Recalibrating timing behavior via expected covariance between temporal cues

Corte, Benjamin J. De; Valle, Rebecca R Della; Matell, Matthew S.

doi:10.7554/elife.38790

“…There is evidence from psychophysical tasks that our brain carries out this causal inference to achieve behavioural flexibility during multisensory integration (Körding et al, 2007;Kayser and Shams, 2015). For example, when localising auditory and visual signals we tend to fuse these when they likely originate from nearby sources, but not when they originate from disparate points in space, suggesting that the probability of fusion is determined by a higher-level inference over the probable cause(s) of sensation (De Corte et al, 2018;Körding et al, 2007;Odegaard and Shams, 2016;Noppeney, 2015a, 2015b;Wozny et al, 2010).…”

Section: Introductionmentioning

confidence: 99%

Causal Inference in the Multisensory Brain

Cao¹,

Summerfield²,

Park³

et al. 2019

Neuron

View full text Add to dashboard Cite

When combining information across different senses humans need to flexibly select cues of a common origin whilst avoiding distraction from irrelevant inputs. The brain could solve this challenge using a hierarchical principle, by deriving rapidly a fused sensory estimate for computational expediency and, later and if required, filtering out irrelevant signals based on the inferred sensory cause(s). Analysing time-and source-resolved human magnetoencephalographic data we unveil a systematic spatio-temporal cascade of the relevant computations, starting with early segregated unisensory representations, continuing with sensory fusion in parietal-temporal regions and culminating as causal inference in the frontal lobe. Our results reconcile previous computational accounts of multisensory perception by showing that prefrontal cortex guides flexible integrative behaviour based on candidate representations established in sensory and association cortices, thereby framing multisensory integration in the generalised context of adaptive behaviour.

show abstract

“…There is evidence from psychophysics that our brain indeed carries out causal inference (CI) to achieve behavioral flexibility during multisensory integration (Kö rding et al, 2007;Kayser and Shams, 2015). For example, when localizing auditory and visual signals, we tend to fuse these when they likely originate from nearby sources, but not when they originate from disparate locations, suggesting that the probability of fusion is determined by high-level inference over the probable cause(s) of sensation (De Corte et al, 2018;Kö rding et al, 2007;Noppeney, 2015a, 2015b;Wozny et al, 2010).…”

Section: Introductionmentioning

confidence: 99%

Causal inference in the multisensory brain

Cao¹,

Summerfield²,

Park³

et al. 2018

Preprint

View full text Add to dashboard Cite

When combining information across different senses humans need to flexibly select cues of a common origin whilst avoiding distraction from irrelevant inputs. The brain could solve this challenge using a hierarchical principle, by deriving rapidly a fused sensory estimate for computational expediency and, later and if required, filtering out irrelevant signals based on the inferred sensory cause(s). Analysing time-and source-resolved human magnetoencephalographic data we unveil a systematic spatio-temporal cascade of the relevant computations, starting with early segregated unisensory representations, continuing with sensory fusion in parietal-temporal regions and culminating as causal inference in the frontal lobe. Our results reconcile previous computational accounts of multisensory perception by showing that prefrontal cortex guides flexible integrative behaviour based on candidate representations established in sensory and association cortices, thereby framing multisensory integration in the generalised context of adaptive behaviour. We thank Robin A. A. Ince for support during initial development of MEG analysis pipeline,

show abstract

“…Yet the discounting of outcomes according to delay is also susceptible to task features beyond the duration of the pre-reward interval: animal preference in intertemporal choice is also sensitive to postreward delays, the delivery of additional rewards in the inter-trial interval as well as to the broader framing of the task (Blanchard and Hayden, 2015;Blanchard et al, 2013;Carter and Redish, 2016;Wikenheiser et al, 2013;Williams et al, 2017). Further, the influence of temporal delay on reward-guided behavior indicates animals and humans learn about the distribution of, and relationship between, the delay to different rewards in a task, consistent with the explicit learning of temporal durations as a property of the state representation of a task (De Corte et al, 2018;Kable, 2013, 2015;Williams et al, 2017). Understanding states as summaries over history thus naturally recruits the concept of a timescale at which the state maintains a representation of previous events: these results imply that the timescale at which a state summarizes history is a malleable, and thus learnable, part of a state representation.…”

Section: Hidden State Representations In Reward-guided Behaviorsmentioning

confidence: 65%

Uncovering the ‘state’: Tracing the hidden state representations that structure learning and decision-making

Langdon

¹

,

Song

²

,

Niv

³

2019

Behavioural Processes

View full text Add to dashboard Cite

We review the abstract concept of a 'state'-an internal representation posited by reinforcement learning theories to be used by an agent, whether animal, human or artificial, to summarize the features of the external and internal environment that are relevant for future behavior on a particular task. Armed with this summary representation, an agent can make decisions and perform actions to interact effectively with the world. Here, we review recent findings from the neurobiological and behavioral literature to ask: 'what is a state?' with respect to the internal representations that organize learning and decision making across a range of tasks. We find that state representations include information beyond a straightforward summary of the immediate cues in the environment, providing timing or contextual information from the recent or more distant past, which allows these additional factors to influence decision making and other goaldirected behaviors in complex and perhaps unexpected ways. Keywords Learning; decision making; reward; timing; dopamine Many computational theories of learning and decision-making rely on the concept of a 'state'-a representation used by an animal, human or artificial agent that describes the current situation of the agent within an environment and which the agent uses to guide their behavior. In reinforcement learning (RL), where a typical task is to choose which of multiple possible actions to perform in order to obtain a possible reward, the current state selectively includes all current and past environmental information an agent treats as relevant for making their decisions to act (Sutton and Barto, 1998). A state representation thus encapsulates knowledge about the structure of a task, providing a map of discrete states that follow one another given events that occur during the task and the actions performed by the agent. While a state is a central concept for RL algorithms, it is an idea that can be ambiguous when applied to neurobiology and behavior. Here, we aim to orient readers unfamiliar with RL to key concepts in the definition of state, and discuss the assumptions (explicit and implicit) that arise when attempting to determine the representation of state internal to an agent acting within a task environment. Through this detailed discussion of the

show abstract

Recalibrating timing behavior via expected covariance between temporal cues

Cited by 14 publications

References 67 publications

Causal Inference in the Multisensory Brain

Causal Inference in the Multisensory Brain

Causal inference in the multisensory brain

Uncovering the ‘state’: Tracing the hidden state representations that structure learning and decision-making

Contact Info

Product

Resources

About