Endogenous Satisficing: A Distributed Theory of Viability Regulation for Autonomous Survival Agents

Abstract An agent that maximizes a specified reward can exhibit sophisticated adaptation while remaining dependent on an externally defined criterion of success. Living systems present a different organizational problem: action must continually regulate the changing conditions of the acting system’s own persistence. We develop endogenous satisficing as a computational framework for this problem. The framework brings together three ingredients: viability-constrained bodily dynamics, aspiration conditions generated from interoceptive state and individually acquired evidence, and distributed competition among partially autonomous behavioral processes. Satisficing is defined operationally as the selection of an action meeting a context-dependent set of locally generated acceptance conditions within bounded sensing and computation, rather than as the global maximizer of a permanently fixed external score. We specify an executable transition system, bodily need variables, local aspiration and risk conditions, a recurrent motor-competition field, event-grounded memory with delayed credit assignment, and a minimal causal self-model criterion based on prediction intervention. Four theoretical consequences follow: internally conditioned action preferences, a distinction between resource transfer and acquisition, the causal importance of model readout, and the risk of changing a previously accurate belief under noisy evidence. Existing SISE-N artificial-survival experiments provide initial mechanistic evidence. In a common-frozen-policy, four-controller comparison over 12 paired maps, distributed controllers achieved longer observed survival than their central-acceptance counterparts, while the additional dynamic-threshold effect within the distributed architecture was small. Subsequent interventions showed that individually learned bodily-outcome predictions altered action selection, whereas injury-probability calibration and survival improvement remained separate empirical questions. The resulting theory provides an experimentally decidable account of functional autonomy grounded in an agent’s own conditions of persistence, without equating distributed control with consciousness or digital simulation with biological life. Keywords: artificial life; endogenous satisficing; viability; distributed control; homeostasis; causal self-model

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-17
DOI
https://doi.org/10.5281/zenodo.24297384
Primary Topic
Embodied and Extended Cognition
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Endogenous Satisficing: A Distributed Theory of Viability Regulation for Autonomous Survival Agents

Jizhang Huang, Huang
5 links
Zenodo (CERN European Organization for Nuclear Research)
Embodied and Extended Cognition
preprint

Endogenous Satisficing: A Distributed Theory of Viability Regulation for Autonomous Survival Agents

Jizhang Huang, Huang
preprint en

Abstract

Abstract An agent that maximizes a specified reward can exhibit sophisticated adaptation while remaining dependent on an externally defined criterion of success. Living systems present a different organizational problem: action must continually regulate the changing conditions of the acting system’s own persistence. We develop endogenous satisficing as a computational framework for this problem. The framework brings together three ingredients: viability-constrained bodily dynamics, aspiration conditions generated from interoceptive state and individually acquired evidence, and distributed competition among partially autonomous behavioral processes. Satisficing is defined operationally as the selection of an action meeting a context-dependent set of locally generated acceptance conditions within bounded sensing and computation, rather than as the global maximizer of a permanently fixed external score. We specify an executable transition system, bodily need variables, local aspiration and risk conditions, a recurrent motor-competition field, event-grounded memory with delayed credit assignment, and a minimal causal self-model criterion based on prediction intervention. Four theoretical consequences follow: internally conditioned action preferences, a distinction between resource transfer and acquisition, the causal importance of model readout, and the risk of changing a previously accurate belief under noisy evidence. Existing SISE-N artificial-survival experiments provide initial mechanistic evidence. In a common-frozen-policy, four-controller comparison over 12 paired maps, distributed controllers achieved longer observed survival than their central-acceptance counterparts, while the additional dynamic-threshold effect within the distributed architecture was small. Subsequent interventions showed that individually learned bodily-outcome predictions altered action selection, whereas injury-probability calibration and survival improvement remained separate empirical questions. The resulting theory provides an experimentally decidable account of functional autonomy grounded in an agent’s own conditions of persistence, without equating distributed control with consciousness or digital simulation with biological life. Keywords: artificial life; endogenous satisficing; viability; distributed control; homeostasis; causal self-model

Zenodo (CERN European Organization for Nuclear Research)
Embodied and Extended Cognition
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon