The Existential Limits of Reason. Vladislav Pedder

      However, theories similar to predictive coding began to actively develop only in the late 20th and early 21st centuries. A key role in this was played by research into neuroplasticity and the brain’s adaptive mechanisms. Neurobiological studies, including investigations of neurotransmitters such as dopamine and the influence of neural networks, allowed for significant insights into how the brain uses prediction and models to perceive the surrounding world. Founders of predictive coding theory, such as Karl Friedrich von Weizsäcker and Gregory Hooper, proposed that the brain is constantly forming hypotheses about the future based on past experience and correlating them with incoming sensory information.

      Bayes’ theorem, proposed by the English mathematician Thomas Bayes in the 18th century, became an important mathematical tool for analyzing and updating probabilistic hypotheses in light of new data.

      The essence of the theorem is that it allows for recalculating the probability of a hypothesis based on new data. Bayes’ theorem describes how the belief (or probability) in a hypothesis is updated in response to new information. In the context of the brain, this theorem can be used to explain how neural networks update their predictions about the future, considering both old and new experiences.

      In the context of predictive coding theory, this theorem and formula illustrate how the brain updates its hypotheses (or predictions) about the world based on new sensory data. When the brain encounters new events (data), it revises its prior probability (predictions) to incorporate these data, which helps improve the accuracy of future predictions.

      Thus, this process reflects a key feature of predictive coding: the brain does not simply react to data, but actively revises its expectations based on new inputs, always striving to minimize prediction errors.

      The application of Bayes’ theorem to neurobiology and cognitive science became possible in the 1980s when scientists began to understand how the brain could use probabilistic methods to solve problems of uncertainty. In this paradigm, the brain is seen as a “Bayesian inference” (interpreter) that formulates hypotheses about the world and updates them in response to sensory information using principles of probability. The Bayesian model suggests that the brain maintains probabilistic models of future events and adjusts them based on prediction errors, which is directly connected to the theory of predictive coding.

      This updating of probabilistic hypotheses is crucial because it allows the brain not only to adapt to changes in the environment but also to account for uncertainty in the world, even when information is incomplete. In this sense, Bayes’ theorem and its applications have become fundamental to understanding how the brain, when faced with uncertainty, can improve its predictions and forecast the future based on prior knowledge.

      Thus, the connection between predictive coding theory and Bayes’ theorem became a key point in the development of neurobiological models explaining how the brain processes information and uses probabilistic computations to predict the future. Bayes’ theory, as the foundation for handling uncertainty and adaptation, provided an important mathematical and cognitive tool for understanding how the brain functions in the context of constant uncertainty and the ever-changing world.

      Predictive Coding as an Adaptive Mechanism

      The principle behind the theory of predictive coding is that the brain does not simply react to external stimuli, but actively predicts them using existing models of the world. The brain constructs hypotheses about what will happen in the future and compares them with current sensory information. If the predictions match reality, the prediction error is minimized, allowing the brain to use its resources efficiently. If an error occurs – when there is a mismatch between the prediction and reality – the brain updates its models of the world, which helps improve perception and adaptation.

      This approach allows the brain to save energy and effort by minimizing the need to process all information from scratch. Instead of interpreting data anew each time, the brain works with simplified models that it constantly updates based on new sensory data. This significantly speeds up information processing and reduces energy expenditure. For example, when a person is walking down the street, their brain does not analyze each step individually but simply uses its predictions about what should happen in the next second.

      Predictive Coding operates at different levels, ranging from simple sensory signals (such as sounds or colors) to complex social interactions and abstract ideas. At lower levels, the brain predicts basic sensory signals, such as shapes and movements, while at higher levels, it predicts more complex phenomena, such as people’s intentions or social interaction scenarios.

      The Role of Hormones, Neurotransmitters, and Microbiota in Prediction

      The effectiveness of predictive coding mechanisms also depends on various external and internal factors. Hormones, neurotransmitters, gut microbiota, and injuries can significantly influence the brain’s ability to predict and adapt.

      Cortisol, the stress hormone, can impair the brain’s ability to adjust its predictions. For example, high levels of cortisol may disrupt the process of updating the world model, leading to persistent perceptual errors and increased anxiety. Neurotransmitters such as dopamine play a key role in reward and motivation processes, as well as in strengthening or weakening certain brain predictions. Recent studies have also shown that gut microbiota can influence cognitive functions and even the brain’s predictive abilities, as microbes interact with the central nervous system, affecting our mood and perception.

      Injuries, especially brain injuries, can disrupt the neurobiological processes of prediction, leading to cognitive and emotional disorders. For example, depression and anxiety disorders can be associated with disruptions in the mechanisms of predictive coding, when the brain cannot effectively update its world models.

      Modern brain research shows that the mind actively creates and updates models of the world using predictive coding and Bayesian approaches.

      Predictive coding is the process by which the brain forms hypotheses about what it expects to perceive and compares these hypotheses with actual sensory information. When predictive coding results in a mismatch between the brain’s expectations and sensory input (prediction error), the brain can either update its world model or try to interpret the data through existing hypotheses. If the prediction error is too large, the brain may sometimes perceive it as reality, which can lead to hallucinations. For example, under conditions of sensory deprivation, when sensory information is insufficient, the brain may dominate with its predictions, and visual or auditory images may appear to compensate for the lack of real stimuli. In cases of excessive activation of predictions, such as during stress or neurochemical imbalances (such as excess dopamine), the brain may ignore real information and impose its own interpretation. This partially explains the hallucinations observed in schizophrenia.

      Levels of Predictive Coding:

      Low level (sensory): The brain predicts simple sensory signals (e.g., lines, colors, or sounds). For example, if you hear footsteps, your brain predicts that you will see a person.

      Middle level (perceptual): Predictions include more complex structures – images, sounds of words, or objects. For instance, seeing quick movement in the bushes, you predict that it’s an animal.

      High level (cognitive): At this level, the brain forms complex hypotheses, including social interactions and abstract ideas. For example, based on someone’s behavior, you might predict their intentions..

      Ascending and Descending Signals

      The hierarchy of information processing is based on two types of signals: