Change point detection with multiple alternatives reveals parallel evaluation of the same stream of evidence along distinct timescales

Abstract

In order to behave appropriately in a rapidly changing world, individuals must be able to detect when changes occur in that environment. However, at any given moment, there are a multitude of potential changes of behavioral significance that could occur. Here we investigate how knowledge about the space of possible changes affects human change point detection. We used a stochastic auditory change point detection task that allowed model-free and model-based characterization of the decision process people employ. We found that subjects can simultaneously apply distinct timescales of evidence evaluation to the same stream of evidence when there are multiple types of changes possible. Informative cues that specified the nature of the change led to improved accuracy for change point detection through mechanisms involving both the timescales of evidence evaluation and adjustments of decision bounds. These results establish three important capacities of information processing for decision making that any proposed neural mechanism of evidence evaluation must be able to support: the ability to simultaneously employ multiple timescales of evidence evaluation, the ability to rapidly adjust those timescales, and the ability to modify the amount of information required to make a decision in the context of flexible timescales.

Quantifying decision-making in dynamic, continuously evolving environments

“…In particular, the well-known P300 component has long been argued to be associated with detection of statistical surprise (Donchin, 1981;Duncan-Johnson and Donchin, 1977;Mars et al, 2008;Squires et al, 1976) or, more recently, a correlate of a continuous time-evolving decision variable (Twomey et al, 2015) that is equivalent to the centroparietal positivity (Kelly and O'Connell, 2013;O'Connell et al, 2012). Our centroparietal responses to |Devidence| (fig 4b) are consistent with a changepoint detection account of continuous decision making (Booras et al, 2021), in which decision-relevant input (fig 6) is evaluated for a change in latent state from a baseline period to a response period (Nassar et al, 2019). This account would also explain why these signals are enhanced when response periods are rarer, as a large |Devidence| is more statistically surprising when response periods are rare than when they are common.…”

Dynamic Weighting of Time-varying Visual and Auditory Evidence during Multisensory Decision-making

“…3D-F been demonstrated that different evidence integration strategies can be explained by specific differences in behavioural paradigms. Among important features that influence weighting profiles are the division of evidence during a trial (Bronfman et al, 2016;Levi et al, 2018;Raposo et al, 2012) and choice expectation (Booras et al, 2021;Talluri et al, 2021). When stimulus information is equally informative throughout the trial (Kiani et al, 2008;Levi et al, 2018) and the observer is able to report the decision whenever they have solved the task, like in the paradigm used here, there is no need to integrate late information after the decision has been made.…”

“…Different strategies have been revealed for studying sensory evidence accumulation. A number of studies using fluctuating visual information (i.e., where the visual stimulus reliability changes over time) have demonstrated that observers tend to weight early sensory information more heavily than late information (Huk & Shadlen, 2005;Kiani et al, 2008;Nienborg & Cumming, 2009;Zylberberg et al, 2012;Odoemene et al, 2018;Booras et al, 2021). However, late sensory information integration strategies (Bronfman et al, 2016;Cheadle et al, 2014;Levi et al, 2018) and flat weighting profiles (Bronfman et al, 2016;Odoemene et al, 2018) have also been observed.…”