Andy Clark, “Surfing Uncertainty: Prediction, Action, and Embodied Mind” (Oxford UP, 2016)
Dec 15, 2016
auto_awesome
Andy Clark, author of 'Surfing Uncertainty: Prediction, Action, and Embodied Mind', discusses the predictive processing hypothesis and its relation to embodied cognition, attention modulation, and perceptual experience. They explore the role of top-down models, traditional debates in philosophy, and the relevance of the Cartesian evil demon in embodied cognition. They also delve into the cognitive penetrability of perception, implicit biases, and the importance of good information in minimizing prediction error.
Our brains are prediction machines, processing incoming sensory information based on expectations and using prediction errors to update subsequent predictions and guide action.
Predictive processing operates hierarchically, with multiple levels of processing incorporating predictions from different levels simultaneously, enabling detailed and rich perceptual experiences.
Attention plays a crucial role in predictive processing by modulating the flow of processing based on task demands and adjusting the precision weighting of prediction error signals.
Deep dives
The Role of Predictive Processing in Neural and Cognitive Function
The podcast episode features Andy Clark, Professor of Logic and Metaphysics, discussing his new book 'Surfing Uncertainty: Prediction, Action, and the Embodied Mind'. The book explores the predictive processing hypothesis, which states that our brains are prediction machines. According to this view, our brains process incoming sensory information based on expectations, and only prediction errors are used to update subsequent predictions and guide action. Clark discusses the theory from the perspective of embodied cognition, highlighting how it challenges traditional views of perception and action. He also examines the empirical support for the theory and its implications for epistemology, schizophrenia, and implicit bias.
The Predictive Coding Model and Perception
The predictive coding hypothesis posits that the brain's main goal is to predict the sensory flux and construct a model of the world. By predicting incoming sensory information, the brain forms a grip on how the world is structured. Prediction errors, which occur when the actual sensory input deviates from the predictions, are crucial for learning and updating the model. The podcast episode discusses the role of prediction errors in perception, using examples like sine wave speech to demonstrate how prior knowledge helps in structuring the world into a meaningful percept.
The Hierarchical Nature of Predictive Processing
Predictive processing operates in a hierarchical manner, with multiple levels of processing within the brain. Higher-level predictions are more abstract and spatiotemporally extended, while lower-level predictions deal with more specific sensory details. This hierarchical processing allows for detailed and rich perceptual experiences by incorporating predictions from multiple levels simultaneously. The episode explains that this hierarchical perspective does not dissolve the differences between different scales or levels of processing, but rather reconstructs these differences in terms of predictions and prediction errors at each level.
Attention, Precision, and Conscious Experience
Attention plays a crucial role in predictive processing by modulating the flow of processing based on the task at hand. Automatic and deliberate attention mechanisms prioritize specific predictions by adjusting the precision weighting of prediction error signals. Precision, which reflects the reliability of prediction error signals, influences their impact on ongoing processing. Conscious experience is still a topic of debate and research, but precision weighting is thought to determine which prediction error signals contribute to conscious perception. The episode acknowledges that understanding the relationship between predictive processing and conscious experience is an ongoing area of investigation.
Implications for Cognitive Bias and Schizophrenia
The podcast episode briefly addresses the potential for cognitive bias and the impact of predictive processing on conditions like schizophrenia. Predictive processing models suggest that biases can emerge from the interaction between prior expectations and sensory evidence. While biases can be mitigated through cultural constructs like science and mathematics, humans may still be prone to selectively harvesting information that aligns with their existing models. The episode also highlights the role of predictive processing in understanding the perceptual abnormalities seen in schizophrenia, where false prediction errors contribute to distorted world models. The episode notes that further investigations are needed to fully grasp the complexities of cognitive bias and mental health conditions within the framework of predictive processing.
The predictive processing hypothesis is a new unified theory of neural and cognitive function according to which our brains are prediction machines: they process the incoming sensory stream in the light of expectations of what those sensory inputs ought to be. On this view, only prediction errors are fed forward into the processing stream, and these are used to update subsequent predictions and guide action. In Surfing Uncertainty: Prediction, Action, and the Embodied Mind (Oxford University Press 2016), Andy Clark explains the theory from the perspective of embodied cognition, addressing such questions as how it alters the classical view of cognition as sandwiched between perception and action and how attention is employed to modulate the sensory flow. Clark, who is Professor of Logic and Metaphysics at the University of Edinburgh, also considers the current empirical support for the theory as well as its implications for traditional debates in epistemology, our understanding of schizophrenia and autism, and concerns about implicit bias.