糖心原创

Is bounded reasoning about rationality driven by limited ability?

Rationality and common belief of rationality (RCBR) is a standard benchmark in game theory. Yet, a body of experimental research points to departures from RCBR. These RCBR departures are typically viewed as an artifact of limits in the ability to engage in interactive reasoning, i.e., to reason through sentences of the form “I think, you think, I think, etc ...” We provide a novel identification strategy to test the hypothesis that RCBR departures are determined by limits in interactive reasoning. It benefits from not relying on auxiliary measures of “ability” or “sophistication” that can capture distinct concepts. We conduct an experiment based on this identification strategy and show that at least 60% of subjects have RCBR departures that are not an artifact of limited ability to engage in interactive reasoning. Moreover, the experiment provides insight into how subjects reason when they depart from RCBR. The results suggest that subjects’ reasoning depends on certain natural heuristics.