MAIP 6 Analogical Problem Solving PDF

Preview

Summary

Memory, Attention and Information Processes Analogical Problem Solving ...

Description

MAIP%Notes% % Lecture%6%:%Analogical%Problem%Solving% % Duncker – problem is having a goal but not knowing how to reach it. Broadly – all inclusive, perceiving, learning. Narrowly – several steps, cautious, planning, well-defined goal. Chess-state space analysis for computers. Newell and Simon – state space – set of states (problem space) – initial state, goal state, set of operators. Solution is set of operators that moves from initial state to goal state. State spaces can grow exponentially. Exhaustive search: number of operators N, length of solution L, number of possible paths is NL – combinatorial explosion. This may occur in problems after just the first few moves, so exhaustive search unlikely. 4 strategies: depth first, breadth first, hill climbing, analogical problem solving. Hill-which is best, and then which is best. James – we reason via analogies. Spearman – all intellectual acts use analogical reasoning. Reed et al. – used puzzles as required no background knowledge, solved quickly and easy to formalize. Missionaries and Cannibals/Jealous Husbands – solution and number of moves – about 17 – asked to solve for first time, after passing twice, and with instruction. Crucially, for JH test with instruction did not change time nor number of moves. Kotovsky – Chinese ring puzzle – slide rings onto horizontal bar – after two hours playing with it nobody worked it out. Digital isomorph version on computer – majority solved it in 10-25 mins. This may be because when holding difficult to define state space as variable. Analogical transfer only worked in digital condition. Still, only 30 possible states. Holyoak (1984) – analogical problem solving – construct mental representations of source and target – select source as potentially relevant analogue to target – map components of source and target – extend mapping to generate solution to target. Duncker – tumour laser problem surround. Insight problem. Gick and Holyoak (1980) – told war story – attack/dispersion story, open supply story or tunnel story. Told participants to use story for the tumour one. Spontaneous use of prior analogy sometimes – only 20% in no-hint control group. People remember how to do first but see 2nd as superficially different. See transfer from hard onto easy, but none from easy to hard. Problem of inert knowledge, applying known things to other fields – NSF report science and engineering requires knowledge from other fields to find innovative solutions – Schunn et al. Yet task demands prevent this.

Holyoak and Koh – discussed tumour problem with some not others. Then given flashbulb question – similar, 3-7 days later. Analogical transfer can occur spontaneously – 81-86% correct, 10% correct in control. Yet same context/experimenter may trigger memories. Perhaps greater surface similarity. Then differed story in surface similarity – high showed 69% correct, low surface similarity showed 1338% (depending on structural similarity). Changed context, increased delay – superficial similarity. When called ‘ultrasound’ with different name had best performance. Chi et al. (1981) – Solomon (1997) – wine categorization by wine drinkers vs social drinkers – base on taste or other extraneous factors. Novick – experts and novices – gave baseline random practice, or remote analogue condition which had 1 in 4 of practice questions were remote structural analogue of target. Only experts used it – from 6.3 to 56.3% correct. Another test replaced condition with superficial with no structural similarity to target – both went up, expert impaired more – novice was not impaired by use of inappropriate analogue. This is when problem looks the same as target problem – both distracted. When both embedded in questions, novices used more negative transfer, whilst experts made high attempt of positive transfer. Medin and Ortoney – things that look alike are alike – psychological essentialism – constraining relation between deep properties and superficial ones. Surface matches not reliable in abstract things like algebra. Analogies good for how stuff works – electricity flowing like water. Gentner et al. – surface terminology hyp vs generative analogy hyp – grouped people according to spontaneous use of water or crowd analogy for explaining electricity. Then asked how current in two circuits compare. Use of different analogies led to differences in patterns of inferences from target domain. – Analogies are generative, not just descriptive. Spiro et al. – rowing like muscle sarcomere, or turnbuckle analogy. Orgill et al. looks at use of analogies in biochem textbooks – common. Dunbar – progress of 4 molecular labs over 1 year – ¾ uses analogies – up to 22 in a meeting – were more successful. 60% when problems arise, analogies are useful. Green et al. – cross-domain analogies (nose is to scent, as antenna is to…) – frontopolar activated fMRI (but may depend on difficulty of task). Gillan et al. – could train monkey to answer simple problem – A is to B as C is to D/E straight away after analogical training, could then do key

is to lock as … Also solved letter is to pencil, as tin is to brush (rather than tin opener). Extent of transfer limited by task difficulty/demands/expertise/surface and structural similarity....