An Active Inference Model of Collective Intelligence

Abstract

Collective intelligence, an emergent phenomenon in which a composite system of multiple interacting agents performs at levels greater than the sum of its parts, has long compelled research efforts in social and behavioral sciences. To date, however, formal models of collective intelligence have lacked a plausible mathematical description of the relationship between local-scale interactions between autonomous sub-system components (individuals) and global-scale behavior of the composite system (the collective). In this paper we use the Active Inference Formulation (AIF), a framework for explaining the behavior of any non-equilibrium steady state system at any scale, to posit a minimal agent-based model that simulates the relationship between local individual-level interaction and collective intelligence. We explore the effects of providing baseline AIF agents (Model 1) with specific cognitive capabilities: Theory of Mind (Model 2), Goal Alignment (Model 3), and Theory of Mind with Goal Alignment (Model 4). These stepwise transitions in sophistication of cognitive ability are motivated by the types of advancements plausibly required for an AIF agent to persist and flourish in an environment populated by other highly autonomous AIF agents, and have also recently been shown to map naturally to canonical steps in human cognitive ability. Illustrative results show that stepwise cognitive transitions increase system performance by providing complementary mechanisms for alignment between agents’ local and global optima. Alignment emerges endogenously from the dynamics of interacting AIF agents themselves, rather than being imposed exogenously by incentives to agents’ behaviors (contra existing computational models of collective intelligence) or top-down priors for collective behavior (contra existing multiscale simulations of AIF). These results shed light on the types of generic information-theoretic patterns conducive to collective intelligence in human and other complex adaptive systems.

Applying the Free Energy Principle to Complex Adaptive Systems

“…In closing, it is worth recognising that the majority of submissions presented herein focus chiefly on human systems, despite our call for more wide-ranging applications. Nevertheless, it should be clear that the authors' proposals can be readily extended to other complex adaptive systems, including biological dynamics intrinsic to other lifeforms [34][35][36], collective, group-level behaviour [39,40], and even non-living systems [41]. Taken together, we hope that the collection of papers presented in our Special Issue motivate others to consider how the FEP might be gainfully applied to their own systems of interest, living or otherwise.…”

“…The idea that the FEP can be extended to social systems is also taken up by Kaufmann and colleagues. In An Active Inference Model of Collective Intelligence [39], the authors propose an active inference model of alignment, describing the manner in which withinscale local interactions (e.g., individual agents' behaviors) can align with cross-scale global phenomena (e.g., collective behavior) in multi-scale systems. In so doing, they offer a principled, agent-based model that has the potential to function as a workbench to simulate collective intelligence as an emergent phenomenon, across many scales.…”

Epistemic Communities under Active Inference

“…From this point of view, to select an action is to infer 'what I must be doing, given what I believe and what I sense'. Extensive work has been done in the field of active inference to study social systems and the way in which the minimisation of free energy could give rise to (eventually large-scale) behavioural coordination [3,15,[78][79][80][81][82][83]. However, much of this work is still theoretical.…”

Extracting Symbolic Models of Collective Behaviors with Graph Neural Networks and Macro-Micro Evolution

“…Neural networks [11,14,20,22,26] are another approach to model collective behaviors. In particular, Cranmer et al proposed a method to perform symbolic regression in particle systems that utilizes a GNN as a data interpolator rather than the final product [7].…”