This work first model the individual-level process of learning via analogy with access to a shared repertoire of cultural information, using NK landscapes to represent high-dimensional solution spaces for complex problems and suggests that even when analogy use is costly, it can nevertheless be advantageous by allowing learners to obtain high-quality solutions more efficiently and more often than those who do not learn via analogy.
Abstract
Abstract Analogies are a fundamental part of our cognition and communication. Humans are also a social species with shared cultural repertoires learnt throughout our lifetimes. As such, we expect analogies to enable the learning of complex, novel information by communication that takes advantage of shared cultural information. Here, we demonstrate the plausibility of this proposal and clarify its scope through computational modelling. We first model the individual-level process of learning via analogy with access to a shared repertoire of cultural information, using NK landscapes to represent high-dimensional solution spaces for complex problems. We then analyse a model of population dynamics to consider the conditions under which analogical communication will evolve and be maintained when costly. Our analyses suggest that even when analogy use is costly in terms of both search time and memory constraints, it can nevertheless be advantageous by allowing learners to obtain high-quality solutions more efficiently and more often than those who do not learn via analogy. Content of image described in text.
Cooperation is ubiquitous in both natural and human societies, yet its evolutionary basis remains a major challenge. A long-standing puzzle is whether having more information leads to better decision-making and thus a higher level of cooperation. To address this question, we adopt a recently developed reinforcement learning framework in which individuals learn through trial and error to maximize cumulative rewards - a paradigm that has successfully explained diverse emergent patterns in human behaviors. Specifically, we equip a structured population with the Q-learning algorithm and systematically vary the size of the interactive neighborhood, which serves as a proxy for perceived information. Interestingly, we observe a non-monotonic relationship between cooperation prevalence and neighborhood size in both two-dimensional square lattices and Barabasi-Albert scale-free networks. This inverted U-shaped dependence reveals that an optimal amount of information exists, yielding the highest level of cooperation. Mechanistic analyses show that a moderate neighborhood size enables individuals to strike an optimal balance between information sufficiency and decision-making tractability. This balance allows them to detect reciprocal opportunities while avoiding the deterioration of decision quality due to information overload. Our findings challenge everyday intuition, suggesting that a proper amount of information - not more - is optimal for the emergence of cooperation.
Yi-Hsin Ku, Xin Ou, Jiqiang Zhang et al.· 0 citations
Strategy learning governs the evolution of collective cooperation in multi-agent systems. Although evolutionary outcomes depend strongly on the information available to agents during strategy learning, most studies treat both the source and amount of that information as fixed over time. In reality, however, individuals are continually exposed to external information that varies over time. Here we develop a general framework for evolutionary dynamics with temporal information, which uses temporal networks to characterize dynamic changes in information available for strategy learning, with time-varying connections determining each agent's information state. Across synthetic and empirical networks, we find that temporal information consistently promotes cooperation relative to their static counterparts. We further demonstrate that this advantage becomes more pronounced as temporal networks become sparser. This pattern arises because sparsification amplifies information heterogeneity among agents, creating unequal access to learning information that can facilitate the spread of cooperative strategies, in sharp contrast to static information formulations where heterogeneity often suppresses cooperation. Guided by this insight, we develop an interpretable algorithm that substantially enhances cooperation across diverse networks by generating learning networks with tunable heterogeneity. Our results identify temporal information as a realistic and broadly applicable mechanism for promoting collective cooperation.
It has been proposed (Richerson and Boyd, 2008) that human intelligence is underpinned by a ratchet-like process called Cultural Evolution in which ideas, originated by individuals, can selectively spread by social learning and replace older, less fruitful ones. Useful ideas can thus accumulate beyond the lifetime of individuals. Although both social and individual learning are thought to be achieved by the selective activity-dependent adjustment of synaptic strengths in an artificial neural net-like manner, there have been relatively few attempts to incorporate neural networks into Cultural Evolution models. This has led to controversy and uncertainty about how cultural traits are created, transformed, transmitted and selected. We have constructed a transparent model of Cultural Evolution based on simple neural networks, and here we show that a population of communicating agents can learn progressively better descriptions of its environment. Specifically we generate input vectors by linearly combining hidden"causes", which agents can learn from in a nonlinear, Hebbian, manner, thus discovering synaptic weights that"unmix"the data to reveal the hidden causes. We previously showed that if agents can communicate their current estimates of hidden causes to other agents, in a selective manner we call"Light", the interacting population can learn unmixing weights under conditions where most noninteracting individuals cannot. Here we show that with"Light"a population can learn progressively better solutions, in a ratchet-like manner. Although our model is highly simplified, this simplicity allows insight into cultural evolution mechanisms that have hitherto been obscure, and provides a stepping stone to more complex, but still relatively transparent, analyses.
Human collective intelligence depends on transmission processes: who shares what with whom, how, and when. While these processes emerge from individual cognition, they can also be directed by deliberate top-down protocols. Prior work has studied how transmission shapes collective outcomes primarily through the lens of network structure, varying who shares with whom and when. But networks are state-agnostic: they cannot condition transmission on what agents know or on the state of the collective. Here, we formalize transmission protocols as state-aware programs that route information and resources based on agent and collective states, and we use LLM-guided evolutionary search to design effective protocols in a collective discovery task. Evolved protocols increase collective performance over standard baselines from the literature by up to 37%. Ablations confirm that state-awareness drives this advantage: removing content-dependence while preserving network topology and timing eliminates performance gains. We find that evolved protocols also transfer across domain variations and agent populations. These results demonstrate that effective and generalizable transmission protocols can be discovered in silico, suggesting a path toward AI-assisted design of coordination infrastructure that enhances human collective intelligence.
Cédric Colas, J'er'emy Perez, Eleni Nisioti et al.· 0 citations
Communication typically relies on a shared code, and any change to it must be coordinated between senders and receivers to avoid a breakdown of communication. The honeybee waggle dance illustrates this problem: species with horizontal combs point directly at a food source, while species with vertical combs cannot point directly and instead reference the dance to gravity, decoded against the position of the sun. We model the rise of the first of these codes and its evolutionary transition to the second in populations of bee-like agents, with selection acting at the level of colonies. In a horizontal-comb model, we find that direct pointing evolves readily when food is moderately hard to find by random search alone, whether because sites are few and large or many and small, and fails when food is too sparse to spark dances or so abundant that it is found without signaling. Adding an exogenous benefit for vertical combs, we then find that the transition to the gravity-referenced code is driven mainly by the mutation rate and the magnitude of this benefit, with the coupling between sender and receiver mutations playing a further role at low mutation rates. Given a favorable confluence of these factors, the transition proceeds reliably and without a breakdown of communication.
