A Deep Active Inference Model of the Rubber-Hand Illusion

Rood, Thomas; Gerven, Marcel van; Lanillos, Pablo

doi:10.1007/978-3-030-64919-7_10

Cited by 16 publications

(11 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Future research needs to address the computational model behind the dynamics of illusory experience itself and more realistic visual input 59 . One possibility could be to integrate, in a hierarchical fashion, our AIF model with Bayesian causal inference models previously proposed for body ownership illusions 8,42 .…”

Section: Discussionmentioning

confidence: 99%

Active strategies for multisensory conflict suppression in the virtual hand illusion

Lanillos

Franklin

Maselli

et al. 2021

Sci Rep

Self Cite

View full text Add to dashboard Cite

The perception of our body in space is flexible and manipulable. The predictive brain hypothesis explains this malleability as a consequence of the interplay between incoming sensory information and our body expectations. However, given the interaction between perception and action, we might also expect that actions would arise due to prediction errors, especially in conflicting situations. Here we describe a computational model, based on the free-energy principle, that forecasts involuntary movements in sensorimotor conflicts. We experimentally confirm those predictions in humans using a virtual reality rubber-hand illusion. Participants generated movements (forces) towards the virtual hand, regardless of its location with respect to the real arm, with little to no forces produced when the virtual hand overlaid their physical hand. The congruency of our model predictions and human observations indicates that the brain-body is generating actions to reduce the prediction error between the expected arm location and the new visual arm. This observed unconscious mechanism is an empirical validation of the perception–action duality in body adaptation to uncertain situations and evidence of the active component of predictive processing.

show abstract

Section: Discussionmentioning

confidence: 99%

Active strategies for multisensory conflict suppression in the virtual hand illusion

Lanillos

Franklin

Maselli

et al. 2021

Sci Rep

Self Cite

View full text Add to dashboard Cite

show abstract

“…We first elaborate on intentionality in Active Inference, then provide a proof-of-concept agent endowed with visual input, and finally describe and analyze several simulations of visually-guided behavior. The theoretical work is motivated by basic research showing the critical role of the PPC in goal-directed sensorimotor control through intention coding [3, 16, 11] and extends previous theoretical and applied research on Active Inference [52, 53, 54] and VAE-based vision support [55, 56, 57]. In turn, the simulations were inspired by a monkey reaching task [58].…”

Section: Resultsmentioning

confidence: 99%

Flexible Intentions: An Active Inference Theory

Priorelli

Stoianov

2022

Preprint

View full text Add to dashboard Cite

We present a normative computational theory of how neural circuitry in the Dorsal Visual Stream (DVS) and the Posterior Parietal Cortex (PPC) support visually guided goal-directed actions in a dynamical environment through flexible goal-encoding intentions. It builds on Active Inference, in which perception and motor control signals are inferred through dynamic minimization of generalized prediction errors. The PPC is proposed to maintain a belief over bodily state and relevant objects on the scene, and by manipulating these states, it is involved along with visual, frontal, and motor areas in dynamically generating goal-directed actions including motor planning and movement control. Specifically, the PPC is viewed as a dynamic system computing flexible motor intentions and directing belief over the latent body state towards dynamic future goals, while the DVS is viewed as a generative model translating beliefs into visual sensory predictions in order to infer targets and posture. An intention-driven Active Inference agent that embodied an actuated upper limb with visual and proprioceptive sensors was able to perceive and reach visual targets, behaving correctly under various conditions, including static and dynamic targets, different sensory feedback, sensory precisions, intention gains, and movement policies. The theory is of general interest because it provides a normative basis for a variety of neurophysiological investigations of flexible goal-directed behavior and sensorimotor processes in end-to-end settings and further advances Artificial Intelligence methods for sensorimotor control.

show abstract

“…Sometimes, the rubber hand illusion has been interpreted in the context of Bayesian inference [Samad et al, 2015, Rood et al, 2020]. Bayesian inference says that the subject attempts to eliminate perceptual inconsistency by replacing it with a new hypothesis.…”

Section: Discussionmentioning

confidence: 99%

Ambiguous Body Ownership Experience Caused by Feeling–judgement Conflict: Evidence from Subjective Measurement in Rubber Hand Illusion using Integrated Information Theory

Niizato

Nishiyama²,

Sakamoto³

et al. 2021

Preprint

View full text Add to dashboard Cite

Human body awareness is malleable and adaptive to changing contexts. The illusory sense of body-ownership has been studied since the publication of the rubber hand illusion, where ambiguous body ownership feeling, expressed as “the dummy hand is my hand even though that is not true”, was first defined. Phenomenologically, the ambiguous body ownership is attributed to a conflict between feeling and judgement; in other words, it characterises a discrepancy between first-person (i.e. bottom-up) and third-person (i.e. top-down) processes. Although Bayesian inference can explain this malleability of body image sufficiently, the theory does not provide a good illustration of why we have different experiences to the same stimuli – the difficulty lies in the uncertainty regarding the concept of judgement in their theory.This study attempts to explain subjective experience during rubber hand illusions using integrated information theory (IIT). The integrated information Φ in IIT measures the difference between the entire system and its subsystems. This concept agrees with the phenomenological interpretation – that is, there is conflict between judgement and feeling. By analysing the seven nodes of a small body– brain system, we demonstrate that the integrity of the entire system during the illusion decreases with increasing integrity of its subsystems. These general tendencies agree well with many brain-image analyses and subjective reports; furthermore, we found that subjective ratings were associated with the Φs. Our result suggests that IIT can explain the general tendency of the sense of ownership illusions and individual differences in subjective experience during the illusions.

show abstract

A Deep Active Inference Model of the Rubber-Hand Illusion

Cited by 16 publications

References 20 publications

Active strategies for multisensory conflict suppression in the virtual hand illusion

Active strategies for multisensory conflict suppression in the virtual hand illusion

Flexible Intentions: An Active Inference Theory

Ambiguous Body Ownership Experience Caused by Feeling–judgement Conflict: Evidence from Subjective Measurement in Rubber Hand Illusion using Integrated Information Theory

Contact Info

Product

Resources

About