5/19/2024

DEPARTMENT OF PSYCHOLOGY
Thinking and problem solving

Cognitive Psychology • 8.1 Introduction

Ch. 8 Thinking and problem solving • 8.2 Problem-solving
• 8.3 Creativity
• 8.4 Analogy
• 8.5 Reasoning
• 8.6 Decision-making
• 8.7 Dual-process theories
May 17th, 2024 • Summary
Università di Milano-Bicocca Cognitive Psychology

1 2

Introduction Introduction

• Thinking is essential in everyday life – people • How Psychology study thinking?

constantly face decisions, some big, some small. • They test whether human thinking follows
• the laws of logic (using research on reasoning)
• A little bit of history • the laws of probability (using research on decision-making).
• If the answer is yes => humans behaviour would be
• Aristotle (4th century BC) developed the field of logic to optimal
understand the proper way of thinking
• Key question: Are human rational?
• Boole (1854) developed ‘Boolean logic’
• Procedure:
• First solve a problem using logic or probability theory, which
provides the correct answer
• Blaise Pascal and Pierre de Fermat (1654 ) worked on • Second test if participants not trained in logic or probability
probability theory provide the same answer.

Cognitive Psychology Cognitive Psychology

3 4 1
 5/19/2024

Introduction Thinking and problem solving

• How Psychology study thinking?

• Other approaches:
• problem-solving is interested in the way participants generate • 8.1 Introduction
solutions, which typically consist in a sequence of actions.
• creativity studies the way people come up with novel products • 8.2 Problem-solving
and ideas.
• analogy-making is concerned with the extent to which • 8.3 Creativity
knowledge acquired to solve a certain type of problem can be
used to solve other types of problems. • 8.4 Analogy
• 8.5 Reasoning
• Approaches differ
• in the kind of tasks they employ • 8.6 Decision-making
• in the theoretical assumptions they make.
• in the type of methods they use • 8.7 Dual-process theories

Cognitive Psychology Cognitive Psychology

5 6

Problem solving

• Aim: understand the way people solve problems.

• These problems can be

• simple puzzles (the Tower of Hanoi)
• More complex problems, for example
• best move in a chess position
• diagnosing a patient Figure 8.1 The Tower of Hanoi problem. This problem involves moving the stack on disks
• developing a new theory in science on the first peg to the third peg, with the following constraints: (a) only one disk can be
• Etc… moved at a time; (b) only the upper disk from one of the stacks can be moved (c) larger
disks cannot be put on top of smaller disks.

Cognitive Psychology

7 8 2
 5/19/2024

Problem solving
Behaviourist approaches to problem-solving
• First studies on problem solving: behaviourism (USA
from about 1910 to 1960)
• Explanation: stimulus–response links.
• Participants to experiment: animals (assumption
that could be extrapolated to humans)
• Findings (p.e. Thorndike, 1898):
• solution by trial and error Figure 8.2 Thorndike’s box (Thorndike, 1898).
• Performance improves through reproducing learned
responses.
https://www.youtube.com/watch?v=Vk6H7Ukp6To

Cognitive Psychology

9 10

Problem solving Problem solving

Behaviourist approaches to problem-solving The Gestalt approach to problem-solving
• Trial and error explanation proposed by • Gestalt psychologists (1915 – 1945)
• Some studies on problem-solving.
behaviourism is reasonable with animals and in • They hypothesize two types of problem solving
some cases with humans. • reproductive problem-solving, studies learning during problem solving
• productive problem-solving, studies how people solve new problems.
• BUT there are many instances of problem-solving • They study productive problem solving
where humans find a solution without a lengthy • Key hypothesis: not just creating associations – as argued by
period of trial and error. behaviourists – but also involve restructuring (changing the
structure of either what is the problem at hand)
• Change tends to be sudden and characterised by a feeling of insight and
an ‘Aha!’ experience.
• https://www.youtube.com/watch?v=FwDhYUlbxiQ
• Secondary hypothesis: Previous knowledge often hinders
problem-solving, because it makes restructuring and insight
harder.

Cognitive Psychology Cognitive Psychology

11 12 3
 5/19/2024

Problem solving Problem solving

The Gestalt approach to problem-solving The Gestalt approach to problem-solving
• Example of problems: Duncker’s (1945) tumour • The solution is to send low-intensity rays from
problem several different directions so that they converge on
• A patient has a malignant tumour that cannot be the tumour, destroying it.
removed by surgery. • It is a difficult problem and less than 10 per cent of
• It is necessary to destroy the tumour by radiation. the participants can solve it without additional
• High-density radiation destroys not only the tumour but
also healthy tissue surrounding it
information.
• Low-density radiation is too weak to destroy the tumour.
• How can the tumour be destroyed whilst sparing the
healthy tissue?

Cognitive Psychology Cognitive Psychology

13 14

Problem solving
The Gestalt approach to problem-solving
• Example of problems: Maier’s (1931) two-string
problem
• participants have to tie together two strings that are
hanging from the ceiling.
• the two strings are too far apart to be simultaneously
held by the two hands.
• Various objects are in the room

Figure 8.3 Maier 2-string problem

Cognitive Psychology

15 16 4
 5/19/2024

Problem solving Problem solving

The Gestalt approach to problem-solving The Gestalt approach to problem-solving
• Solution by insight • Insight is connected to subjective ‘Aha!’ experience
• Main characteristics subjective ‘Aha!’ experience
• tie the pair of pliers to one of the strings and then to swing it. • Suddenness
• One then holds the static string, catching the other string when • Obviousness
• confidence that the answer is correct.
it swings close enough to grasp it.
• Support for insight is inconsistent. Probably no sharp divide between
• 40% found the solution spontaneously, additional 38% solving a problem through analysis and through insight.
• ‘Aha!’ experiences are more common with insight problems than with non-insight
with a hint (such as brushing against a string) problems
• BUT successfully solved insight problems do not always lead to such an
• Gestalt psychologists: hints help to restructure the experience, while non-insight problems sometimes do
• AND incorrect solutions often lead to ‘Aha!’ experiences, in particular when
problem => insight problems are solved quickly (Danek & Wiley, 2017).
• verbal protocols showed that solving insight problems different strategies are
• Illustrates the notion of functional fixedness – the used (Fleck and Weisberg, 2013)
• Restucturing
negative influence of knowledge (pliers are perceived • application of previous knowledge
as a tool for bending wire) • various heuristics

Cognitive Psychology Cognitive Psychology

17 18

Problem solving
The Gestalt approach to problem-solving
• Mental sets Phenomenon (Einstellung effects) - an idea
that comes immediately to mind during a familiar
problem blocks alternative solutions from being taken
into account.

• Finding: Luchins (1942)

• Participants were asked to imagine that they have three
jugs to measure different amounts of water.
• The task was to obtain a certain quantity of water using Figure 8.4 Luchins water jug problem
these three jugs.
• The capacity of the three jugs was different for each
problem.
Cognitive Psychology

19 20 5
 5/19/2024

Problem solving Problem solving

The Gestalt approach to problem-solving The Gestalt approach to problem-solving
• Participants were first trained on five problems that had the same solution as the • Magicians often create mental sets , as shown by an experiment by
first problem. They learnt the solution quickly. Thomas and Didierjean (2016).
• They then received two problems that had two possible solutions: • A brown-backed playing card was surrounded by six red-backed playing
• same complex solution of the first five problems cards all presented face down.
• a shorter solution.
• The experimenter – a magician – then explained
• Only 20 per cent of the participants spotted the simple solutions;
• Manipulation group: he could influence their choice by suggestion
• Finally, participants were presented with an eighth problem, which had only one • Control group: no explanation.
novel solution.
• Procedure: choose one of six red-backed cards.
• Nearly 70% participants could not solve this problem.
• Failure was not due to the simple solution being hard to find in the last three • When the brown-backed card was turned over by the experimenter, it
problems was the same as the card they had selected.
• Participants in a control condition, where the first five problems were solved using different • Participants were then invited to explain the trick.
methods, had no difficulty finding the simple solution for these three problems. • Control group: 83% guessed trick (all the cards are the same)
• The experimental group failed to find it because the similarity of the final • Manipulation group: 87% did not gues correctly, even after being told that the
problem to the previous ones brought the usual (but now inappropriate) method experimenter in fact did not use suggestion.
to mind, blinding them to alternatives.
• Conclusion: mental set effects can be very strong, and that a single
• Presence of such mental sets shows that people have a strong tendency to use exposure to a wrong explanation can block correct and simpler solutions.
repetitive thinking.

Cognitive Psychology Cognitive Psychology

21 22

Problem solving Problem solving

Problem space theory Problem spaces, states, and search
• Alternative proposal: Problem space theory • Problem space theory (Newell & Simon, 1956,
• Background: Herbert Simon’s theory of bounded 1972)
rationality (Simon, 1957; Gobet & Lane, 2012). • problem-solving is a search through a problem space.
• humans cannot find the best solutions to most problems • The initial state is the starting point of a problem and
because of the limitations of their cognitive system (e.g. includes the information given at the beginning.
limited STM capacity). • The goal state is the desired end state.
• They have to content themselves with highly selective • The operators consist of the set of legal (permissible)
search and choose options that are good enough but not operations or moves that can be performed.
necessarily optimal. • To find the solution to a problem, one must link the initial
• rely on heuristics: rules of thumb and shortcuts that are efficient, state to the goal state by moving through a number of
cheap computationally and work most of the time, but that can states, which are created by the application of operators.
fail every so often.
• In many problems, this can be likened to searching
• Thus, people are satisficers, not maximisers. through a maze.
Cognitive Psychology Cognitive Psychology

23 24 6
 5/19/2024

Problem solving Problem solving

Problem spaces, states, and search Problem spaces, states, and search
• The internal problem space usually includes
• Two types of problem spaces • representation of the start state
• External problem space, which can be found through task • Representation of the goal state
• allowed operators
analysis, contains all the possible states in a given • previous domain-specific knowledge (‘strong heuristics’)
problem. • domain-general knowledge (‘weak heuristics’).
• relatively small for simple problems, but they can be very large • The internal problem space is progressively augmented as search
(10120 possible chess games) progresses and the problem-solver learns more about the problem.
• Analysing external problem spaces is useful because they offer • internal problem space is usually much smaller than the external
problem space
normative measures about optimal solutions and thus provide
• Example of strong heuristic: analysing data from a psychology
information about the difficulty of a problem. experiment look for a linear relationship between two variables.
• The internal problem space is the internal representation • Examples of weak heuristics:
that people construct of the problem • trial and error
• means–end analysis (reduce the difference between the start state and the goal
state).

Cognitive Psychology Cognitive Psychology

25 26

Problem solving
Problem spaces, states, and search
• The number of possible moves in each state is small
(a maximum of two moves per state).
• Moving the tower to the middle peg (left part of the
figure) considerably increases the length of the
solution, and requires some backtracking.
• Things get quickly more complicated: try drawing
the problem space of the three-disk Tower of
Figure 8.5 Problem space of the two-disk tower of Hanoi. Hanoi!

Cognitive Psychology

27 28 7
 5/19/2024

Problem solving Problem solving

Problem spaces, states, and search Problem spaces, states, and search
• Problems come in different forms. • The structure of the external problem space does not completely
correlate with the difficulty of a problem
• Some problems, which have the same problem space and
therefore should have the same difficulty, differ markedly in the
• well-defined problems: the initial state, the goal state, time it takes to solve them.
and the operators are specified. • Example: Monster-move problem, an isomorph of the Tower of
• Examples: Tower of Hanoi and algebra problems. Hanoi (i.e. the two problems have the same structure when their
external space is analysed)
• Three monsters a small, medium-sized, and large monster hold a large,
• ill-defined problems: at least some of this information is small, and medium-sized globe, respectively.
missing. • They want to move the globes around so that each monster holds a
globe proportionate to its size. The rules are as follows (Kotovsky et al.,
• Examples: Most insight problems are of this kind, and so are most 1985, p. 251):
real-life problems such as buying a house or finding a partner. • Only one globe may be transferred at a time.
• If a monster holds two globes, only the largest of the two globes may be
transferred.
• A globe may not be transferred to a monster who is holding a larger globe.

Cognitive Psychology Cognitive Psychology

29 30

Problem solving
Problem spaces, states, and search
• Kotovsky et al. (1985)
• on average two minutes to solve the problem in the
Tower of Hanoi version
• 14 minutes to solve the monster-move version.
• It took even longer (29 minutes) to solve another version
(the monster-change) in which the globes shrink and
expand.
• These differences can be explained by the difficulties
Figure 8.6 Monsters-move problem: Example of the start state and goal state of the involved in learning and applying rules, and increased
Monster-Move problem, with the corresponding states of the tower of Hanoi. working memory load.

Cognitive Psychology

31 32 8
 5/19/2024

Problem solving Problem solving

Strengths and weaknesses of problem space theory Strengths and weaknesses of problem space theory

• Strengths of problem space theory(Gobet & Lane, 2015). • One relative weakness of this approach is that the
• problem space theory offers a normative theory of problem-solving
• participants used identifiable strategies, often means–end analysis.
problems that are studied tend to be puzzles:
• an analysis of the task environment and its complexity often made it possible artificial and fairly simple problems.
to predict the strategies used by the participants and their solution times.
• this analysis and the generation of external problem spaces make it possible
to measure the differences between human participants and an ideal solver.
• In general, the results show that people do not opt for the optimal solution
but rather use heuristics, which supports Simon’s theory of bounded
rationality.
• the models developed in this tradition tend to be well specified, and
are often implemented as computer programs.
• Examples include the Logic Theorist (Newell et al., 1958), one of the very first
artificial intelligence (AI) programs, which proves theorems in mathematical
logic and has led to a theory of creativity

Cognitive Psychology Cognitive Psychology

33 34

Problem solving Problem solving

Box 8.1 Expertise and thinking Box 8.1 Expertise and thinking
• Important application: study of expertise. How do experts solve • Another key difference: kind of internal problem
problems quickly, with much better results than non-experts? representation used.
• Fiding: De Groot’s (1978) study of chess players • For example, in physics, experts use fundamental principles (e.g. the
• top players very rapidly zero in onto the key aspects of a chess position concept of gravity), while novices use superficial aspects (e.g. the
“a world champion understands a problem situation better after five kind of device involved, pulley vs inclined plane; Larkin et al., 1980;
seconds than a strong amateur after 15 minutes” Chi et al., 1981).
• When anticipating possible moves and sequences of moves, strong • Having better representations affects the kind of search used. If one
players were very selective and rarely considered more than three or uses fundamental principles, it is easier to start from the givens and
four moves in a given position. generate unknown quantities, and thus to carry out forward search.
• Finding: Simon and Chase (1973) explaining mechanism • Expertise also impacts how external representations are
• With practice and study, chess players acquire a large number of used.
perceptual patterns, called chunks, which are domain specific. In chess, • For example, Tabachnek-Schijf et al. (1997) were interested in the
these patterns encode the location of a few pieces on the board. way diagrams were used in economics, specifically in problems about
• Chess players also learn what actions are favourable given a pattern. supply and demand. They found that experts were able to use graphs
• Simon and Chase estimated that chess masters have acquired about effortlessly and effectively. By contrast, novices displayed great
50,000 chunks. A later estimate based on computer simulations puts this difficulties with both using and generating graphs.
number even higher, at about 300,000 patterns (Gobet & Simon, 2000).

Cognitive Psychology Cognitive Psychology

35 36 9
 5/19/2024

DEPARTMENT OF PSYCHOLOGY
Thinking and problem solving

Cognitive Psychology • 8.1 Introduction

Ch. 8 Thinking and problem solving • 8.2 Problem-solving
• 8.3 Creativity
• 8.4 Analogy
• 8.5 Reasoning
• 8.6 Decision-making
• 8.7 Dual-process theories
May 20th, 2024 • Summary
Università di Milano-Bicocca Cognitive Psychology

37 38

Creativity Creativity
• Creativity = form of problem-solving with more complex problems (e.g. • Most of the support for Wallas’s theory comes from biographies
discovering a new law in psychology) and autobiographies of creative scientists and artists.
• Wallas (1926) proposed that creativity involves four broad stages (The first three
based on work by Poincaré,1913) • BUT self-reports are unreliable – human memory is reconstructive
• The first stage, preparation, is characterised by conscious attempts to solve the and cannot be fully relied upon for old memories!
problem, often from multiple directions. With more difficult problems, these attempts • Also the history of art and science contains many counter-
are unsuccessful,
• One must move to the second stage, incubation. The problem is left aside, but examples that clearly show that creators depend on a systematic
unconscious processing remains active and multiple ideas get combined behind the search for a solution rather than unconscious processing.
scenes, so to speak. (example Edison)
• This is followed by the illumination stage, where the solution comes as a flash of
insight, sometimes at an unexpected moment.
• Finally, there is the verification stage, where there is a conscious and deliberate effort
to check the solution. • The most influential theory of creativity is Newell and Simon’s
• Wallas’s four-stage theory emphasises unconscious processing, but the (1972) problem space theory
experimental evidence for this mechanism is sparse. • nothing mysterious about creativity: it is just a kind of problem-solving.
• P.e. Penaloza and Calvillo (2012) that a two-minute incubation was beneficial, • Creators selectively explore a problem space, using their knowledge and
but only when the problems were presented with misleading clues. a variety of strategies to keep the search efficient.
• Strong support from two sources: experimental psychology and artificial
intelligence.

Cognitive Psychology Cognitive Psychology

39 40 10
 5/19/2024

Creativity Creativity

• Experimental psychology support: Qin and Simon • Artificial intelligence support: a number of
(1990) carried out a ‘replication’ of Kepler’s computer programs embodying selective search can
discovery of the Third Law of Planetary Motion, be described as showing creativity.
which describes the motion of planets around the • Strong support for the theory, as this shows that the
Sun – an important discovery in astronomy. mechanisms proposed by problem space theory are
• They gave undergraduate students the data that Kepler sufficient for creativity.
used; however, there was no information about the • There are two main approaches, which sometimes
context (i.e. the labels ‘distance’ and ‘period’ were overlap.
replaced by X and Y). • The first approach is mainly interested in AI: the use of computers
to produce intelligent or creative behaviour, not necessarily in the
• The correct equation was found within one hour by about same way as humans.
one third of the students; in line with problem space • The second approach, falling under the umbrella ‘computational
theory, the successful students performed a limited modelling’, embodies theories of human behaviour that are
search with efficient heuristics. tested against empirical data.

Cognitive Psychology Cognitive Psychology

41 42

Creativity Thinking and problem solving

• Example of AI approach
• computer program displaying creativity: Logic Theorist (Newell et al., 1958), which
specialised in logic and was able to improve on the proofs constructed by some of the best
mathematicians of the twentieth century.
• Applied to the visual arts was AARON (Cohen, 1981), a production system program that drew • 8.1 Introduction
pictures following rules and constraints. (In a production system, the information is coded as
IF-THEN rules.) While limited to just a few thematic styles, AARON displayed creativity and
produced paintings of high aesthetic value (Boden, 1990). • 8.2 Problem-solving
• Example of simulation of human behaviour
• KEKADA (Kulkarni & Simon, 1988). The program was able to replicate in detail a complex
scientific discovery: the discovery of the urea cycle which earned Krebs the Nobel Prize in
• 8.3 Creativity
Physiology or Medicine in 1932.
• KEKADA used a number of heuristics implemented as productions. Not only was it able to • 8.4 Analogy
make theoretical inferences and evaluate the acceptability of its knowledge, but it could also
make predictions and propose novel experiments to test them. According to Kulkarni and
Simon, KEKADA’s behaviour matched the way Krebs carried out his research fairly well. • 8.5 Reasoning
• Gobet and Sala (2019) review recent developments in creative AI and argue that
psychologists can take advantage of these developments to design novel ways of
studying human creativity and develop new theories of creativity.
• 8.6 Decision-making
• 8.7 Dual-process theories
• Summary
Cognitive Psychology Cognitive Psychology

43 44 11
 5/19/2024

Analogy Analogy

• Analogy making is another important type of thinking. • Making analogies is very hard indeed
• It can be defined as the transfer of the solution of one • Finding: Series of experiments carried out by Gick and Holyoak
(1980) using Duncker’s (1945) tumour problem
problem to another problem. It is an efficient and • Question: providing analogical knowledge to the participants improve
sometimes creative reuse of knowledge. their success rate?
• Story about a general wanting to attack a well-defended fortress. The
• Examples of analogy from history of science: Rutherford’s roads leading to it were mined and would explode in the presence of
use of the solar system to explain the structure of the atom. large groups of soldiers; thus, it was not possible to send the entire army
on a single road. What the general did was to divide his force into small
• Models to explain analogy making (Gentner, 1983; Lovett & groups that approached the fortress simultaneously from different
Forbus, 2017) usually incorporate two mechanisms. directions. When they converged on the fortress, the whole army could
attack it with full power.
• First, an analogue or source (i.e. a problem similar to the one • Clearly, the fortress problem provides an analogy to the tumour problem:
currently faced) must be retrieved from long-term memory (p.e solar the soldiers correspond to the rays, the general to the surgeon, the
system) fortress to the tumour, and the risk of detonating mines to the risk of
• Second, a mapping should be made between the source and the hurting healthy tissue.
current or target problem (correspondence between the sun and the • In both cases, the solution is to spread one’s forces (the soldiers and the
rays, respectively) before hitting the target.
atom nucleus).

Cognitive Psychology Cognitive Psychology

45 46

Analogy Thinking and problem solving

• Making analogies is very hard indeed
• Finding: Series of experiments carried out by Gick and Holyoak
(1980) using Duncker’s (1945) tumour problem • 8.1 Introduction
• Results:
• only 20% used the story spontaneously and solved the problem successfully.
• Performance was better when they received a hint that the fortress story could be
• 8.2 Problem-solving
of use with the tumour problem; the percentage of success jumped to 92 per cent.
• This difficulty of using analogy has been replicated in several • 8.3 Creativity
studies.
• Seven experiments of Anolli et al. (2001) reproduced the key result that • 8.4 Analogy
information from the source analogue is not spontaneously transferred
to the target problem but is used only when participants receive a cue
that it is relevant.
• 8.5 Reasoning
• One possible explanation is that, with two stories that do not • 8.6 Decision-making
share any surface similarities, it is difficult to appreciate that they
have the same deep structure in common. • 8.7 Dual-process theories
• Summary
Cognitive Psychology Cognitive Psychology

47 48 12
 5/19/2024

Summary Reasoning

• The main subfields of the psychology of thinking are • Do people follow the laws of logic when they think? If
problem-solving, expertise, creativity, analogy making,
reasoning, and decision-making. not, how they depart from these laws?
• Behaviourists proposed that problem-solving can be • Research can be divided into two main topics:
explained by learning stimulus–response links. Gestalt • deductive reasoning, where the conclusions necessarily follow
psychologists devised clever experiments, but their
theoretical explanations were vague. from the premises
• Deductive reasoning task = A problem that has a well-defined
• Problem space theory proposed that problem-solving structure in a system of formal logic where the conclusion is certain.
consists of search within a problem space. It has received
support from both experiments and computer simulations.
The theory has also been successfully applied to the • inductive reasoning, where only probable conclusions can be
subfields of expertise and creativity. derived from the available information
• Some authors have proposed that analogy making is at the • Inductive reasoning task = A problem that has a well-defined
heart of thinking, but experiments have shown that people structure in a system of formal logic where the conclusion is highly
struggle when trying to use analogies. probable but not necessarily true.

Cognitive Psychology Cognitive Psychology

49 50

Reasoning Reasoning
Deductive reasoning Deductive reasoning – Truth tables

• Basics of deductive reasoning (form of logic). • Truth table for the logical conjunctive AND.
• building blocks: propositions (e.g. ‘the sky is blue’), which A B A AND b
can be either true or false.
T T T
• propositions can modified or joined by logical operators
T F F
(NOT, AND, OR, IF-THEN …)
F T F
• Propositional logic allows one to establish the truth of
modified propositions F F F

• Example
• P1 = ‘the sky is blue’ is TRUE , P1 =‘roses are red’ is TRUE
• compound proposition P1 AND P2 is true?
• formalism of truth tables to deduce the correctness of a new
proposition mechanically

Cognitive Psychology Cognitive Psychology

51 52 13
 5/19/2024

A B A AND b

T T T
Reasoning Reasoning T F F

Deductive reasoning – Truth tables Deductive reasoning F T T

F F T
• Truth table for the conditional operator IF-THEN. • Correct and incorrect inferences
A B A AND b • It is also possible to assume that the implication is true, to give
T T T
the truth value of A (or B), and ask what can be inferred about the
truth value of the other statement.
T F F • Example
F T T • ‘IF I study hard, THEN I’ll pass the exam’ TRUE
• I studied hard; can we conclude that I’ll pass the exam?
F F T
• From Table if A is true and ‘IF A THEN B’ is also true, then B is also true.
=> correct inference to conclude that I’ll pass the exam. This type of
• P1 = ‘it’s raining; P2 = I’m wet inference is called the modus ponens.
• if statement A is false, then the value of statement B does not matter: the IF- • The meaning of the statements is irrelevant! what matters is their
THEN statement is always true. form. Example:
• Important consequences: if a statement is incorrect in a logical system, then by a • If psychology is a science, then the moon is blue
chain of inferences any statement can be derived. • Psychology is a science
• Thus, if one or several statements are false in a scientific theory, then anything • Therefore the moon is blue
can be concluded from this theory.
• This is a correct inference, even though it is semantically incorrect.

Cognitive Psychology Cognitive Psychology

53 54

A B A AND b A B A AND b

T T T T T T
Reasoning T F F Reasoning T F F

Deductive reasoning F T T Deductive reasoning F T T

F F T F F T
• Correct and incorrect inferences • Correct and incorrect inferences
• The second correct inference is called modus tollens. • Modus ponens and modus tollens are the only two
• ‘IF P THEN Q’ is true
• Q is not true; the inference is: therefore P is not true. This valid inferences with conditional problems.
inference can be abbreviated as ‘IF NOT Q then NOT P’. • Two inferences are incorrect, although very
• This is a much less intuitive rule. Modus tollens is at the heart common.
of much experimental psychology, where the aim of running an
experiment is to refute a theory. • Affirming the consequent
• Consider the following example, which uses some information • ‘IF P THEN Q’ is true
you learned in this chapter. • Q is true, and the conclusion is ‘Therefore P is true’.
• If Gestalt theory is correct, then solving insight problems requires an
‘Aha!’ experience (IF P THEN Q) • Why is it incorrect?
• Insight problems are sometimes solved without an ‘Aha!’ experience • Example: if it rains, then I’m wet; I’m wet. To conclude that it
(Webb et al., 2016) (not Q) rained is incorrect, because I could be wet for other reasons (for
• Therefore, Gestalt theory is incorrect (not P). example, I took a shower).

Cognitive Psychology Cognitive Psychology

55 56 14
 5/19/2024

A B A AND b

T T T
Reasoning T F F Reasoning
Deductive reasoning F T T Deductive reasoning
F F T

• Correct and incorrect inferences • Empirical results

• Modus ponens and modus tollens are the only two • Overall, the results clearly indicate that people do
valid inferences with conditional problems. not perform well with reasoning tasks.
• Two inferences are incorrect, although very • Case: premises (P and Q) are given, and the
common.
• denying the antecedent question is whether certain conclusions are valid.
• ‘IF P THEN Q’ is true • 100% of participants correctly use modus ponens
• P is not true; the conclusion is ‘Therefore Q is not true’. • 60% correctly use modus tollens.
• Why it is incorrect? • 50% commit the invalid inferences of affirming the
• consider again our raining example. To conclude that I’m consequent and denying the antecedent (Gilhooly, 1996).
not wet because it is not raining is incorrect; for example,
if I take a shower on a sunny day, I would be wet.
Cognitive Psychology Cognitive Psychology

57 58

Reasoning
Deductive reasoning

• Empirical results
• Finding: Wason’s (1968) selection task
• You are presented four cards
• decide whether the following rule is correct or not:
• if a card has a vowel on one side, it has an even number on the Figure 8.7 The Wason Selection Task
other side.
• You should turn over only the card(s) needed for testing the rule. Note: The rule is ‘If there is an A on one side of the card, then there is a 2 on the other
Which card(s) do you turn over? side of the card.’ Select only those cards that would need to be turned over in order to
decide whether or not the rule is correct. Each card has a letter on one side and a
number on the other.

Cognitive Psychology

59 60 15
 5/19/2024

Reasoning Reasoning
Deductive reasoning Deductive reasoning
• Empirical results • Empirical results
• The card ‘A’ is a correct choice, and this is an application of
the modus ponens. Intuitively, this makes sense: if there is
no even number on the back of ‘A’, the cards would not • Ample evidence that people struggle with reasoning
follow the rule. tasks when they are presented with abstract material.
• By contrast, the card ‘2’ is incorrect – it is equivalent to • They do much better when the material consists of
affirming the consequent. concrete information.
• What we need to do is to try to falsify the rule, using modus • Example: variation of Wason’s task.
tollens; this can be done by turning over the card ‘7’ and • As a police officer, you must implement the rule that only people over
18 can drink beer in a pub. So, the rule is: if a customer in the bar is
expecting to see a ‘K’ on the other side. drinking beer then that individual must be older than 18. You are a
• Only few people choose the correct cards. police officer trying to enforce the rule, and you see four customers in
the bar. The first customer is having a beer; the second is sipping
• Most people select the cards ‘A’ and ‘2’ (46 per cent) coke; the third is clearly older than 18; and the last seems to be 16
• even ‘A’ alone (33 per cent) years old
• Only 4 per cent selected the two correct cards, ‘A’ and ‘7’ (Johnson- • Which customer(s) do you need to check?
Laird & Wason, 1970).

Cognitive Psychology Cognitive Psychology

61 62

Reasoning
Deductive reasoning

• Empirical results
• The problem is an isomorph to Wason’s selection
task
• Results: about 75 per cent of the participants get
the answer correct: the person drinking beer and
Figure 8.8 Drinking beer
the person looking under 18 (Griggs & Cox, 1982).
• Remember that only 4 per cent of the participants
got the correct answer in the abstract version.

Cognitive Psychology

63 64 16
 5/19/2024

Reasoning Reasoning
Deductive reasoning Deductive reasoning
• Theories • Theories
• A first group of theories propose that people use • Second approach: mental models (Johnson-Laird, 2001),
content-free logical rules, similar to those used in • people build an internal representation of the information
formal logic (Rips, 1994) based on abstract rules. contained in a problem
• Reasoning consists in applying these rules in the same way • then mentally inspect this representation to establish whether
logicians use them to derive a proof. Thus, in ideal conditions,
human reasoning would be perfect. it is correct. In particular, they try to find counter-examples
• However, people commit errors because they fail to that would invalidate an inference.
understand the statements of the problems, use the wrong • The mental model approach has been predominantly
strategy to tackle the problem, or suffer from limits in working
memory. used with a type of reasoning problems called syllogism,
• Weakness: they assume full rationality only to add a such as:
number of ad hoc assumptions to limit it and thus • All the artists are beekeepers.
explain the pattern of errors seen in the data. • Some of the beekeepers are chemists.
• Can we conclude that all the artists are chemists?

Cognitive Psychology Cognitive Psychology

65 66

Reasoning Reasoning
Deductive reasoning Deductive reasoning
• If you are told that Caroline is taller than Hannah and that Hannah is • Finding: participants perform better on the concrete version of Wason’s
taller than Julian, it is likely that you imagine these persons in a row in selection task
your mind’s eye. • => theory based on pragmatic reasoning schemas (Cheng and Holyoak,
• If you are now told that Paul is smaller than Julian, you can add Paul to 1985).
your mental picture, and you can ‘see’ that Paul is the smallest person. • schemas are general and abstract, but less so than logical rules as they are
• Mental models in reasoning are based on the same idea, but the details constrained by the satisfaction of a goal.
rapidly become more complicated. • Example: permission schema (if one wants to do X, then one must satisfy
condition Y).
• Mental model theory does a fairly good job at predicting the time it • In everyday reasoning, people do not use rules or construct internal
takes participants to solve reasoning problems of varying difficulty. representations of the problem at hand, but rather use schemas based on their
experience.
• Weaknesses of mental models: • For example, in the pub version of Wason’s task, but not in the original abstract
• gives importance to the way the statements of a problem are encoded to build version, people can use a permission schema. When it is difficult to apply such a
models, it lacks mechanisms specifying how encoding occurs. schema, as in abstract problems, then people tend to make mistakes.
• Prediction: problems requiring the construction of many mental models are • Strength: accounts for the fact that people do better with concrete
harder to solve, because this drains working-memory resources. BUT empirical
data do not support problems than abstract problems.
• verbal protocols offer little evidence that people try to find counter-examples, • Weakness:
which is a central assumption of the theory.. • it is silent about the first phase of reasoning, where verbal information is
encoded.
• It is limited to problems of the type IF-THEN, and does not say much about other
Cognitive Psychology
types of reasoning problems. Cognitive Psychology

67 68 17
 5/19/2024

Reasoning Reasoning
Inductive reasoning Inductive reasoning

• Inductive reasoning: • Inductive reasoning:

• 2-4-6 task
• it is possible to reach only probable conclusions from the • participants receive the sequence 2–4–6 and their task is to find the rule
given information. behind the sequence.
• They can produce as many sequences of three numbers as they want,
• conclusions must go beyond the information given. writing down a brief description of the rule they entertain, and are given
a simple YES or NO feedback from the experiment.
• There is more uncertainty than with deduction, and, • After having received sufficient positive feedback, participants would
indeed, no guarantee that the conclusions that are stop and state the rule they had found.
reached are correct. • Typical test sequences were 8–10–12 or 14–16–18, with YES as feedback.
• Most participants formed the rule ‘sequence of even numbers increasing
by two’. However, the rule that the experimenter had in mind was ‘any
sequence of increasing numbers’.
• In Wason’s (1960) original study, only 6 out of 29 participants found the
correct rule without proposing any incorrect ones, with 13 finding an
incorrect rule, nine proposing two incorrect rules or more, and one not
proposing any rule. Importantly, few participants tried to disprove their
rule by sequences such as 8–10–6 or 3–9–44.

Cognitive Psychology Cognitive Psychology

69 70

Reasoning Reasoning
Inductive reasoning Inductive reasoning
• Inductive reasoning: • H: Deductive and inductive reasoning rely on different
• Experimenters have tried all sorts of manipulations, but it has turned out
surprisingly difficult to improve participants’ performance in the 2–4–6 task. Ù
mechanisms.
• One of the rare successful manipulations consists in telling participants that the • Disconfirmation: Stephens et al. (2018)
experimenter has two rules in mind, the DAX rule and the MED rule. • Using three data sets, they compared models that assumed
• DAX is the original ‘any sequence of increasing numbers’ rule, while MED covers two different processes with models assuming only one
all the number triplets not following that rule (so MED is the complement of process.
DAX: any triplet not made of ascending numbers; e.g. 2–2–8).
• Only one model was able to account for all the data, and it was
• At the beginning, participants are given the triplet 2–4–6 as an instance of DAX. a model which incorporated only one mechanism: argument
• In this version of the experiment, about twice as many participants state the strength. This strength is evaluated on a single dimension, with
correct rule in their first attempt, as compared to the standard version of the different decision thresholds used for inductive and deductive
task (Wharton et al., 1993). judgments.
• However, participants took advantage of the two rules only if they were told • If Stephens et al.’s conclusions are replicable, this means that
explicitly that they are complementary. One possible explanation for the
increased success in this experiment is that participants do not need to refute a the experimental results on deductive reasoning and inductive
rule; rather, what they do is to confirm its complement. reasoning can be explained by a single theory, which would be
an important step in our understanding of thinking.

Cognitive Psychology Cognitive Psychology

71 72 18
 5/19/2024

Thinking and problem solving Decision making

Expected utility

• Focus: decision-making under uncertainty - the way

people make decisions with only imperfect information.
• 8.1 Introduction
• Influential approach - Von Neumann and Morgenstern (1944):
• 8.2 Problem-solving • people, perhaps unconsciously, use all the available
• 8.3 Creativity information in order to make an optimal decision => people are
maximisers
• 8.4 Analogy • one attaches a utility (i.e. a value) to each outcome and assesses the
probability that this outcome will occur.
• 8.5 Reasoning • By multiplying the utility of each outcome by its probability and
taking the sum over all possible outcomes, one gets the expected
• 8.6 Decision-making utility of an option.
• 8.7 Dual-process theories • The same procedure is applied to all options. The best solution is the
option with the largest expected utility (see Box 8.3).
• Summary
Cognitive Psychology Cognitive Psychology

73 74

Decision making Decision making

Expected utility Euristics and biases

• Example of how expected utility is computed. • BUT people are probably not maximisers (Simon
• You have to choose between two lottery tickets (i.e. two bounded rationality)
options).
• First ticket • Finding: Kahneman and Tversky
• outcome 1, probability .1: winning £10 • No computation of expected utility
• Outcome 2, probability .9: losing £1.
• Expected value i= (£10 x .10) + (–£1 x .90) = £1.00 – £0.90 = £0.10. • humans often violate the predictions of the theory.
• With this ticket, you are expected to win £0.10. • humans often use heuristics, but that these heuristics can
• Second ticket lead to systematic biases.
• outcome 1, probability .1: winning £20
• Many heuristics. Examples
• Outcome 2, probability .9: losing £2.
• Expected value i= (£20 x .10) + (–£2 x .90) = £2.00 – £1.80 = £0.20. • availability heuristic
• With this ticket, you are expected to win £0.20. • representativeness heuristic
• you should choose the second ticket.

Cognitive Psychology Cognitive Psychology

75 76 19
 5/19/2024

Decision making Decision making

The availability heuristic The availability heuristic

• Example 1: Statistics show that flying is much safer than • Example 2: if a word of three letters or more is sampled
travelling by car. In fact, you are 10,000 times less likely at random from an English text, is it more likely that the
to die in a plane accident than in a car accident. word (a) starts with ‘r’ or (b) has ‘r’ as its third letter?
However, people generally think that travelling by car is • When facing this question, participants usually try to generate
safer. How can that be? words starting with an ‘r’ and words having an ‘r’ as the third
• People use the availability heuristic: basing one’s letter.
estimate of probability on how easy it is to retrieve • However, it is much easier to generate words with a given
examples from long-term memory. letter in the first position than in the third position, in part
• Typically, there are few casualties in a car accident. because people occasionally do the former but never the latter.
• By contrast, there are many deaths in an airplane crash and the • Thus, participants conclude that words with an ‘r’ in the first
accident is widely covered in the news, with horrific pictures. position are more common. In fact, there are more words with
• As a consequence, more details are stored in long-term an ‘r’ in the third position.
memory, which makes retrieval of plane accidents easier.

Cognitive Psychology Cognitive Psychology

77 78

Decision making Decision making

The representativeness heuristic The representativeness heuristic
• representativeness heuristic - people make decisions on • FINDING: Kahneman and Tversky (1973)
probability by retrieving from LTM instances that are similar
to the problem at hand. However, people typically use only • Group 1 story: a person had been chosen randomly from
few instances to make a decision, which means that the a set of 100 individuals consisting of 70 engineers and 30
information is not reliable. lawyers (high-engineer group).
• Law of large number: many observations lead to more • What’s the probability that she is an engineer? Participants
reliable conclusions than few observations => we should try correctly answered that the probability was .70.
to base our decisions on a large sample of instances
• BUT people very often do the opposite. • Group 2 story: Same question, but now with the 100
• Classic example: gambler’s fallacy. individuals consisting of 30 engineers and 70 lawyers
• Fair coin. 3 heads/10 tails. (low-engineer group).
• Gambler’s fallacy: it is more likely to land on heads in the coming • The second group correctly answered that the probability was .30.
tosses!
• BUT a coin does not have any memory and the probability of landing
heads or tails is still 50%

Cognitive Psychology Cognitive Psychology

79 80 20
 5/19/2024

Decision making Decision making

Euristics and biases Euristics and biases
• All groups: second individual had been selected, and received the • The participants were given the following description (Kahneman &
following description Tversky, 1973, p. 242):
• Jack is a 45-year-old man. He is married and has four children. He is
generally conservative, careful, and ambitious. He shows no interest in • Dick is a 30-year-old man. He is married with no children. A man of high ability
political and social issues and spends most of his free time on his many and high motivation, he promises to be quite successful in his field. He is well
hobbies which include home carpentry, sailing, and mathematical liked by his colleagues.
puzzles.
• The probability that Jack is one of the 30 (70) engineers in the sample of • Estimate the probability that Dick is one of the 30 (or 70) engineers in the sample.
100 is —-%.
• Participants in both groups estimated that the probability of being • As this last description does not provide any information with respect to
Dick’s profession, the estimates should be the same as in the first
an engineer was .90! question and reflect the a priori probabilities.
• BUT participants knew the a priori probability. • However, in both the high-engineer and low-engineer groups,
• To answer the second question correctly, one needs to take into participants estimated that the probability was .50. T
account background information (base rate information) about • hus, participants provided a different answer when given no additional
the population. However, participants ignored the base rate, information and when given irrelevant information – a rather strong
presumably because they thought Jack fitted the description of an violation of rationality!
engineer better than that of a lawyer.

Cognitive Psychology Cognitive Psychology

81 82

Decision making Thinking and problem solving

Euristics and biases: evaluation
• Heuristics were seen in a positive way in research on
problem-solving (for example in Newell and Simon’s
research), in research on decision-making they shed a bleak • 8.1 Introduction
light on human cognition
• But are humans really dumb? Maybe not and that • 8.2 Problem-solving
Kahneman and Tversky underestimate our abilities
(Gigerenzer, 2000). • 8.3 Creativity
• The term ‘probability’ they use in the statements might be
understood as referring to similarity and not to mathematical
probability. Most people do not have a formal training in probability
• 8.4 Analogy
theory and thus have a poor understanding of concepts linked to
probability. • 8.5 Reasoning
• Humans are adapted to the environment in which they live and take
advantage of some of its features, such as its statistical structure. • 8.6 Decision-making
However, artificial problems in the laboratory do not provide such a
structure. • 8.7 Dual-process theories
• Heuristics are useful most of the time, but they lead to errors in
Kahneman and Tversky’s experiments because these authors cleverly
identified when these heuristics would break down.
• Summary
Cognitive Psychology Cognitive Psychology

83 84 21
 5/19/2024

Dual Process theory Dual Process theory

• The cognitive system consists of two kinds of cognition.
• This partition is a feature of numerous theories
• First
• evolved early (Fiedler & von Sydow, 2015, count no less than 30
•
•
shared by many animals
based on perception and learning
theories!), which differ in
• intuitive, operates automatically and with little effort • the nomenclature used (e.g. System 1 and 2; Type 1 and
• uses simple heuristics.
• Experimentally, it is independent of manipulations such as working memory load Type 2 processing)
and time pressure. • the exact nature of each form of cognition.
• Second
• newer evolutionarily,
• uniquely human,
• characterised by analytical and rule-based thinking that requires attention and
effort.
• slow and deliberative
• highly sensitive to experimental manipulations of working memory and time
pressure.
• correlates with intelligence.

Cognitive Psychology Cognitive Psychology

85 86

Dual Process theory Dual Process theory

• Evans’s (2006) theory provides a good example of this • Kahneman (2003) applied similar ideas to the field
dual approach in the field of reasoning. of judgement and decision-making.
• heuristics are implemented in System 1
• analytical processes in System 2. • Just like Evans’s (2006) theory, Kahneman’s (2003)
• When a reasoning problem is encountered, theory involves serial processing:
• System 1 rapidly constructs a mental model using simple heuristics. • System 1 enters into action first, generating a rapid and
• Only then does System 2 enter the picture to possibly revise or even
replace the original mental model, using deliberate and slow intuitive answer.
analytical processes. • In a second step, System 2 improves and possibly changes
• Which system is engaged is affected by a number of factors, the original answer.
including
• the amount of time available to provide an answer,
• the level of intelligence of the participants
• The instructions provided by the experimenter.

Cognitive Psychology Cognitive Psychology

87 88 22
 5/19/2024

Dual Process theory Dual Process theory

• Do experts solve problems by pattern recognition – • Dual-process theories have been highly influential in the
rapid identification of the key features of a problem last two decades or so.
that lead automatically to a solution – or by look- • good empirical support for the existence of the two
ahead search and analytical thinking? different modes of processing,
• Simon and Chase (1973) emphasised the former • number of criticisms have been levelled at these
mechanism theories.
• Holding (1992) argued that the importance of perception, • First, they do not propose anything new (Brakel & Shevrin,
intuition, and knowledge had been overstated by Simon 2003), as the key ideas were already present in Freud’s
and Chase, and that experts mostly solve problems by (2001/1915) distinction between primary and secondary
analysing the consequences of possible options. processes.
• More recently, researchers have emphasised that most • Second, these theories are not well specified, and could be
domains of expertise engage both processes (Gobet, characterised as circular restatements of the phenomena
2016). under study (Gigerenzer, 2010).

Cognitive Psychology Cognitive Psychology

89 90

Dual Process theory Dual Process theory

• number of criticisms have been levelled at these
• number of criticisms have been levelled at these theories.
• Finally, the assumption of seriality is not plausible (De Neys,
theories. 2006) as it is not clear how System 2 knows that it should be
• Third, the features of the two systems do not map onto engaged.
the two modes of processing as neatly as proposed; for • possible solution: assume that the two systems operate in parallel,
but this would seem to throw away the benefits provided by System 1
example, heuristics are used both intuitively and (i.e. it is fast and intuitive).
deliberately (Gigerenzer, 2010). • Another possibility is that the two systems closely interact. Such a
theory was developed to explain expert problem-solving (Gobet,
1997). The theory, implemented as a computer program, shows how
pattern recognition (corresponding to System 1) is applied recursively
during search (corresponding to System 2). Every time a new problem
state is generated in imagination, pattern recognition suggests
possible courses of action. If these suggestions fail, deliberative
heuristics are used. When decisions must be taken rapidly (e.g. within
one second), there is time for only pattern recognition.

Cognitive Psychology Cognitive Psychology

91 92 23
 5/19/2024

Summary Summary
• The main subfields of the psychology of thinking are problem-solving, • Expected utility theory offers a normative theory of decision-
expertise, creativity, analogy making, reasoning, and decision-making. making under uncertainty. Humans’ decisions do not follow the
• Behaviourists proposed that problem-solving can be explained by predictions of this theory, in part because they use heuristics that
learning stimulus–response links. Gestalt psychologists devised clever sometimes lead to systematic biases.
experiments, but their theoretical explanations were vague. • Dual-process theories have been developed to explain the
• Problem space theory proposed that problem-solving consists of search empirical results in most subfields of thinking. These theories
within a problem space. It has received support from both experiments postulate the existence of two systems: System 1 is fast and
and computer simulations. The theory has also been successfully applied intuitive whilst System 2 is slow and deliberative.
to the subfields of expertise and creativity.
• Some authors have proposed that analogy making is at the heart of
• Research on thinking has been dominated by the question of
thinking, but experiments have shown that people struggle when trying rationality. In general, the data do not support the hypothesis of
to use analogies. full rationality. Rather, they support the hypothesis of bounded
rationality. Due to the limits of attention and working memory,
• Research into deductive reasoning compares the way people reason with biases affecting retrieval from long-term memory, and the
solutions obtained by the application of logic, which offers a normative difficulty of searching large problem spaces, humans are
theory. People commit many errors when faced with reasoning satisficers rather than optimisers, and choose solutions that are
problems. With inductive reasoning, people suffer from confirmation
bias. good enough rather than optimal.

Cognitive Psychology Cognitive Psychology

93 94

24