A Bayesian psychophysics model of sense of agency

Abstract

Sense of agency (SoA) refers to the experience or belief that one’s own actions caused an external event. Here we present a model of SoA in the framework of optimal Bayesian cue integration with mutually involved principles, namely reliability of action and outcome sensory signals, their consistency with the causation of the outcome by the action, and the prior belief in causation. We used our Bayesian model to explain the intentional binding effect, which is regarded as a reliable indicator of SoA. Our model explains temporal binding in both self-intended and unintentional actions, suggesting that intentionality is not strictly necessary given high confidence in the action causing the outcome. Our Bayesian model also explains that if the sensory cues are reliable, SoA can emerge even for unintended actions. Our formal model therefore posits a precision-dependent causal agency.

The evolving sense of agency: Context recency and quality modulate the interaction between prospective and retrospective processes

“…Regarding prospective processes, previous studies have shown that a dynamic, evolving context conceivably influences the sense of agency through the dynamic modulation of predictive mechanisms (Wolpert et al, 2011;Clark, 2013) instantiated in sensorimotor and executive brain networks (Gallese & Sinigaglia, 2010;Di Plinio et al, 2019b). As suggested by existing models of the sense of agency (Friston & Kiebel, 2009;Friston, 2012;Thilakarathne & Treur, 2015;Buhrmann & Di Paolo, 2017;Legaspi & Toyoizumi, 2019), these predictive mechanisms are putatively mediated by implicit sensorimotor learning (McDougle et al, 2016;Kim et al, 2019). On the one hand, we confirmed the existence of a classic, prospective sense of agency in our experiment (Haggard et al, 2002;Moore & Haggard, 2008;Voss et al, 2010;Synofzik et al, 2008Synofzik et al, , 2013.…”

“…By mediating the development of action awareness, implicit sensorimotor learning may be a fundamental process in establishing the sense of agency (McDougle et al, 2016;Kim et al, 2019). Implicit learning is an endogenous process involved in generating and updating sensorimotor predictions in the brain (Friston & Kiebel, 2009;Wolpert et al, 2011;Legaspi & Toyoizumi, 2019;Di Plinio, Perrucci, & Ebisch, submitted) that affects behavioral temporal estimates (Taylor et al, 2014;Stark-Inbar et al, 2017). Therefore, implicit learning plausibly supports a continuous, trial-after-trial update of the sense of agency.…”

Statistical Learning Model of the Sense of Agency

“…Many researchers have assumed that Bayesian inference is the fundamental and leading principle algorithm in cognitive brain function (Wolpert and Ghahramani, 2000;Körding, 2008;Sanborn et al, 2010;Pouget et al, 2013;Penny, 2015;Legaspi and Toyoizumi, 2019), known as a "Bayesian coding hypothesis" or a "Bayesian brain hypothesis" (Knill and Pouget, 2004). Gershman (2019) summarizes the hypothesis as follows: (1) the brain is equipped with an internal model, i.e., the likelihood function and prior distribution, and (2) Bayes' conditionalization or its approximations work to update the prior distribution.…”

“…While this approach would definitely help the SoA studies, to the best of our knowledge, very few studies have proposed a mathematical formulation of the generative process of the SoA except for the work by Moore and Fletcher (2012) and Legaspi and Toyoizumi (2019). Legaspi and Toyoizumi (2019) proposed a concrete and analyzable formula for the "Bayesian cue integration model" proposed by Moore and Fletcher (2012), which had been described only in the abstract formula. In their model, cue integration corresponds to multi-modal sensory integration, e.g., audiovisual integration.…”

“…As noted by the authors, their hypothesis is that Bayesian cue integration is the "general principle" behind the SoA generation. One of the important contributions by Legaspi and Toyoizumi (2019) was a proposal of a quantitative definition for the SoA. According to our understanding of their work, at least three important issues exist regarding their model to be discussed.…”