Skip to content Skip to menu

Paula Parpart

My research is on integrating Bayesian inference models of cognition with simple heuristics (e.g., fast and frugal heuristics). When people make decisions under uncertainty, such as choosing an apartment to rent, one common view is that they rely on heuristic algorithms, which can be viewed as shortcuts in that they do not fully utilize all available information. Heuristics are often contrasted with full-information decision models, which make proper use of the available information. In my research, I developed a formal framework that puts both heuristics and standard regression models on equal footing by treating them as extreme cases of the same rational Bayesian model (a full-information model). This integration helps explain why heuristics can sometimes perform better than full-information models.

I use both computational modelling and behavioural experiments, and rely on several machine learning techniques such as ridge regression to characterize the formal mathematical relationship between simple heuristics and more traditional regression approaches. The implications range from new machine learning methods to the discovery of new psychological mechanisms that have previously been ignored which are situated between traditional regression approaches and frugal heuristics. A second stream of research looks at active learning and heuristics.

Contact details

Room 204B
26 Bedford Way
London WC1H 0AP

Google map


On the web