Normative Feeling: Socially Patterned Affective Mechanisms

TL;DR

The paper investigates how normative punishment could have shaped the evolution of mood across species by embedding social regularities into affective mechanisms. It implements an agent-based, energy–based system with evolvable mood, contrasting competition (non-normative) and punishment (normative) in a shared-resource setting, and introduces an axiomatic normativity framework to identify multi-equilibria norms. The results show that punishment fosters a mood mechanism where negative affect signals social sanctioning, promoting resource conservation and preventing population collapse, while enabling mood-based signaling without costly enforcement. This work highlights a decentralized, culturally-agnostic pathway for the emergence of social preferences and demonstrates how normative processes can reprogram cognitive and physiological systems by embedding cultural patterns into psychological dispositions, with implications for understanding mood evolution and social regulation across taxa. Its framework suggests avenues for mood signaling and future work on multi-dimensional affect and topology-aware models with potential cross-species relevance.

Abstract

Breaking a norm elicits both material and emotional consequences, yet how this coupling arose evolutionarily remains unclear. We investigate this question in light of emerging work suggesting that normativity's building blocks emerged earlier in evolution than previously considered, arguing that normative processes should inform accounts of how even ancient capacities such as mood evolved. Using a definition of normative processes we developed, we created an agent-based model with evolvable affect in a shared resource dilemma, comparing competition (non-normative) versus punishment (normative) conditions. Critically, different mood mechanisms emerge under each condition. Under competition, agents evolve a "bad mood -> consume more" response, creating a tragedy of the commons leading to resource depletion and population collapse. Under punishment, agents evolve a "bad mood -> consume less" mechanism, where negative affect functions as an implicit signal of social sanction, promoting resource conservation. Importantly, once normative logic is imprinted through punishment, it creates an evolutionary pathway for mood-based signalling that operates without costly physical enforcement. Our findings demonstrate how normative processes enable social preferences to emerge in a distributed manner within psychological mechanisms, showing how normative processes reprogram cognitive and physiological systems by embedding cultural patterns into psychological dispositions.

Figures (7)

• Figure 1: An evolvable model of mood mechanism. Where different stimuli ($S$) and how they affect $Mood$ (fluctuating affective state) and how mood affects behaviour, with weights ($W$) affecting the direction and strength of these effects. Therefore the weights define agents' mood mechanism. Note: the dotted lines between $S_2$ and $S_n$ implies that there are more stimuli that can be in-between $S_2$ and $S_n$.
• Figure 2: This figure depicts the average $\mu_{eat}$ for different simulation runs over time, with the black line being the average across runs and each coloured line the average of a population within a run. Comparing the competition and punishment condition, we see that competition tends to one equilibrium (the maximum value) whereas the punishment condition results in a wide range of equilibria that cover the whole spectrum of $\mu_{eat}$. (a) Competition: In brief, in competition there is a positive correlation between avg-$W_{hunger}$ (bad mood) and avg-$W_{bite}$ indicating a "bad mood → consume more" response (b) Punishment (axiomatic normativity): In brief, there is a positive correlation between avg-$W_{energy-gain}$ (good mood) and avg-$W_{bite}$ indicating a "bad mood → consume less" mechanism. Additionally, there is a positive correlation avg-$W_{others-being-punished}$ vs. avg-$W_{bite}$ indicating that others will lower their consumption when they see others being punished.
• Figure 3: Here we see the effect of evolution on the correlations between different stimuli weights, behavioural weights which encode mood mechanism i.e. how stimuli affect mood and how mood affects behaviour respectively )as well as other evolvable agent traits) and how they are correlated with one another across populations after evolution. (a) competition (b) punishment (axiomatic normativity)
• Figure 4: This figure depicts various averages of different traits plotted against each other for the condition with punishment (normative) to see the relation between traits over different runs. Each blue circle is a different simulation run and the blue line indicates the general trend. Here we see that avg-$W_{energy-gain}$ vs avg-$W_{bite}$ is the strongest positive correlation, indicating on average that agents will eat more if they are gaining energy and east less if they are losing energy. Further we see that there's a negative correlation between avg-$W_{sanction}$ vs avg-$W_{energy-gain}$, indicating that agents will punish more (since sanction threshold is lower) when they are gaining energy. We also see negative correlation for avg-$W_{others-being-punished}$ vs avg-$W_{bite}$, indicating agents will eat less if they see others being punished. (a) avg-$W_{energy-gain}$ vs avg-$W_{bite}$ (b) avg-$W_{hunger}$ vs avg-$W_{energy-gain}$ (c) avg-$W_{hunger}$ vs avg-$W_{bite}$ (d) avg-$W_{others-being-punished}$ vs avg-$W_{bite}$ (e) avg-$W_{sanction}$ vs avg-$W_{energy-gain}$ (f) avg-$W_{sanction}$ vs avg-$W_{bite}$
• Figure 5: This figure depicts various averages of different traits plotted against each other for the condition with competition (non-normative) to see the relation between traits over different runs. Each blue circle is a different simulation run and the blue line indicates the general trend. Here we see, in contrast to the punishment condition, a strong correlation between avg-$W_{hunger}$ vs avg-$W_{bite}$ and weak correlations for avg-$W_{energy-gain}$ vs avg-$W_{bite}$ and avg-$W_{hunger}$ vs avg-$W_{energy-gain}$. This means that on average agents will eat more if they are in a hungry state and losing energy (a) avg-$W_{energy-gain}$ vs avg-$W_{bite}$(b) avg-$W_{hunger}$ vs avg-$W_{energy-gain}$ (c) avg-$W_{hunger}$ vs avg-$W_{bite}$