Относится к сборнику статей теори МВАП https://t.me/thinking_cycles
The homeostatic regulation system in living organisms is based on vital parameters that deviate from their normal range and require corrective actions to restore balance.
For each specific deviation scenario, evolutionarily shaped motivational systems activate distinct behavioral styles (e.g., feeding, exploratory, sexual, defensive, etc.) that guide behavior toward restoring homeostasis.
For each such behavioral style (hereafter simply “styles”) and their combinations, evolutionary processes have shaped dedicated reaction sequences (instincts).
To implement this in an artificial system, the following homeostatic regulation mechanisms can be identified, ensuring an algorithmic sequence:
In such a system, the active styles serve as recognizers of the current internal state and, in response to external stimuli, trigger specific reactions and reaction chains (see fornit.ru/71450).
Before the evolutionary emergence of the neocortex—and the consequent development of hierarchical perception/action representations (fornit.ru/70785)—individual active styles did not form coherent groups. Only with the neocortex could such groups become branches within the perception tree (fornit.ru/66797). Within each such branch, all subsequent perceptual representations become dependent on the emotional state formed by the group of active styles (fornit.ru/70312).
At this phylogenetically ancient level (pre-neocortex), behavioral responses are organized as chaotic overlays of reaction chains (“evolutionary patches”). However, with the emergence of a unique ID for each style combination group (i.e., an emotion—represented as an index in biological implementations), it became possible to store the entire hierarchy of representations—from the emotion itself down to the final perceptual episode in memory. This enabled flexible manipulation during problem-solving processes aimed at restoring homeostatic balance.
It turns out that storing only the ID of the final perceptual node in a branch is sufficient to reconstruct all intermediate representations—including the emotion itself.
In designing an artificial living creature, it becomes essential to define combinations of active styles that:
Below is the table of behavioral styles implemented in the artificial creature Beast:
The greater the number of distinct emotions, the more precisely the system can respond to environmental conditions, enabling more diverse and flexible behavioral solutions. The optimal design for any artificial living creature balances minimal sufficiency with maximum differentiation—or, equivalently, maximizing the number of emotions while minimizing the number of styles per emotion and ensuring maximal functional uniqueness.
This implies that it is entirely feasible to predefine a fixed number of emotions (Ne) corresponding to the most common life situations and fix the maximum number of styles per emotion (Ns). The creature will then develop on this basis, with its behavioral repertoire constrained primarily by Ne, and only weakly dependent on Ns—provided Ns ≥ 2 (since the most significant styles dominate, and including 2–3 is usually sufficient; additional styles have minimal impact).
Crucially, the number of emotions (Ne)—not the number of base styles (Ns)—determines the expressive and adaptive power of the system. Therefore, it is reasonable to fix Ns = 3 (maximum of three styles per emotion).
For Ne = 12 base styles and Ns = 3, the total number of non-redundant style combinations (ignoring antagonisms and treating combinations as unordered sets) equals 1,680,592.
However, assigning innate reaction patterns and instincts to each of these millions of hypothetical emotions would be computationally and temporally infeasible—and practically unnecessary. The gain in behavioral flexibility would be marginal compared to a system with just ~1,000 well-chosen emotions.
Indeed, the vast majority of these combinations would describe negligible situational nuances that do not require unique behavioral responses.
Moreover, in urgent survival situations (e.g., suffocation), motivational competition becomes so sharply prioritized that combinatorial methods ignoring this hierarchy are invalid. For instance, the need to breathe overrides all other motivations—only behaviors aimed at restoring respiration are activated. Similarly, thirst, extreme temperature regulation, and hunger follow a strict priority order.
Thus, under extreme deviation from homeostasis, Ns should be limited to 1 (the highest-priority style only). Likewise, in the “Good” state—when a need is already being satisfied and the organism is maintaining that state—Ns should also be 1. Only in the “Normal” state (or minor deviations) is it safe and beneficial to combine multiple styles.
Ns = 1 ensures:
The “Normal” state (or minor deviations) allows the system to:
In the Beast artificial system, Ns was never limited in “Bad” or “Good” states (though always ≤3), but during heuristic design of innate reactions, the excess of style combinations felt clearly redundant.
Several key theories and research programs support the view that emotions emerge from more fundamental motivational systems—though often using different terminology: motivational systems, behavioral modes, action systems, primary processes, survival circuits, etc.
Both Anokhin’s and Ukhtomsky’s frameworks treat emotions as derivatives of the current motivational state, which itself arises from parameter deviations.
Empirical studies estimate that humans can reliably distinguish 20–30 basic and mixed emotions, and with fine-grained analysis, up to several hundred emotional states.
Neuroimaging and behavioral data suggest the brain encodes emotions not as discrete categories, but as points in a continuous multidimensional space (e.g., valence, arousal, control, novelty).
However, in verbalization and recognition, humans use a limited set of stable labels—typically 30–50 in everyday life.
Linguistically, while hundreds of emotion terms exist across languages, most are contextual nuances or blends (e.g., “tоска” [Russian melancholy], “hope”, “shame”). The core emotional lexicon for native speakers is typically 20–30 concepts.
Thus, empirically, the functionally significant and discriminable number of human emotions is 20–30, extendable to ~50–100 with cultural and individual nuances.
This strongly supports designing emotion systems by selecting style combinations based on realistic life situations, with Ns adjusted according to global state: Bad, Good, or Normal (fornit.ru/70332).
(Critical deviation; single-style focus)
(Need satisfied; maintaining favorable state)
(Baseline mode; complex behavior via style combinations)
As seen, emotional architecture need not differ fundamentally between cats and humans. However, for creative or unsatisfied individuals, the set of “Normal”-state emotions can be expanded:
(Extended list of 30 human “Normal”-state emotions, N15–N30, follows with combinations like “Exploring personal limits”, “Public creativity”, “Family comfort creation”, “Strategic conflict avoidance”, etc.—all built from meaningful, adaptive behavioral style combinations.)
Notably, this heuristic approach avoids the need for explicit antagonist tables or style-weight calculations, as combinations are selected based on adaptive behavioral meaning. This intuitive, context-driven selection proves far more efficient and effective than brute-force combinatorial methods or blind generalization.
In conclusion, overly meticulous automated methods for emotion selection yield diminishing returns compared to heuristic, functionally grounded design.