Pyotr Anokhin’s Systems Approach in the Digital Era
At first glance, the Theory of Functional Systems (TFS), developed by Pyotr Anokhin, belongs to the realm of "old" physiology. In reality, it proves to be closer to modern artificial intelligence, control theory, and cybernetics than to classical textbooks on reflexes. We regard TFS not as a historical curiosity, but as a rigorous model of how a system acts purposefully—specifically, how it achieves a result and maintains its stability. This model is inherently scalable: the same logic applies to an individual neuron, an entire nervous system, an artificial network, a hospital, a city, or a planetary network. In other words, we can describe a hospital or a city using the same conceptual apparatus as a single neuron.
The Rejection of Linear Determinism
For a long time, the primary framework of neurophysiology and its application to other sciences was the reflex—the "stimulus-response" mechanism. There is a stimulus, followed by a response; it appears simple and elegant. This picture accurately describes very brief and primitive reactions, but it collapses the moment we attempt to explain complex actions: planning, learning, errors, shifts in strategy, and long-term goals.
The reflex arc is always a response to what has already occurred. It depicts the organism as something that "waits for an impact" and then reacts. In such a model, there is no place for the future—only for a reaction to the past.
Anokhin proposed a different way of viewing behavior. He argued that the stimulus is not the primary factor; the primary factor is the result toward which the system strives. Instead of the linear "impact-reaction" line, he draws a closed functional ring: needs, motivation, context, memory, result prediction, action, result control, and correction. Thus, behavior is not an automatic reaction to the past, but a continuous verification of hypotheses about the future: "If I act this way, will I achieve the result the system requires to preserve itself?"
The Functional Ring and Self-Learning
We replace the reflex arc with the functional ring. At the center of this ring is the principle of "result-feedback." The system first forms an image of the result (what must happen to consider the action successful), and only then selects the means, acts, measures how closely the actual outcome matches the planned one, and corrects itself.
This implies that an intellectual act is not a "one-off response," but a multi-stage cycle of self-verification:
Action → Verification → Correction → New Action
Error here is not a failure, but fuel for learning. The discrepancy between the expected and the factual triggers a restructuring of connections. This is the very principle upon which the plasticity of the biological brain and the adaptability of the most advanced neural networks are based. In machine learning terms, this is the analogue of "reward/penalty": if the result is better than expected, the strategy is reinforced; if worse, it is modified.
TFS as an Algorithm for Natural Intelligence
In the digital era, this logic becomes particularly valuable. If we want to create agentic systems—programs, robots, or infrastructural "brains" capable of independent action—it is insufficient for them to merely react to commands or events. They require an internal mechanism that allows them to constantly verify their course: "Does this action lead to the result I must maintain?"
Modern Reinforcement Learning (RL) algorithms, predictive control models, and "adaptive critics" are all technical embodiments of the same idea: there is a model of the world, a target result, an action, an error between the expected and the obtained (in the form of a reward or penalty), and an update to the strategy.
In the language of TFS, this cycle is structured through specific sequential stages. It begins with afferent synthesis, which involves the collection and analysis of all available internal and external information. This leads directly to decision making, where a single course of action is chosen. Next, the system establishes an acceptor of action results, creating a predictive model of the outcome to serve as an internal standard. This is followed by efferent synthesis to implement the action. Once the action is taken, the system relies on return afferentation to receive data about the actual result. Finally, a stage of correction modifies the action or the system itself based on the mismatch error.
Thus, TFS becomes the "stability algorithm" not only for biology but for Natural Intelligence as a whole—from an individual cell to the future Bio-Digital Noosphere. This hypersystem must similarly form a result image, act, measure deviation, and correct itself to preserve its integrity within a changing Universe.
You can learn more by reading our e-book or listening to our audiobook
Comments