A hierarchical Bayesian model of “memory for when” based on experience sampling data

Abstract

Participants wore a smartphone, which collected GPS, audio, accelerometry and image data, in a pouch around their necks for a period of two weeks. After a retention interval of one week, they were asked to judge the specific day on which each of a selection of images was taken. To account for people’s judgements, we proposed a mixture model of four processes uniform guessing, a signal detection process based on decaying memory strength, a week confusion process and a event confusion process in which the sensor streams were used to calculate the similarity of events. A model selection exercise testing all possible subsets of the processes favoured a model that included only the event confusion model. GPS similarities were found to be the most significant predictors, followed by audio and accelerometry similarities and then image similarities.

10 Figures and Tables

Cite this paper

@inproceedings{Yim2017AHB, title={A hierarchical Bayesian model of “memory for when” based on experience sampling data}, author={Hyungwook Yim and Vishnu Sreekumar and Nathan J. Evans and Paul Garrett}, year={2017} }