Adaptation-By-Proxy: Contagion Effect of Social Buffering in an Artificial Society

Abstract

The "social buffering" phenomenon proposes that social support facilitates wellbeing by reducing stress in a number of different ways. While this phenomenon may benefit agents with social support from others, its potential effects on the wider social group are less clear. Using a biologicallyinspired artificial life model, we have investigated how some of the hypothesised hormonal mechanisms that underpin the "social buffering" phenomenon affect the wellbeing and interactions of agents without social support across numerous social and physical contexts. We tested these effects in a small, rank-based society, with half of the agents endowed with numerous hormonal mechanisms associated with "social buffering", and half without. Surprisingly, our results found that these "social buffering" mechanisms provided survivalrelated advantages to agents without social support across numerous conditions. We found that agents with sociallyadaptive mechanisms themselves become a proxy for adaptation, and suggest that, in some (artificial) societies, "social buffering" may be a contagious phenomenon.

The Long-Term Efficacy of “Social Buffering” in Artificial Social Agents: Contextual Affective Perception Matters

“…Experiments were conducted in the simulation environment described in Section 2 . Following the approach in our previous work ( Khan et al, 2020 , 2019 ; Khan and Cañamero, 2021 ), we used a society of six artificial agents, where three agents share affective social bonds and the remaining three are unbonded. We have found that this group size (six agents) to be a suitable number of agents that 1) allows the society to be split into multiple (non-dyad) groups, 2) permits for the appropriate level of complexity for analysis, and 3) allows for analysis at the levels of the overall society, the sub-groups, and the individual agents (which we call the macro, meso, and micro-levels, respectively).…”

“…In all of these works, however, the effects and analysis of these social bonds have been at a dyadic level (a human-agent pair bond), as opposed to the group or social level. Attempting to address those limitations, our previous work has investigated some of the adaptive effects of affective social bonds in a society of artificial (virtual) agents (Khan et al, 2020(Khan et al, , 2019Khan and Cañamero, 2021).…”

“…Building on our own previous work ( Khan et al, 2019 , 2020 ; Khan and Cañamero, 2021 ), and also following the approach that has been used extensively in related approaches ( Lewis and Cañamero, 2019 ; Lones et al, 2017 ; Cañamero and Lewis, 2016 ; Montebelli et al, 2008 ; Iizuka and Di Paolo, 2008 ), our agent model accounts for several biologically-inspired abstractions of hormonal mechanisms that act as (hormonal) modulators on an agent’s physiology and behaviours. These hormonal mechanisms are inspired by two biological hormones: cortisol (CT) and oxytocin (OT).…”

“…Finally, all scalar values in the equations above ( c = 0.1; o = 0.3, i = 0.5) are apriori values used to regulate the values of respective equations, which were tuned to our specific environment prior to previous investigations ( Khan et al, 2020 , 2019 ; Khan and Cañamero, 2021 ).…”

“…TactInt (Tactile Intensity) and Speed t are discussed in the Agent Model section. All parameter values are apriori values taken from previous investigations(Khan et al, 2020;Khan and Cañamero, 2021).…”