formulation. The conjunction rule is
conceptual distinction between single applied in direct comparisons, but not in
events and frequency representations is between-subjects experiments, indicates
sufficiently powerful to make this that the key variable that controls
allegedly-stable cognitive illusion adherence to the conjunction rule is not
disappear" (1993,p. 294). the contrast between single events
and frequencies, but the opportunity to
detect the relation of set inclusion.
Conjunction Errors-c) The methodological issue
Gigerenzer’s critique Authors’ response

1. Gigerenzer appears to deny the 1. . In our view, this is hardly more reasonable
relevance of the between-subjects design on than the claim that a
the ground that no individual subject can be randomized between-subject design cannot
said to have committed an error demonstrate that one drug is more effective
than another because no individual subject has
experienced the effects of both drugs.

2. The within-subject design, in which critical 2. The two designs address different questions,
items are presented in immediate succession, especially in cases of conflict between a
provides subjects with judgmental heuristic (e.g., representativeness)
information that is not available to subjects in and a compelling formal principle (e.g., the
a between-subjects design. conjunction rule). Thus, the between-subjects
First, it often reveals the intent of the design (indirect test) is appropriate when we
researcher, by drawing attention to the wish to understand "pure" heuristic reasoning;
independent variable that is manipulated. the within-subject design (direct test) is
Second, the subject has a chance to detect and appropriate when
correct errors and inconsistencies in the we wish to understand how conflicts between
responses to different items. rules and heuristics are resolved
Overconfidence
Gigerenzer’s critique Authors’ response
1. Consistent with his agnostic normative 1. . This argument overlooks the fact that in
stance, Gigerenzer argues that overconfidence most experiments the subjects were explicitly
should not be viewed as a bias because instructed to match their stated confidence to
judgments of confidence are meaningless to a their expected accuracy. The presence of
frequentist. overconfidence therefore indicates that the
subjects committed at least one of the
following errors: (a) overly optimistic
expectation or (b) a failure to use the scale as
instructed.

2. This account of overconfidence, which 2. Random sampling of items is not sufficient


draws on the theory of probabilistic mental to eliminate overconfidence. Additional
models (Gigerenzer, Hoffrage, & support for this conclusion comes from
Kleinbolting, 1991), encounters both observations of overconfidence in the
conceptual and empirical difficulties. prediction of natural events (e.g., economic
recessions, medical diagnoses, bridge
tournaments), where biased selection of items
is not an issue.
Overconfidence
Gigerenzer’s critique Authors’ response
3. Gigerenzer portrays the discrepancy 3. But he is wrong. On the contrary, we drew a
between individual and aggregate assessments distinction between two modes of judgment
as incompatible with our theoretical position. under uncertainty, which we labeled the inside
and the outside views (Kahneman & Tversky,
1979, 1982b; Kahneman & Lovallo, 1993). In
the outside view (or frequentistic approach)
the case at hand is treated as an instance of a
broader class of similar cases, for which the
frequencies of outcomes are known or can be
estimated. In the inside view (or single-case
approach) predictions are based on specific
scenarios and impressions of the particular
case.
4. Gigerenzer appears unwilling to 4. We have neither ignored nor blurred the
apply normative criteria to judgments of single distinction between judgments of single and of
and of repeated events. repeated events. Our disagreement here is
normative, not descriptive. We believe that
subjective probability judgments
should be calibrated .
Discussion
• Gigerenzer's critique employs a highly unusual strategy.
First, it attributes to us assumptions that we never made
(e.g., that judgmental heuristics are independent of
content and context or that judgments of probability and
of frequency yield identical results).
• The adoption of an "outside view" that brings to bear the
statistics of past cases can sometimes improve the
accuracy of judgment concerning a single case
(Kahneman & Lovallo, 1993; Kahneman & Tversky,1979).
• The frequency formulation sometimes makes available
strong extensional cues that subjects can use to avoid
conjunction errors in a within-subject design.
• There are substantial biases in judgments of frequency,
often the same biases that affect judgments of
probability (Tversky & Kahneman,1983).
Conclusion
• The view that "both single-case and frequency judgments
are explained by learned frequencies (probability
cues),albeit by frequencies that relate to different
reference classes“ (Gigerenzer, 1991, p. 106) appears far
too restrictive for a general treatment of judgment under
uncertainty.
• First, this treatment does not apply to events that are
unique for the individual and therefore excludes some of
the most important evidential and decision problems in
people's lives.
• Second, it ignores the role of similarity, analogy,
association, and causality. There is far more to inductive
reasoning and judgment under uncertainty than the
retrieval of learned frequencies.