Thinking: Problem Solving, Judgment, and Cognitive Flexibility

A solution can be logically simple yet difficult to notice. A familiar tool suggests its usual purpose, a vivid story feels more probable than a broader category, and a previously successful rule survives even after the situation changes. Thinking involves both building a useful representation and checking whether the first answer deserves our confidence.

Learning goals and Feynman practice

Explain functional fixedness and mental sets, distinguish algorithms from heuristics, solve three classic reasoning problems, and connect executive control with distributed brain systems. Separate an experimental observation from the mechanism proposed to explain it.

Read a section, close it, and teach the mechanism to a beginner. Give a new example and one limitation. Attempt each puzzle before opening its answer. When you get stuck, identify the exact assumption blocking progress and repair it.

1. Problem solving: represent the gap before searching for a move

A problem has an initial state, a goal state, possible operations, and constraints. Solving it means finding a way to bridge the gap. Some problems specify these elements clearly; others require discovering what the actual goal or constraint is.

Suppose you must move a large table through a doorway. “Push harder” searches within one representation. “Change its orientation or remove detachable legs” changes the available operations. A better representation can make a previously invisible solution accessible.

Functional fixedness and mental sets

Functional fixedness is difficulty seeing an object’s possible uses beyond its customary function. In the two-string problem, two hanging strings cannot both be reached at once. A heavy tool can serve as a pendulum weight, bringing one string back within reach while the solver holds the other. The obstacle is partly recognizing a useful physical property rather than only the tool’s conventional purpose.

Mental set is broader: a familiar strategy continues to guide problem solving when another approach would be better. A spreadsheet specialist may try increasingly elaborate formulas when reorganizing the data would solve the problem more simply.

Prior knowledge is not inherently harmful. It makes expert solutions possible, but can also narrow the candidate actions considered. Ask: “Which constraints are stated, and which have I supplied from habit?”

Teach it back: Explain the difference between not knowing a tool’s physical properties and knowing them but failing to consider an unfamiliar use.

2. Restructuring, analogy, and insight

Restructuring changes how a problem is represented. Analogy transfers a useful relationship from another domain: for example, understanding queue congestion by comparing it with a bottleneck in a transport network. Good transfer depends on shared structure, not merely surface resemblance.

Insight is the subjective experience of a solution becoming suddenly clear. The feeling can accompany a valid solution, but it is not proof of correctness. An answer that feels obvious still needs checking against the constraints.

Incubation is setting a problem aside before returning to it. A break may reduce fixation, change cues, or allow a fresh approach. Sleep can support some forms of learning and reorganization, but “sleeping on it” does not guarantee a solution or establish that every insight was generated by an unconscious solver during the break.

Teach-back check: how would you test whether a break improves insight?

Compare matched problems and groups with and without a break, controlling total working time, prior exposure, and task difficulty. Record correct solutions rather than only the feeling of insight. A benefit still would not uniquely identify its underlying mechanism.

3. Algorithms and heuristics: procedures have conditions

An algorithm is a specified procedure for a task. Some algorithms guarantee a correct result when their assumptions hold, the input belongs to the defined domain, and the procedure is correctly completed. Not every algorithm guarantees the answer to every problem: some are approximate, computationally impractical, or designed for a different objective.

A heuristic is a simplifying strategy that can produce useful judgments efficiently without guaranteeing correctness. “Try the most constrained move first” may save search time. “Choose the most familiar option” may work in one environment and fail in another.

Kahneman and Tversky’s work showed how useful judgment shortcuts can produce systematic errors. Heuristics are not synonymous with irrationality, and deliberate thinking is not automatically correct. The practical question is whether the shortcut suits the environment and stakes.

Approach	Example	What needs checking?
Exact procedure	Calculate a conjunction probability from an explicitly specified model.	Whether the model and inputs are appropriate.
Search heuristic	Start with a promising candidate solution.	Whether better alternatives were overlooked.
Similarity judgment	Classify a case by resemblance to familiar examples.	Whether relevant base rates were ignored.
Memory-based estimate	Use readily recalled events to estimate frequency.	Whether recall is representative of actual frequency.

4. Representativeness: resemblance is not probability

Representativeness uses similarity to a prototype as evidence for category membership. It can be informative, but may lead someone to neglect base rates, the background prevalence of the relevant categories.

Imagine a fictional town with many more shopkeepers than astronomers. A resident enjoys observing the sky. That detail may fit an astronomer stereotype, but a probability judgment also requires considering how common each occupation is and how often people in each occupation share the interest. The example supplies no numerical probabilities, so it does not justify an exact answer.

The Linda conjunction problem

A description presents Linda as concerned with social justice. Compare:

• A: Linda is a bank teller.
• A and B: Linda is a bank teller and active in a feminist movement.

The second description may feel more representative. Under the intended probability interpretation, however, it cannot be more probable than the broader first event. Tversky and Kahneman used such problems to study the conjunction fallacy. [1]

$P(A \cap B) \leq P(A)$

Every bank teller who is active in that movement is already included among bank tellers. Adding a condition narrows the set. Some task wordings invite pragmatic interpretations, such as reading the first statement as “bank teller and not active”; explicitly separating the events makes the probability rule clear.

Teach-back check: explain the conjunction rule without a formula

Among 100 people, suppose 20 are bank tellers. The number who are both bank tellers and movement activists cannot exceed those 20. This is an illustrative count, not a claim about Linda’s actual population.

5. Availability: easy to retrieve is not necessarily frequent

The availability heuristic uses the ease of retrieving or imagining examples to estimate frequency or probability. Tversky and Kahneman investigated how memory accessibility can influence judgments. [2]

In the classic letter-position task, words beginning with a letter such as K are easier to generate than words with that letter in the third position. This can bias estimates toward first-position frequency. The direction of the example is the key lesson; an exact ratio depends on which word corpus and counting rules are used.

A vivid news report can similarly make a rare event feel common. Ease of recall reflects exposure, emotional salience, and retrieval cues as well as actual frequency. Before estimating risk, ask what observations were sampled and what denominator is missing.

Teach it back: Why does listing ten memorable incidents fail to tell you the incident rate without knowing how many opportunities there were?

6. Wason selection: search for a counterexample

Four cards show A, D, 4, and 7. Each has a letter on one side and a number on the other. Test the rule: If a card has a vowel on one side, it has an even number on the other. Which cards must be turned?

Try before opening the answer.

Solution and the reason for each card

Turn A and 7. A would violate the rule if an odd number were behind it. Seven would violate it if a vowel were behind it. D cannot violate a rule that restricts vowels only. Four does not need a vowel: the rule does not say that every even-numbered card must have one.

The only forbidden combination is vowel plus odd number. The logical statement is one-way: “if vowel, then even” does not imply “if even, then vowel.”

People often select cards that match terms mentioned in the rule or seem likely to confirm it. Confirmation bias, matching tendencies, task interpretation, and perceived relevance can all contribute; one error should not automatically be assigned a single cause.

A practical repair is to ask: What possible observation would show this rule to be false? This is useful for testing claims, although real evidence also requires evaluating reliability and alternative hypotheses.

7. Disjunctive reasoning: consider every live case

Jack is looking at Anne. Anne is looking at George. Jack is married; George is unmarried. Is a married person looking at an unmarried person?

Use the puzzle’s intended two-category assumption: Anne is either married or unmarried. Do not silently introduce another category or relationship.

Solution: both branches lead to yes

If Anne is married, she is looking at unmarried George. If Anne is unmarried, married Jack is looking at her. Therefore the answer is yes under the stated assumptions. Anne’s unknown status changes which pair establishes the result, not whether such a pair exists.

Fully disjunctive reasoning checks the relevant alternatives rather than stopping at an unknown fact. The lesson is not to ignore missing information; it is to determine whether that information can change the conclusion.

Teach it back: Invent a different two-case problem in which the conclusion stays the same in both branches. Then change one premise so that the missing fact matters.

8. Executive control: keeping goals and updating strategies

Prefrontal regions, including dorsolateral prefrontal cortex, contribute to maintaining goals, working with information, planning, and adjusting behaviour. These abilities depend on networks involving parietal, striatal, and other systems, not a single general “thinking centre.”

Task	What the person does	Relevant difficulty	Interpretive limit
Wisconsin Card Sorting Test	Infers a sorting rule from feedback and adjusts when it changes.	Perseveration: continuing an old rule despite feedback.	Errors can also reflect attention, working memory, or understanding; they are not unique to frontal injury.
Cognitive estimation	Produces a plausible estimate when exact knowledge is unavailable.	Combining knowledge, constraints, and reasonableness checks.	Culture, education, and factual knowledge affect answers.
Tower of London	Plans moves to reach a goal configuration under rules.	Sequencing, subgoals, and avoiding impulsive moves.	Motor, visual, comprehension, and working-memory demands also contribute.

A patient with prefrontal damage may struggle with these tasks, but “cannot adapt at all” is too categorical. Lesion location, extent, connections, and individual history affect performance. One test cannot diagnose a specific lesion or establish a single lost faculty.

Estimation as a checkable chain

For an unfamiliar quantity, break the estimate into components, state assumptions, and check the resulting scale. To estimate seats in a small cinema, estimate rows and seats per row, multiply, then compare with the room’s apparent size. This is a reasoning strategy rather than a validated clinical test.

Teach it back: Why can an implausible estimate reflect poor monitoring even when the arithmetic itself is correct?

9. Iowa Gambling Task: outcomes, bodily signals, and interpretation

In the Iowa Gambling Task, participants choose among decks with different patterns of gains and losses. Some offer appealing immediate gains but worse long-run outcomes in the task. Studies by Bechara, Damasio, and colleagues found disadvantageous choices and altered anticipatory skin-conductance patterns in patients with ventromedial prefrontal damage. [3]

The somatic marker hypothesis proposes that bodily or bodily represented signals can help guide decisions based on experience. This does not mean that every gut feeling is accurate or that autonomic arousal uniquely encodes the objectively best choice.

Later work using more detailed questions found that participants could have more explicit task knowledge than earlier measures detected, challenging a simple claim that advantageous decisions necessarily arise before conscious knowledge. [4] Learning, reversal, working memory, loss frequency, and task details can also influence choices.

Skin conductance indexes autonomic activity, not a direct readout of wisdom or emotion. The task supports investigating interactions among valuation, experience, physiological signals, and control; it does not isolate a universal brain switch for safe decisions.

Teach-back check: why does a physiological difference before a choice fail to prove that it caused the choice?

The signal may accompany another learning process or reflect shared underlying influences. Strong causal interpretation requires additional evidence beyond temporal order and correlation, and the meaning of the signal must be independently established.

10. Worked method: improve a difficult decision

Suppose a team prefers a familiar proposal because its story is compelling and several recent successes come easily to mind.

1. Represent the problem: state the goal, constraints, and what counts as success.
2. Check the mental set: identify assumptions inherited from earlier projects.
3. Generate alternatives: change the representation or use a structurally relevant analogy.
4. Check denominators: compare success rates, not only memorable examples.
5. Try to falsify: specify evidence that would undermine the preferred proposal.
6. Branch the uncertainty: ask whether the decision changes under each plausible case.
7. Monitor the outcome: update the rule when evidence changes rather than repeat it automatically.

This is an application of the concepts, not a guarantee of a correct decision under uncertainty. Heuristics remain useful when time and information are limited; targeted checks help identify where they may mislead.

Retrieval workshop and spaced review

From memory, explain functional fixedness and mental set with different examples. Solve the card and relationship puzzles again without memorizing only the answer. Explain representativeness and availability, then identify what additional evidence would test each explanation.

Repair: “Heuristics are bad, algorithms are always correct”

Heuristics can be efficient and accurate in suitable environments. Algorithmic guarantees depend on the procedure, its domain, assumptions, and correct execution. A precise calculation using the wrong model can still produce a poor decision.

Apply: a new sorting rule repeatedly fails. What should you inspect?

Check whether you understood the feedback and goal, whether the environment changed, whether your old strategy is being reused automatically, and whether another representation fits better. Persistent errors alone do not identify one neurological cause.

Review tomorrow and in a week. Score each explanation: 0 for a label, 1 for a definition, 2 for a mechanism plus example and limitation. Repair the weakest explanation first.

One-minute explanation: Thinking depends on how a problem is represented and which strategies receive priority. Familiarity and memory make efficient judgments possible but can bias probability estimates and hypothesis tests. Checking sets, counterexamples, and alternative cases improves reasoning. Executive networks help maintain goals and update strategies, while valuation and bodily responses interact with learning. No single task or brain region explains the whole process.

Sources and further reading

The supplied outline’s citation fragments did not include a bibliography. These primary studies support the judgment and decision-making examples; the cinema and team scenarios are teaching aids.

1. Tversky, A., and Kahneman, D. (1983). Extensional versus intuitive reasoning: the conjunction fallacy in probability judgment.
2. Tversky, A., and Kahneman, D. (1973). Availability: a heuristic for judging frequency and probability.
3. Bechara, A., Damasio, H., Tranel, D., and Damasio, A. R. (1997). Deciding advantageously before knowing the advantageous strategy.
4. Maia, T. V., and McClelland, J. L. (2004). A reexamination of the evidence for the somatic marker hypothesis.