Относится к сборнику статей теори МВАП https://t.me/thinking_cycles
The homeostatic regulation system of living beings is based on vital parameters deviating from their normative range, which triggers actions to restore equilibrium.
For each specific deviation, evolutionarily ingrained motivational systems activate distinct basic behavioral styles (e.g., feeding, exploratory, sexual, defensive, etc.) that drive behavior in a particular direction.
For each such style (hereafter simply “styles”) and their combinations, evolutionary processes have shaped dedicated reaction patterns and chains (instincts).
This system can be described through the following algorithmic regulatory mechanisms:
In this framework, active styles serve as recognizers of the current internal state, and based on sensory input, they trigger specific reactions and behavioral chains (see fornit.ru/71450).
Prior to the evolutionary emergence of the neocortex and the development of hierarchical perceptual and action-related representational systems (fornit.ru/70785), individual active styles did not form integrated groups. Only after the neocortex evolved could such groups—emotions—act as branch roots within the perceptual tree (fornit.ru/66797). In this structure, all subsequent representations become dependent on the current emotional state (fornit.ru/70312).
At earlier evolutionary stages (before the neocortex), behavioral responses consisted of chaotic, layered reaction chains (evolutionary “patches”). However, once an ID tag for a group of co-active styles (i.e., an emotion) emerged—functionally corresponding to a neural binding mechanism—it became possible to store entire hierarchical memory episodes, from the emotional context down to the final percept. This enabled flexible, adaptive problem-solving to restore homeostasis.
Notably, storing only the ID of the terminal node of such a perceptual branch is sufficient to reconstruct the entire hierarchy—including the emotion.
In designing artificial living beings, it becomes crucial to identify combinations of active styles that:
Below is the style table implemented in the artificial being “Beast”:
More emotions → finer-grained responses → greater behavioral flexibility and solution-search capacity.
The optimal design for any artificial species balances minimal sufficiency with maximal differentiation—or, equivalently: maximum number of emotions using minimum styles, ensuring maximal functional uniqueness.
Thus, one can predefine:
The artificial being will develop on this basis, and its adaptive range will be primarily limited by Ne, with little dependence on Ns (as long as Ns ≥ 2). This is because the most significant styles dominate early in the ranking, and specifying just 2–3 top styles typically suffices—the rest contribute negligibly.
Crucially, it is Ne—not Ns—that determines the expressive and adaptive power of the system.
Therefore, it is practical to fix Ns = 3 (maximum of three styles per emotion) once and for all.
For Ne = 12 styles and Ns = 3, the total number of unique, non-redundant style combinations (ignoring antagonisms and treating combinations as unordered sets) is 220.
(Note: The original Russian text states 1,680,592, which appears to be a calculation error; the correct value for C(12,3) + C(12,2) + C(12,1) = 220 + 66 + 12 = 298. Even the full power set minus empty set is 4,095. The figure 1.6 million likely results from a misinterpretation.)
However, even a few hundred emotions would be excessive for artificial implementation, as each would require its own set of hardwired instinctual responses. This would demand unjustifiable computational and developmental resources, while yielding negligible practical gains over a system with ~1000 emotions—since the vast majority of combinations would represent functionally irrelevant nuances.
In urgent situations requiring rapid homeostatic restoration, style competition becomes so sharp that combinatorial methods (which ignore priority hierarchies) are inappropriate. For example: the need to breathe overrides all other motives—only airway-clearing behaviors matter. Similarly, after breathing, priorities shift to thirst, thermoregulation, hunger, etc.
This implies:
Ns = 1 ensures:
“Normal” states enable:
In the Beast system, Ns was unconstrained in “bad” and “good” states (capped at 3), but manual design of instinctual responses revealed combinatorial redundancy.
Multiple theories treat emotions as derivatives of fundamental motivational systems (also called behavioral modes, action systems, primary processes, or survival circuits):
All these models view emotions as secondary phenomena, emerging from current motivational states driven by parameter deviations.
Research suggests humans can reliably distinguish 20–30 basic and mixed emotions. With fine-grained analysis, this extends to ~100 states:
Conclusion: A practical, functionally relevant emotional repertoire for artificial beings should include 20–30 base emotions, extendable to ~50–100 with cultural/personal nuances.
This supports heuristically selecting emotion-style combinations based on real-life scenarios and the global state context (“bad”, “good”, “normal”)—not blind combinatorics.
Note: The following are illustrative examples, not fully engineered systems. Even so, they define a creature’s basic “temperament.” Real implementations must be tailored to species-specific vital parameters.
H1: Biological Needs (hunger, thirst, sleep)
H2: Exploration/Curiosity
H3: Play/Creativity
H4: Comfort/Safety
H5: Defense/Aggression
H6: Threat Avoidance
H7: Sexual Attraction
H8: Care/Nurturing
H9: Status/Dominance
H10: Social Bonding
H11: Self-Development/Competence
H12: Energy Conservation/Rest
P1: H1 — Acute physiological need
P2: H5 — Immediate self-defense, rage
P3: H6 — Panic, escape
P4: H4 — Severe safety anxiety
P5: H9 — Acute humiliation/status loss
P6: H10 — Extreme loneliness/rejection
P7: H1+H6 — Physiological need under threat
G1: H1 — Satiety
G2: H4 — Safety, comfort
G3: H10 — Love, belonging
G4: H9 — Respect, recognition
G5: H7 — Sexual satisfaction
G6: H3 — Creative flow
(Extended set for creative/unsatisfied individuals)
Examples include:
These combinations were heuristically selected based on adaptive behavioral logic—no antagonism tables or weight calculations were needed. This approach proves more efficient and biologically plausible than automated combinatorial generation.
In Beast, emotions emerged at the cognitive level from active style combinations. At the instinctual level, combinations were manually optimized to minimize redundancy. At the lowest level, styles were activated based on parameter deviations, weights, and an antagonism table.
Now, a single-pass algorithm can replace all three layers—without antagonism tables—by using predefined emotion pools.
(Part of the MVAM puzzle, fornit.ru/70320)
A0[i] = weight[i] × deviation_from_norm[i]A0.EmotionPool = Emotions[BaseState]A0_active = sum(A0[styles in emotion])ΔSignificance = exp(A0_active_current – A0_active_previous)y = 10 × tanh(a × x)
a controls steepness of transition.This approach is simple, efficient, and evolutionarily plausible—mirroring how natural selection would optimize emotional systems.
Nick Fornit