Modeling burglars' occupancy decisions: fast and frugal or rational and complex?

The performance of two full-information models (i.e., Dawe's rule and Franklin's rule) and a simple heuristic model (i.e., Matching heuristic) in predicting occupancy decisions by burglars was examined. Burglars (N = 40) were presented with photographs of 20 homes (previously coded on physical characteristics such as 'open window') and asked to infer whether or not the home was occupied. Performance of each model was measured by (a) relative accuracy- the percentage of photographs for which the models correctly predicted each burglar's occupancy decisions and (b) absolute accuracy - the percentage of photographs for which the models correctly predicted actual occupancy. In terms of relative predictions, the matching heuristic was more frugal than the other two models, using, on average, 1.08 cues in comparison to all available cues. Additionally, the matching heuristic was also the most accurate model, predicting 80% of participants' decisions accurately. When modelling the actual occupancy of residences the matching heuristic was again more frugal. While Franklin's rule was slightly more accurate than both the matching heuristic and Dawe's rule, there were no significant differences between the models based on the accuracy of their absolute predictions. Overall, it appears that a simple model of decision making is able to predict residential burglars' occupancy decisions and actual occupancy states equally as well as or better than complex models