Interoceptive Attention as Dynamic Homeostatic Prioritization in a Foraging Agent
Abstract
Biological systems must regulate competing needs under limited perceptual bandwidth, where sharpening one estimate costs the capacity to sharpen the others. Any fixed-budget system therefore has to decide where to allocate its perceptual precision. We study this in a foraging agent that must keep several bodily needs satisfied to survive, modelled with active inference. At each step it reads its own body-state beliefs, identifies the most-needed channel, and reallocates a fixed budget of interoceptive precision toward it, so that the same precision-shaped likelihood feeds both belief update and planning. In AffectWorld, a four-channel foraging gridworld, this selective allocation more than doubles learning-phase survival at matched budget against a uniform-precision agent ($0.414$ vs $0.199$ across 11 layouts, n=32 seeds each, paired cluster-bootstrap p ≤ 10−4). Two further results sharpen the mechanism. The benefit runs through planning as well as perception, since denying the shaped likelihood to the planner alone removes about half of it. It is also need-aligned, since aiming precision at the least-needed channel does worse than spreading it evenly. The attended channel additionally learns its own dynamics about twice as fast, and stays ahead even at matched observation count, a behavioural trace of the same precision routing, visible in learning speed, not survival.
Related
- Dynamic workforce modulation and foraging efficiency in Eusocial insect colonies
- Dynamic hypernode injection and graph attention fusion for multimodal sentiment analysis of behavioral signals
- Cross-attention encoding models reveal dynamic spatiotemporal routing across human higher visual cortex
- Cell-type specific and nanoscale control of hippocampal presynaptic homeostatic plasticity
- Parvalbumin interneurons and dentate gyrus homeostatic dysregulation shape epileptogenesis in Angelman syndrome model mice
BibTeX
@misc{grimbly2026interoceptive,
title = {{Interoceptive Attention as Dynamic Homeostatic Prioritization in a Foraging Agent}},
author = {St John Grimbly and Nicolas Kuske and Evert A. Boonstra and Bruce A. Bassett and Charel van Hoof and Rowan Hodson and Benjamin Rosman and Ryan Smith and Mark Solms and Jonathan P. Shock},
howpublished = {arXiv},
year = {2026},
month = aug,
doi = {10.48550/arxiv.2608.04232},
eprint = {2608.04232},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2608.04232},
}
RIS
TY - UNPB
TI - Interoceptive Attention as Dynamic Homeostatic Prioritization in a Foraging Agent
AU - St John Grimbly
AU - Nicolas Kuske
AU - Evert A. Boonstra
AU - Bruce A. Bassett
AU - Charel van Hoof
AU - Rowan Hodson
AU - Benjamin Rosman
AU - Ryan Smith
AU - Mark Solms
AU - Jonathan P. Shock
PY - 2026
DA - 2026/08/04
PB - arXiv
DO - 10.48550/arxiv.2608.04232
UR - https://arxiv.org/abs/2608.04232
AB - Biological systems must regulate competing needs under limited perceptual bandwidth, where sharpening one estimate costs the capacity to sharpen the others. Any fixed-budget system therefore has to decide where to allocate its perceptual precision. We study this in a foraging agent that must keep several bodily needs satisfied to survive, modelled with active inference. At each step it reads its own body-state beliefs, identifies the most-needed channel, and reallocates a fixed budget of interoceptive precision toward it, so that the same precision-shaped likelihood feeds both belief update and planning. In AffectWorld, a four-channel foraging gridworld, this selective allocation more than doubles learning-phase survival at matched budget against a uniform-precision agent ($0.414$ vs $0.199$ across 11 layouts, n=32 seeds each, paired cluster-bootstrap p ≤ 10^−4). Two further results sharpen the mechanism. The benefit runs through planning as well as perception, since denying the shaped likelihood to the planner alone removes about half of it. It is also need-aligned, since aiming precision at the least-needed channel does worse than spreading it evenly. The attended channel additionally learns its own dynamics about twice as fast, and stays ahead even at matched observation count, a behavioural trace of the same precision routing, visible in learning speed, not survival.
ER -