Increased interactivity and numeracy foster Bayesian reasoning

Ideally, we should be able to reason appropriately with uncertain information. In reality, research has shown that reasoning under uncertainty is often flawed and intervention efforts designed to improve Bayesian reasoning performance have met with mitigated success. Over the years, the accumulated evidence has tended to suggest that people’s poor performance in tasks requiring them to draw inferences based on statistical information is mainly due to reliance on heuristic thinking, as a result of their general lack of numeracy skills, lower cognitive abilities, or lack of motivation to engage in effortful thinking. We surmised, instead, that participants’ struggle to engage in Bayesian reasoning, together with researchers’ mitigated success in helping participants overcome their difficulties, originates from the material resources customarily used to present information in Bayesian problems—namely, paper-and-pencil questionnaires—because such resources severely constrain what participants can do to discover the correct solution. We report a series of three experiments showing that performance can be substantially improved when material resources afford richer interactions with the probabilistic information presented in the problems, independently of the information format used and without training. We conclude by discussing the implication of adopting a distributed cognition approach to better understand how people’s actual thinking capability may be realised within and outside the laboratory.
This paper was presented on the 20th August, 2.15pm during the 24th Subjective Probability, Utility and Decision-Making conference at the IESE Business School, Barcelona.

Additional resources

Here you’ll find two examples of our recordings of participants solving a Bayesian tasks using playing cards. The first participant was successful, the second wasn’t.

Part. 27, Problem 1, Successful:

Part. 24, Problem 1, Unsuccessful: