International Society for History, Philosophy, and Social Studies of Biology

twitter 2015
     facebook 2015


WEDNESDAY, JULY 8  /  15:30 - 17:00  /  DS-R510
Organized session / standard talks
Is the brain a prediction machine? Exploring the Bayesian revolution in neuroscience

Madeleine Ransom (University of British Columbia, Canada)

While philosophers of science and epistemologists are well acquainted with Bayesian methods of belief updating, there is a new Bayesian revolution sweeping neuroscience and perceptual psychology. First proposed by Helmholtz, predictive coding is the view that the human brain is fundamentally a hypothesis generator. On this view, the processes by which the brain tests its self-generated hypotheses against sensory evidence are seen as conforming to a hierarchical Bayesian operation; each level of the hierarchy involves a hypothesis space, with higher levels entertaining hypotheses about more complex and slower regularities as compared to the lower levels. The higher-level hypothesis spaces serve to generate and constrain the lower-level hypothesis spaces, thus enabling the lower-levels to predict the evidence. When there is a mismatch between the predicted and actual evidence, a prediction error is produced and is relayed up the hierarchy, where it is used to revise the hypothesis. Through the iterative interaction between top-down signals (which encode predictions) and bottom-up signals (which encode prediction error) the generative models that can predict the evidence most accurately are selected. Given the crucial role of sensory evidence in supervising the hypothesis testing process, it is no surprise that the view has garnered the most significant empirical support as a theory of perception. Nonetheless, increasing numbers of neuroscientists are also adopting the predictive coding framework in some capacity in order to elucidate attention, decision making, dreaming, hallucinations, felt agency, interoception and the emotions. Not since dynamical systems theory has there been a theoretical framework as popular. However, it is unclear that the success of the predictive coding theory of perception will extend to these other areas. This session will critically examine and explore the prospects of predictive coding theories of attention, dreaming and the emotions.

Three problems for the predictive coding theory of attention

Madeleine Ransom (University of British Columbia, Canada)

Attention has been a central topic of study in neuroscience and psychology due to its pivotal role in guiding perception and thought. It has also recently garnered significant interest in philosophy due in part to the close link between attention and consciousness and the lack of a coherent theory of attention. While predictive coding has most prominently offered a theory of perception, the Bayesian framework also promises to deliver a comprehensive theory of attention that falls out of the perceptual theory without the need for positing additional machinery. On this account, attention is optimization of the precision of prediction errors. In perceptual inference, prediction errors are measurements of the difference between predicted and actual sensory data. Expected precisions are a measure of how reliable, or precise, we expect the prediction error signal to be in a given context: how likely is it in a given situation that the incongruent data constitutes legitimate prediction error as opposed to noise? Optimizing expected precisions is the process of guiding perceptual inference by directing processing resources towards the prediction errors with the higher expected precisions – we attend to what is expected to be the most informative, and use this information to preferentially revise our perceptual hypotheses. I argue here that this theory of attention faces significant challenges on three counts. First, while the theory may provide a successful account of endogenous spatial attention, it fails to model feature-based attention – a central aspect of attention that any theory must explain. Second, it does not accommodate non-perceptual forms of attention such as attention to one’s thoughts. Third, it fails to accommodate the influence of affectively salient objects or high cost situations in guiding and capturing attention.

Evaluating the predictive coding model of dreaming

Sina Fazelpour (University of British Columbia, Canada)

The predictive coding framework promises the potential of a grand unifying theory in which any cognitive function can be understood on the basis of the brain's overarching function of hypothesis testing, carried out at various levels of the cortical hierarchy by a single kind of computational process with the shape of a Bayesian inferential operation. Within the hierarchically structured hypothesis space, the brain's generative model makes predictions whose probabilities are updated in proportion to how well they explain away the current sensory evidence. While the framework has proved successful in dealing with cognitive functions constrained by sensory input, it is difficult to see how it can be extended to prominent cognitive phenomena, such as dreaming, that proceed in a largely decoupled fashion from environmental stimuli, given the crucial supervisory role played by sensory input within the framework. Nonetheless, Friston and Hobson have recently proposed a predictive coding model of dreaming has been assigned the functional role of optimizing the statistical efficiency of the brain's generative model by minimizing the model's redundancy and complexity. Furthermore, the function of complexity minimization is carried out by Bayesian inferential processes aimed at explaining unpredicted oculomotor input the only sort of input available to the system during REM sleep. My aim here is to critically examine three foundational issues facing the model, with a view towards developing constructive guidelines for future research. First, at the phenomenological level, what empirically testable implications does this functional role, assuming its correctness, have for the sort of content within a dream episode Second, with regards to the processing level, are Bayesian inferential processes in general suited to the task of reducing a model's complexity. Third, is the Bayesian operation in light of occulomotor input in particular capable of delivering the assigned functional role?

Predicting emotions, emotional predicting

Mark Miller (University of Edinburgh, United Kingdom)

While predictive coding frameworks have primarily been applied to exteroceptive signals and the ways in which we model the outside world, there is a growing interest in how same functional models may be used to describe the processing of interoceptive signals. Anil Seth (2013) has recently proposed a predictive coding theory of emotion. According to the model cascading top-down predictions about the source of interoceptive signals counterflow with bottom-up interoceptive prediction errors. The integration of the various predictive representations results in the felt aspect of an emotion. The model is intended to extend traditional cognitive appraisal theories of emotion by filling out the neurocomputational mechanisms underlying the interaction between the affective (eg. neural and physiological arousal) and appraisal (eg. memories, evaluations, predictions, etc.) elements commonly considered to make up an emotional experiences (Schachter & Singer, 1962). I will argue that such an account of emotion ends up being embodied in ways that refute the cognitivist assumptions of existing appraisal theories. My argument will be based on affective neuroscience descriptions of the anterior insula (AI). Seth’s account of processing in the AI lends substantial support to recent network models of the brain (Pessoa 2014) that aim at dissolving the boundaries between emotion and cognition, and also between notions of a thinking brain and a feeling body. The aim of the paper will be to highlight the ways in which predictive coding can contribute to live debates in emotion theory, as well as suggest how affective neuroscience can in turn facilitate a better understand predictive coding theories of mind. The hope is to contribute support for a theoretical framework that bridges predictive coding and embodied cognitive science.