Australasian Mathematical Psychology Conference 2019

Sequential testimony is as good as independent testimony in judgments under uncertainty

Belinda Xie
Psychology, University of New South Wales
Danielle Navarro
University of New South Wales
Brett Hayes
University of New South Wales

Suppose three friends tell you one thing, but then you directly observe evidence to the contrary. How much weight do you give to the social evidence, relative to your own private data? Does this answer depend on whether the three friends had listened to each other or not? And would it differ if seven friends agreed rather than three? We used the balls-and-urns task to explore these questions. Following Whalen et al. (2017), we considered the case where helpful friends are each shown one randomly-sampled ball from the selected urn, before giving their best guess about which urn was the selected urn. We manipulated whether the friends provide independent testimony by providing their guesses based solely on their own randomly-sampled ball, or sequential testimony, in which each friend considers the guesses given by all previous friends. The participant then observed one randomly-sampled ball from the selected urn and made a judgment about which urn was the source of the draws. We summarise a series of experiments demonstrating participants’ limited ability to discriminate between independent and sequential testimony. We then discuss a preliminary attempt to model participants’ updating of beliefs after each piece of evidence. We find that participants’ behaviour deviates from the Bayesian model of learning specified by Whalen et al., (2017), and propose alternative models that may better capture experimental data.