Tracking the contribution of inductive bias to individualised internal models

Abstract

Internal models capture the regularities of the environment and are central to understanding how humans adapt to environmental statistics. In general, the correct internal model is unknown to observers, instead they rely on an approximate model that is continually adapted throughout learning. However, experimenters assume an ideal observer model, which captures stimulus structure but ignores the diverging hypotheses that humans form during learning. We combine non-parametric Bayesian methods and probabilistic programming to infer rich and dynamic individualised internal models from response times. We demonstrate that the approach is capable of characterizing the discrepancy between the internal model maintained by individuals and the ideal observer model and to track the evolution of the contribution of the ideal observer model to the internal model throughout training. In particular, in an implicit visuomotor sequence learning task the identified discrepancy revealed an inductive bias that was consistent across individuals but varied in strength and persistence.

Tracking human skill learning with a hierarchical Bayesian sequence model

“…The other, contemporaneous alternative account [ 15 ] is mechanistic, and so is rather closer to ours. [ 15 ] fitted their model to the same data set that we present here, therefore the differences between our models and fitting procedures are straightforwardly understood and we will elaborate on them here.…”

“…The other, contemporaneous alternative account [ 15 ] is mechanistic, and so is rather closer to ours. [ 15 ] fitted their model to the same data set that we present here, therefore the differences between our models and fitting procedures are straightforwardly understood and we will elaborate on them here. They use an infinite hidden Markov model (iHMM) [ 37 , 38 ] which is a nonparametric Bayesian extension of the classical HMM, capable of flexibly increasing the number of states given sufficient evidence in the data.…”

“…Apart from the prior structure itself, our modeling approach differed from that employed by [ 15 ] in three ways. First, [ 15 ] assumed a stationary model structure for each session, while we sought to capture how participants update their internal model trial by trial.…”

“…Apart from the prior structure itself, our modeling approach differed from that employed by [ 15 ] in three ways. First, [ 15 ] assumed a stationary model structure for each session, while we sought to capture how participants update their internal model trial by trial. This is important for the initial training session and the interference sessions, where quick within-session model updating is expected.…”

“…A detailed description of the task, procedure, and participants can be found in [ 15 ] where this data was first published. In brief, we tested participants’ long-term sequence learning on a serial reaction time task with second-order dependence (the Alternating Serial Reaction Time task or ASRT [ 11 ]).…”