Trends in Neurosciences
Volume 44, Issue 8, August 2021, Pages 619-628
Journal home page for Trends in Neurosciences

Opinion
Useful misrepresentation: perception as embodied proactive inference

https://doi.org/10.1016/j.tins.2021.04.007Get rights and content

Highlights

  • Converging lines of evidence place allostatic physiological regulation as a core imperative of perception. From the perspective of predictive processing, we evaluate how such an optimisation is reflected by biases in expectations underlying perception.

  • Within perceptual inference, expectations and attention are often assigned functionally orthogonal roles, whereby attention alters the gain on sensory signals based upon their estimated utility, while expectations bias perception according to regularities in the environment.

  • We argue that expectations are utility driven, biasing perceptual inference towards useful outcomes that may be inaccurate but, importantly, facilitate actions that preemptively avoid biologically unfavourable regulatory conditions. Internal models are thus structured, tuned, and updated in ways that usefully misrepresent the world.

According to the predictive processing framework, perception is geared to represent the environment in terms of embodied action opportunities as opposed to objective truth. Here, we argue that such an optimisation is reflected by biases in expectations (i.e., prior predictive information) that facilitate ‘useful’ inferences of external sensory causes. To support this, we highlight a body of literature suggesting that perception is systematically biased away from accurate estimates under conditions where utility and accuracy conflict with one another. We interpret this to reflect the brain’s attempt to adjudicate between conflicting sources of prediction error, as external accuracy is sacrificed to facilitate actions that proactively avoid physiologically surprising outcomes. This carries important theoretical implications and offers new insights into psychopathology.

Section snippets

Better safe than sorry: adaptive biases in perception

Imagine walking in a forest and coming across a seemingly ambiguous object lying on the ground, for instance a thin, elongated, and curved dark shape. Upon encountering this object, the task of your brain is to infer what it is so that it can formulate an appropriate behavioural response (Figure 1). According to predictive processing (PP) (see Glossary), perception of this object will entail a process of approximate Bayesian inference [1., 2., 3.], which combines sensory evidence (‘likelihood’)

Active inference, embodied prediction, and pragmatic perceivers

Biological agents are characterised by the ability to maintain themselves within a limited range of states that are compatible with continued survival [9]. This can be described in information-theoretical terms, where organisms preferentially occupy a set of high-probability states through minimising ‘surprising’ sensory exchanges with the environment. According to PP, this is achieved via the deployment of a hierarchical generative model, which minimises the prediction error (or average

Attention and expectations: are they functionally orthogonal?

Perceptual decision-making is optimised by distinct underlying computational mechanisms. Signal detection theory [23] models and distinguishes these influences in terms of sensitivity, the discriminability of a stimulus (d’), and bias, the probability of a given response (C). While both may improve perceptual accuracy through improving the detection of a target stimulus, they are dissociable in terms of their effects upon detecting a target’s absence, where increased sensitivity results in a

Expectations are utility driven: support from the empirical literature

Defending the claim that expectations are utility driven is complicated by the fact that accuracy and utility often coincide in perceptual inference. However, the probability-driven versus utility-driven models should lead to divergent predictions under conditions where accuracy and utility conflict with one another [8]. Empirically, this has been achieved by evaluating the effects of expectations in relation to ambiguous stimuli that give rise to percepts of comparable probability yet of

At what cost? Utility biases efficiently minimize interoceptive prediction error

On the surface, utility biases may seem to contradict the principle of prediction error minimisation, since they systematically lead organisms towards predictions that, on average, deviate from actual states in the world. However, they may be suitably accommodated once we revisit the notion that prediction error can arise in both the external and internal milieu. With this in mind, we view these respective models to be minimizing two conflicting sources of prediction error. On the one hand,

Concluding remarks

The core function of cognition is to successfully regulate the internal milieu [14], and perceptual models are thought to emerge as a consequence of this imperative [68]. We suggest that this optimisation is mediated by biases in the structure and tuning of predictive information in the brain, leading to internal models that usefully misrepresent the environment. Allostasis thus plays an essential role in constraining perceptual outcomes, consistent with the notion that the ‘brain recovers the

Acknowledgements

We acknowledge the Deutscher Akademischer Austauschdienst and the Berlin School of Mind and Brain for scholarship funding to J.M.M. and the Berlin Institute of Health for a Clinical Fellowship to P.S. in relation to this research. P.S. is further supported by grants from Deutsche Forschungsgemeinschaft (STE 1430/8-1 and STE 1430/9-1). Special thanks to Chiara Caporuscio, Marika Constant, and the members from the Visual Perception Laboratory for the fruitful discussions and motivational support.

Declaration of interests

The authors declare no competing interests in relation to this work.

Glossary

Active inference
a framework for incorporating the role of action into the PP framework. While perception functions to change one’s predictions of the world to accommodate the sensory evidence, action functions to change the sensory input to fulfil these predictions. Action can thus be cast as a type of ‘self-fulfilling prophecy’ where we actively engage with and sample the world to bring about the very predictions that our internal model deploys.
Allostasis
the process of preserving physiological

References (97)

  • A. Clark

    Surfing Uncertainty: Prediction, Action, and the Embodied Mind

    (2015)
  • J. Hohwy

    The Predictive Mind

    (2013)
  • P. Sterzer

    The predictive coding account of psychosis

    Biol. Psychiatry

    (2018)
  • J. Hohwy

    Predictive coding explains binocular rivalry: an epistemological review

    Cognition

    (2008)
  • D.D. Johnson

    The evolution of error: error management, cognitive constraints, and adaptive decision-making biases

    Trends Ecol. Evol.

    (2013)
  • M.G. Haselton

    The evolution of cognitive bias

  • M.G. Haselton et al.

    The paranoid optimist: an integrative evolutionary model of cognitive biases

    Personal. Soc. Psychol. Rev.

    (2006)
  • S.K. Lynn et al.

    “Utilizing” signal detection theory

    Psychol. Sci.

    (2014)
  • T. Parr et al.

    Generalised free energy and active inference

    Biol. Cybern.

    (2019)
  • L.F. Barrett

    The theory of constructed emotion: an active inference account of interoception and categorization

    Soc. Cogn. Affect. Neurosci.

    (2017)
  • M. Allen et al.

    From cognitivism to autopoiesis: towards a computational framework for the embodied mind

    Synthese

    (2018)
  • M. Allen et al.

    The body as first prior: interoceptive predictive processing and the primacy

  • M. Kirchhoff

    The body in action: predictive processing and the embodiment thesis

  • J. Schulkin et al.

    Allostasis: a brain-centred, predictive mode of physiological regulation

    Trends Neurosci.

    (2019)
  • A.K. Seth

    Interoceptive inference, emotion, and the embodied self

    Trends Cogn. Sci.

    (2013)
  • P. Sterling

    Homeostasis vs allostasis: implications for brain function and mental disorders

    JAMA Psychiatry

    (2014)
  • A.K. Seth et al.

    Being a beast machine: the somatic basis of selfhood

    Trends Cogn. Sci.

    (2018)
  • P. Cisek

    Cortical mechanisms of action selection: the affordance competition hypothesis

    Philos. Trans. R. Soc. Lond. B Biol. Sci.

    (2007)
  • J.J. Gibson

    The Ecological Approach to Visual Perception

    (1979)
  • G. Pezzulo et al.

    Navigating the affordance landscape: feedback control as a process model of behavior and cognition

    Trends Cogn. Sci.

    (2016)
  • G. Pezzulo

    Hierarchical active inference: a theory of motivated control

    Trends Cogn. Sci.

    (2018)
  • A.W. Corcoran

    From allostatic agents to counterfactual cognisers: active inference, biological regulation, and the origins of cognition

    Biol. Philos.

    (2020)
  • D.M. Green et al.

    Signal Detection Theory and Psychophysics

    (1966)
  • R. Ratcliff et al.

    The diffusion decision model: theory and data for two-choice decision tasks

    Neural Comput.

    (2008)
  • C.N. White

    Using decision models to decompose anxiety-related bias in threat classification

    Emotion

    (2016)
  • H. Feldman et al.

    Attention, uncertainty, and free-energy

    Front. Hum. Neurosci.

    (2010)
  • C. Summerfield et al.

    Expectation (and attention) in visual cognition

    Trends Cogn. Sci.

    (2009)
  • C. Summerfield et al.

    Expectation in perceptual decision making: neural and computational mechanisms

    Nat. Rev. Neurosci.

    (2014)
  • V. Wyart

    Dissociable prior influences of signal probability and relevance on visual contrast sensitivity

    Proc. Natl. Acad. Sci. U. S. A.

    (2012)
  • F.P. De Lange

    How do expectations shape perception?

    Trends Cogn. Sci.

    (2018)
  • P. Kok

    Less is more: expectation sharpens representations in the primary visual cortex

    Neuron

    (2012)
  • S. Kumar

    Encoding of predictable and unpredictable stimuli by inferior temporal cortical neurons

    J. Cogn. Neurosci.

    (2017)
  • G. Hesselmann

    Predictive coding or evidence accumulation? False inference and neuronal fluctuations

    PLoS One

    (2010)
  • M.I. Garrido

    Bayesian mapping reveals that attention boosts neural responses to predicted and unpredicted stimuli

    Cereb. Cortex

    (2018)
  • A. Alink et al.

    Can expectation suppression be explained by reduced attention to predictable stimuli?

    NeuroImage

    (2021)
  • C. Teufel et al.

    Forms of prediction in the nervous system

    Nat. Rev. Neurosci.

    (2020)
  • M. Bar

    From objects to unified minds

    Curr. Dir. Psychol. Sci.

    (2021)
  • W. Wiese

    Action is enabled by systematic misrepresentations

    Erkenntnis

    (2017)
  • M. Baltieri et al.

    The dark room problem in predictive processing and active inference, a legacy of cognitivism?

  • J. Vanrie

    Bistability and biasing effects in the perception of ambiguous point-light walkers

    Perception

    (2004)
  • G.W. Alpers

    Binocular rivalry between emotional and neutral stimuli: a validation using fear conditioning and EEG

    Intern. J. Psychophysiol.

    (2005)
  • S. Marx et al.

    Reward modulates perception in binocular rivalry

    J. Vis.

    (2015)
  • G. Wilbertz

    Reinforcement of perceptual inference: reward and punishment alter conscious visual perception during binocular rivalry

    Front. Psychol.

    (2014)
  • G. Wilbertz

    fMRI-based decoding of reward effects in binocular rivalry

    Neurosci. Conscious.

    (2017)
  • J.G. Neuhoff

    Adaptive biases in visual and auditory looming perception

  • R.E. Jackson et al.

    Evolved navigation theory and the descent illusion

    Percept. Psychophys.

    (2007)
  • B. Akdoğan et al.

    The effects of payoff manipulations on temporal bisection performance

    Acta Psychol.

    (2016)
  • E. Becker et al.

    Sensitivity and response bias in fear of spiders

    Cogn. Emot.

    (2004)
  • Cited by (10)

    • Multistability, perceptual value, and internal foraging

      2022, Neuron
      Citation Excerpt :

      Elevating visual perception from a mere sensory inference to being a key contributor to affective decision-making requires us to consider how the values of sensory objects influence what we look at and also how we interpret what we see. There are prominent examples of value-laden interpretations of images (for a review, see Martin et al., 2021); however, we focus on what we choose to look at. Indeed, it is known that different aspects of eye movements (e.g., velocity, amplitude, reaction time, and gaze duration) are modulated by various sorts of values in visual-based affective decision-making tasks (see e.g., Shams-Ahmar and Thier, 2020).

    • Knowledge-augmented face perception: Prospects for the Bayesian brain-framework to align AI and human vision

      2022, Consciousness and Cognition
      Citation Excerpt :

      Accordingly, theoretical explanations of context-effects on facial expression perception increasingly focus on predictive processing (e.g., Otten et al., 2017; Trapp, Schweinberger, Hayward, & Kovács, 2018; Pereira, Barbosa, de Haan, & Ferreira-Santos, 2019; Feldman Barrett, 2016). To reiterate the Bayesian-brain view of visual perception in a nutshell: Rather than a passive process providing us with an objective, veridical image of the physical world, perception is viewed as an active, constructive process with the purpose of providing a useful basis for adaptive behavior situated in our environment (Buzsáki, 2019; Lupyan & Clark, 2015; Seth & Tsakiris, 2018), even if that sometimes means to ”usefully misrepresent” the world (Martin, Solms, & Sterzer, 2021). In this view, the content of perception is specified through predictive processing–as the brain’s ”best guess” about the causes of its sensory input that minimizes the difference (prediction error) between actual sensory signals and the predictions based on constantly updated predictive models (Friston, 2009; Clark, 2013; Otten et al., 2017).

    • The role of conflict processing in multisensory perception: Behavioural and electroencephalography evidence

      2023, Philosophical Transactions of the Royal Society B: Biological Sciences
    View all citing articles on Scopus
    View full text