Abstract:

Automatic item generation is a principled approach to item design that is grounded on cognitive theory of item solution, of which a key step is to identify and extract task features systematically via principled structure analyses of the items. The current study aims to develop such a principled structure analysis method of algebra story problems that serves the purpose of automatic item generation by synthesizing four structure analysis methods that are currently available in the literature, that is, propositional analysis, network language analysis, relation-functional analysis and task analysis map. It is shown that task features extracted from the first three methods correspond to the three types of interim mental representations formed during algebra story problem solving, that is, semantic structure, situational model and problem model, respectively, whereas those from the fourth reveal the cognitive demands at various stages of problem solving. However, further studies are required to investigate the psychological reality, thoroughness, and the range of applicability for such methods before an integrated method of structure analysis of algebra story problems can be formulated for automatic item generation.

Key words: automatic item generation, algebra story problem, cognitive model, principled structure analysis method