Analogical Problem Solving

· people may use information about related problems during problem solving

· Gick & Holyoak (1980)

· subjects read an analogy story about a solution to a problem

· then are given a problem that can be solved with an analogous solution

· does the analogy story help?

· Experiment 1

· read one of 3 stories (demonstrating 3 types of solution)

· solve radiation problem

· encouraged to use analogy story

· 3 solution types: dispersion, open supply, tunnel

· results: subjects more likely to propose solution used in analogy story

· people can use solutions from analogous problems

· how do people use analogy in problem solving?

1. form a representation of problems

2. map the representation of the story onto the problem

· based on similar relations, not similar elements

3. use the mapping to generate the solution

· evidence for mapping

· verbal protocols

· evidence of mapping, although amount of mapping varies across subjects

· e.g., whole story vs. solution

· Experiment 2

· compare two stories that both have dispersion solution

· parade story: fewer correspondences with radiation story

· attack story: more correspondences

· result: attack story leads to more dispersion solutions

· people don’t always use the analogy if not instructed to

· Experiment 4:

· subjects read attack story

· half of subjects given hint to use story; half not

· story embedded in other, unrelated stories

· results:

· Hint condition: 92% dispersion solution

· No Hint: 20% dispersion solution

· analogical problem solving is “neither automatic nor invariably applied”

Why don’t people spontaneously use the analogy story?

· the story and the problem are formally analogous, but their content is quite dissimilar

· the story and the problem may be stored in a content-dependent way

· not as abstract problem type

· Ross (1984)

· subjects learned to solve probability problems

· different principles -- e.g., permutation

· taught using golf tournament example

· later, solve new problems using those principles

· problem types:

1. same story line as study (e.g., golf)

2. novel story line

3. same story line as used for different principle

· results:

1. same story line: 77% correct

2. novel story line: 43% correct

3. other story line: 23% correct

· subjects retrieved superficially similar problems to generate solution

· this is often a useful thing

· people can retrieve past solutions to formally similar (superficially dissimilar) problems under some conditions

· e.g., after comparing a number of instances of the same problem type

· memory for past problems in AI: case-based reasoning

· rely on knowledge of how similar problem was solved

· steps:

· case selected from memory

· selected case adapted to fit present problem

· adapted solution applied and evaluated

· memory is updated

· example: appraisal of antiques

Incubation Effects

· struggle with a difficult problem ® set it aside for some time ® come back to it ® solve it quickly

· Silveira (1971): cheap-necklace problem

· 3 conditions:

· control

· half-hour break

· four-hour break

· results:

· control: 55%

· half-hour: 64%

· four-hour: 85%

· not evidence of unconscious problem solving!

· verbal protocols: people don’t come back to problem with solutions worked out

· possible explanation: set effects

· people initially use inappropriate methods

· with break: forget those methods