Chapter 1 Decomposition Patterns in Problem Solving

Abstract The origin of biases in problem solving is analysed in the context of games and puzzles. To discover a game strategy, individuals decompose the problem according to their “intuitions”, i.e. their ways to categorize and conceptualize the game's properties. It is shown that, according to Bellman's principle, by decomposing a problem into subproblems and optimizing each of them, normally players do not achieve an optimal solution of the original problem. The global solution to a problem may therefore be suboptimal, in relation to the pattern of decomposition that has been adopted. Therefore, biases are interpreted as suboptimal behaviour originated by the decomposition pattern that individuals adopt, and ultimately by their categorization of the problem. After introducing a simplified version of Rubik cube, two decomposition patterns are identified, one of which maintains optimality while the other is weakly suboptimal. It is shown, with preliminary experiments, that individuals who have discovered the weakly suboptimal solution behave in a fully coherent way with their strategy, and therefore that errors can be explained as rational behaviour within an incomplete representation.