Insight problem solving is characterized by impasses, states of mind in which the thinker does not know what to do next. ==The authors hypothesized that impasses are broken by changing the problem representation, and 2 hypothetical mechanisms for representational change are described: the relaxation of constraints on the solution and the decomposition of perceptual chunks. These 2 mechanisms generate specific predictions about the relative difficulty of individual problems and about differential transfer effects.== The predictions were tested in 4 experiments using matchstick arithmetic problems. The results were consistent with the predictions. Representational change is a more powerful explanation for insight than alternative hypotheses, if the hypothesized change processes are specified in detail. Overcoming impasses in insight is a special case of the general need to override the imperatives of past experience in the face of novel conditions.
#references
Title: Constraint relaxation and chunk decomposition in insight problem solving
Design: have people do two blocks: in second block people got a lot btter, so probably best to consider block 1 to really feel out the ffects of problem types
Vary problem difficulty by varying constraint hardness and chunk tightness, both together and independently
If constraints are not considered explicitly, possible to over/under-constrain problem representation (through implicit reminding) (p. 1535)
Citations here…
Therefore, if problem solver is able to relax unnecessary constraints, they can find a problem representation that admits solution, and overcome the impasse
Encountering a novel problem activates some chunks from past experiences. Not all of them will be helpful; and some will be actively unhelpful. And some will need to be decomposed (compression reversed??)
Decomposing is harder for “tighter” chunks: where the components are not (obviously) meaningful units on their own
Thus, problems with tight chunks are harder to solve.
Evidence for this: when they made it harder to decompose chunks (even with identical constraint types), people were about half as likely to solve the problem (p. 1540).
“chunks” for Knoblich are not similar to Design Patterns or Frames - they are more like visual features, that exist at multiple levels of hierarachies. (cf.
Knoblich characterizes Chunks as #> “patterns that capture recurring constellations of features or components” (p. 1535)
Which itself goes back fruther to de Groot, back in the mid-1960’s
Be careful, though! Their idea of Chunking is closer to the idea of compression - packing a huge amount of information into a smaller unit, abstracting away a lot of things.
Observed some transfer effects across blocks: once you see a constraint relaxed or decompose a chunk, a later problem that .
Somewhat anecdotal in Experiment 1, and directly validated in Experiment 2
