Wednesday 25 October 2017, 1.00PM to 2.00pm
Speaker(s): Terri Kneeland (UCL)
Host: Makoto Shimoji
Abstract: It is well-understood that bounded reasoning about rationality can have important behavioral consequences. The literature has typically viewed such bounds as an artifact of the difficulties of interactive reasoning---i.e., the difficulties of reasoning through sentences of the form ``I think that you think that I think…” However, in principle, bounded reasoning about rationality need not be determined by such limits in ability: Subjects may not be willing to believe their opponent is ``rational, believes rationality, etc…,'' even if they are fully capable of doing so. Is bounded reasoning about rationality entirely determined by limited ability to engage in interactive reasoning? To address this question, we develop a novel identification strategy based on disentangling rationality bounds from, what we call, cognitive bounds. If we identify a gap between the subject's rationality and cognitive bounds, then bounded reasoning about rationality is not entirely determined by limits in ability. We use Kneeland (2015) experimental dataset to show that such a gap exists. Moreover, we show that, in that case, non-degenerate beliefs about rationality are an important determinant of behavior. We argue that this has important implications for out-of-sample predictions.
Location: ARRC Auditorium (A/RC014)
Admission: All welcome