Genetics-based inductive inference model to represent human decision strategies in a supervisory control task

This paper presents a genetics-based inductive inference model to represent human decision strategies in a supervisory control task. It has been suggested that, as task complexity increases, human decision strategies tend to shift from compensatory to noncompensatory ones. Therefore, the key feature of the model is a robust inductive system that utilizes noncompensatory strategy rules. The model utilizes a multi-objective optimization learning method, and seeks to optimize genetic rule set fitness along three dimensions: completeness, specificity and parsimony. Results of model performance on human data collected in a dynamic command-and-control environment suggest that the model is capable of differentiating decision strategies. Implications of model application in other domains are also discussed.