Pedro M. P. Curvo, Mara Dragomir, Salvador Torpes
et al.
This study evaluates and extends the findings made by Piatti et al., who introduced GovSim, a simulation framework designed to assess the cooperative decision-making capabilities of large language models (LLMs) in resource-sharing scenarios. By replicating key experiments, we validate claims regarding the performance of large models, such as GPT-4-turbo, compared to smaller models. The impact of the universalization principle is also examined, with results showing that large models can achieve sustainable cooperation, with or without the principle, while smaller models fail without it. In addition, we provide multiple extensions to explore the applicability of the framework to new settings. We evaluate additional models, such as DeepSeek-V3 and GPT-4o-mini, to test whether cooperative behavior generalizes across different architectures and model sizes. Furthermore, we introduce new settings: we create a heterogeneous multi-agent environment, study a scenario using Japanese instructions, and explore an "inverse environment" where agents must cooperate to mitigate harmful resource distributions. Our results confirm that the benchmark can be applied to new models, scenarios, and languages, offering valuable insights into the adaptability of LLMs in complex cooperative tasks. Moreover, the experiment involving heterogeneous multi-agent systems demonstrates that high-performing models can influence lower-performing ones to adopt similar behaviors. This finding has significant implications for other agent-based applications, potentially enabling more efficient use of computational resources and contributing to the development of more effective cooperative AI systems.
We study the existence and multiplicity of weak solutions for the following quasilinear elliptic system: \[ \begin{cases} -\mathrm{div}(A_1(x,u_1)\nabla u_1) + \displaystyle\frac{1}{2} D_{u_1}A_1(x,u_1)\nabla u_1 \cdot \nabla u_1 = λ_1 u_1 + g_{β,1}(u) & \text{in } Ω, \\[3mm] -\mathrm{div}(A_2(x,u_2)\nabla u_2) + \displaystyle\frac{1}{2} D_{u_2}A_2(x,u_2)\nabla u_2 \cdot \nabla u_2 = λ_2 u_2 + g_{β,2}(u) & \text{in } Ω, \\[2mm] u_1 = u_2 = 0 & \text{on } \partialΩ, \end{cases} \] where $λ_1, λ_2 < μ_1$, the first Dirichlet eigenvalue of the Laplacian, and $Ω$ is a bounded domain. The nonlinearity derives from a potential $G_β$ with subcritical growth. Due to the lack of differentiability of the associated energy functional, we employ nonsmooth critical point theory and variational methods based on the concept of weak slope. We prove the existence of least energy solutions in both the cooperative ($β> 0$) and competitive ($β< 0$) regimes.
The phenomenon of group cooperation constitutes a fundamental mechanism underlying various social and biological systems. Complex networks provide a structural framework for group interactions, where individuals can not only obtain information from their neighbors but also choose neighbors as cooperative partners. However, traditional evolutionary game theory models, where nodes are the game players, are not convenient for directly choosing cooperative partners. Here, we exchange the roles of nodes and edges and innovatively propose the "edge game" model, using edges in complex networks as virtual game players for group games. Theoretical analysis and simulation experiments show that by configuring a synergy factor (r) that satisfies the "moderate cooperation" condition, a stable cooperative structure can be achieved for any network at the evolutionary equilibrium. Specifically, when there is no constraint on the number of cooperators per node, the condition for the evolution of cooperation in the network is r > kmax, where kmax is the maximum degree of the nodes. When there is a threshold constraint, in nearest-neighbor coupled networks (with degree k), the condition for "moderate cooperation" is k < r < 2k. In heterogeneous networks, a variable synergy factor scheme is adopted, where the synergy factor for each game group (rx) is defined to be proportional to the degree of the central node (kx) in the group (rx = n-fold*kx), "moderate cooperation" can be achieved when 1 < n-fold < 2. If the value of r exceeds the range, it may lead to "excessive cooperation" with node overload. Comparing algorithm performance and time complexity, edge games demonstrate advantages over other optimization algorithms. Simple and universal, the edge game provides a new approach to addressing multi-agent cooperation problems in the era of machine intelligence.
Consensus-based collective decision-making is a common feature of political life in hunter-gatherer (forager) societies. In this paper, we ask why. Synthesizing evidence from anthropology and experimental social psychology, we argue that consensus-based decision-making is an adaptive design feature of groups for efficiently processing information and facilitating good judgments, thereby increasing collective intelligence. Consensus-based decision-making facilitates information flow in several ways: by encouraging high viewpoint diversity; by placing an emphasis on independent judgments; by encouraging leadership that is informal, temporary, and knowledge-based; by encouraging inclusive deliberative norms; and by aiding in the construction of a shared reality and meaning among group members. We further argue that consensual political behavior is part of a coordination game: deliberation preceding decisions serves as a form of pregame talk that refinesmutual expectations of effort and reward, signals cooperative intent, elicits cooperation, and refinescoordination, all of which is hypothesized to result in better group decisions (i.e. higher collective intelligence). We sketch an evolutionary scenario for the emergence of consensus-based decision-making, which likely prevailed as a modal political form during the Palaeolithic.
The positive impact of cooperative bots on cooperation within evolutionary game theory is well documented; however, existing studies have predominantly used discrete strategic frameworks, focusing on deterministic actions with a fixed probability of one. This paper extends the investigation to continuous and mixed strategic approaches. Continuous strategies employ intermediate probabilities to convey varying degrees of cooperation and focus on expected payoffs. In contrast, mixed strategies calculate immediate payoffs from actions chosen at a given moment within these probabilities. Using the prisoner's dilemma game, this study examines the effects of cooperative bots on human cooperation within hybrid populations of human players and simple bots, across both well-mixed and structured populations. Our findings reveal that cooperative bots significantly enhance cooperation in both population types across these strategic approaches under weak imitation scenarios, where players are less concerned with material gains. However, under strong imitation scenarios, while cooperative bots do not alter the defective equilibrium in well-mixed populations, they have varied impacts in structured populations across these strategic approaches. Specifically, they disrupt cooperation under discrete and continuous strategies but facilitate it under mixed strategies. These results highlight the nuanced effects of cooperative bots within different strategic frameworks and underscore the need for careful deployment, as their effectiveness is highly sensitive to how humans update their actions and their chosen strategic approach.
The complete cooperation and the complete defection are two typical strategies considered in evolutionary games in many previous works. However, in real life, strategies of individuals are full of variety rather than only two complete ones. In this work, the diversity of strategies is introduced into the weak prisoners' dilemma game, which is measured by the diversity of the cooperation tendency. A higher diversity means more cooperation tendencies are provided. The complete cooperation strategy is the full cooperation tendency and the complete defection strategy is without any cooperation tendency. Agents with other cooperation tendencies behave as partial cooperators and as partial defectors simultaneously. The numerical simulation shows that increasing the diversity of the cooperation tendency promotes the cooperation level, not only the number of cooperators but also the average tendency over the whole population, until the diversity reaches its saturated value. Furthermore, our work points out maintaining cooperation is based on the cooperation efficiency approximating to the reward of cooperators and that the cooperation efficiency oscillates and quickly decreases to zero when cooperator clusters cannot resist the invasion of defectors. When the effect of the noise for the Femi update mechanism is considered, a higher diversity of strategies not only improves the cooperation level of the whole population but also supports the survival of more rational agents.
In this work, we address the cooperation problem among large language model (LLM) based embodied agents, where agents must cooperate to achieve a common goal. Previous methods often execute actions extemporaneously and incoherently, without long-term strategic and cooperative planning, leading to redundant steps, failures, and even serious repercussions in complex tasks like search-and-rescue missions where discussion and cooperative plan are crucial. To solve this issue, we propose Cooperative Plan Optimization (CaPo) to enhance the cooperation efficiency of LLM-based embodied agents. Inspired by human cooperation schemes, CaPo improves cooperation efficiency with two phases: 1) meta-plan generation, and 2) progress-adaptive meta-plan and execution. In the first phase, all agents analyze the task, discuss, and cooperatively create a meta-plan that decomposes the task into subtasks with detailed steps, ensuring a long-term strategic and coherent plan for efficient coordination. In the second phase, agents execute tasks according to the meta-plan and dynamically adjust it based on their latest progress (e.g., discovering a target object) through multi-turn discussions. This progress-based adaptation eliminates redundant actions, improving the overall cooperation efficiency of agents. Experimental results on the ThreeDworld Multi-Agent Transport and Communicative Watch-And-Help tasks demonstrate that CaPo achieves much higher task completion rate and efficiency compared with state-of-the-arts.The code is released at https://github.com/jliu4ai/CaPo.
The emergence of large-scale cooperation during the Holocene remains a central problem in the evolutionary literature. One hypothesis points to culturally evolved beliefs in punishing, interventionist gods that facilitate the extension of cooperative behaviour toward geographically distant co-religionists. Furthermore, another hypothesis points to such mechanisms being constrained to the religious ingroup, possibly at the expense of religious outgroups. To test these hypotheses, we administered two behavioural experiments and a set of interviews to a sample of 2228 participants from 15 diverse populations. These populations included foragers, pastoralists, horticulturalists, and wage labourers, practicing Buddhism, Christianity, and Hinduism, but also forms of animism and ancestor worship. Using the Random Allocation Game (RAG) and the Dictator Game (DG) in which individuals allocated money between themselves, local and geographically distant co-religionists, and religious outgroups, we found that higher ratings of gods as monitoring and punishing predicted decreased local favouritism (RAGs) and increased resource-sharing with distant co-religionists (DGs). The effects of punishing and monitoring gods on outgroup allocations revealed between-site variability, suggesting that in the absence of intergroup hostility, moralizing gods may be implicated in cooperative behaviour toward outgroups. These results provide support for the hypothesis that beliefs in monitoring and punitive gods help expand the circle of sustainable social interaction, and open questions about the treatment of religious outgroups.
Cooperation is a vital social behavior that plays a crucial role in human prosperity, enabling conflict resolution and averting disastrous outcomes. With the increasing presence of autonomous agents (AAs), human-agent interaction becomes more frequent in modern society. We investigate the impact of cooperative and defective AAs on human cooperation within the framework of evolutionary game theory, particularly in one-shot social dilemma games. Our findings reveal that cooperative AAs have a limited impact on prisoner's dilemma, but facilitate cooperation in stag hunt games. Surprisingly, defective AAs, rather than cooperative AAs, promote complete dominance of cooperation in snowdrift games. Meanwhile, in scenarios with weak imitation strength, cooperative AAs are able to maintain or even promote cooperation in all these games. Additionally, the results obtained from structured populations also imply that the effectiveness of AAs in promoting cooperation can be maximized by carefully considering their design and application in a given context.
Francesco Nava, Francesco Margoni, Nilmini Herath
et al.
Abstract Cooperation is one of the most advantageous strategies to have evolved in small- and large-scale human societies, often considered essential to their success or survival. We investigated how cooperation and the mechanisms influencing it change across the lifespan, by assessing cooperative choices from adolescence to old age (12–79 years, N = 382) forcing participants to decide either intuitively or deliberatively through the use of randomised time constraints. As determinants of these choices, we considered participants’ level of altruism, their reciprocity expectations, their optimism, their desire to be socially accepted, and their attitude toward risk. We found that intuitive decision-making favours cooperation, but only from age 20 when a shift occurs: whereas in young adults, intuition favours cooperation, in adolescents it is reflection that favours cooperation. Participants’ decisions were shown to be rooted in their expectations about other people’s cooperative behaviour and influenced by individuals’ level of optimism about their own future, revealing that the journey to the cooperative humans we become is shaped by reciprocity expectations and individual predispositions.
پژوهش حاضر با هدف بررسی تاثیر جهتگیری کارآفرینی سبز بر دیدگاه مصرف کننده و بهبود مزیت رقابتی پایدار انجام شد. جامعه آماری مورد مطالعه تعداد 150 شرکت تعاونی تولیدی کشاورزی استان خوزستان بود. از بین این شرکتها، بر اساس روش نمونهگیری با استفاده از جدول مورگان، 111 شرکت به عنوان نمونه آماری انتخاب شدند. دادههای تحقیق با استفاده از پرسشنامه شامل 30 سوال جمعآوری شد. به منظور تشریح رابطه بین جهتگیری کارآفرینی سبز، مزیت رقابتی پایدار و دیدگاه مصرفکننده از مدل معادلات ساختاری استفاده شد. دادههای به دست آمده از پرسشنامه با استفاده از روش حداقل مربعات جزیی و نرمافزارSmart-PLS3 تحلیل شدند. پایایی پرسشنامه با استفاده از روش آلفای کرونباخ و ضریب پایایی ترکیبی اندازهگیری شد. مقدار هر دو ضریب برای سازههای تحقیق بیشتر از 9/0 بود. نتایج نشان داد جهتگیری کارآفرینی سبز در شرکتهای تعاونی تولیدی کشاورزی استان خوزستان، بر دیدگاه مصرف کننده اثر مثبت و معناداری دارد. نتایج تحقیق، اثر جهتگیری کارآفرینی سبز بر دیدگاه مصرف کننده را با ضریب مسیر 707/0 و ضریب معناداری 43/17و نیز بر مزیت رقابتی پایدار را با ضریب مسیر 823/0 و ضریب معناداری 893/13تایید کرد. همچنین نتایج حاکی از تاثیر مثبت و معنادار دیدگاه مصرفکننده بر مزیت رقابتی پایدار با تاثیرپذیری از جهتگیری کارآفرینی سبز با ضریب معناداری 086/2 بود. با توجه به یافتههای این پژوهش، شرکتهای تعاونی کشاورزی میتوانند با جهتگیری کارآفرینی و نوآوری سبز، علاوه بر بهبود عملکرد اقتصادی و اجتماعی شرکت، به کاهش پیامدهای مخرب زیست محیطی کمک کنند.
Intersection management with mixed cooperative and non-cooperative vehicles is crucial in next-generation transportation systems. For fully non-cooperative systems, a minimax scheduling framework was established, while it is inefficient in mixed systems as the benefit of cooperation is not exploited. This letter focuses on the efficient scheduling in mixed systems and proposes a two-stage policy that makes full use of the cooperation relation. Specifically, a long-horizon self-organization policy is first developed to optimize the passing order of cooperative vehicles in a distributed manner, which is proved convergent when inbound roads are sufficiently long. Then a short-horizon trajectory planning policy is proposed to improve the efficiency when an ego-vehicle faces both cooperative and non-cooperative vehicles, and its safety and efficiency are theoretically validated. Furthermore, numerical simulations verify that the proposed policies can effectively reduce the scheduling cost and improve the throughput for cooperative vehicles.
Alloparental care is central to human life history, which integrates exceptionally short interbirth intervals and large birth size with an extended period of juvenile dependency and increased longevity. Formal models, previous comparative research, and palaeoanthropological evidence suggest that humans evolved higher levels of cooperative childcare in response to increasingly harsh environments. Although this hypothesis remains difficult to test directly, the relative importance of alloparental care varies across human societies, providing an opportunity to assess how local social and ecological factors influence the expression of this behaviour. We therefore, investigated associations between alloparental infant care and socioecology across 141 non-industrialized societies. We predicted increased alloparental care in harsher environments, due to the fitness benefits of cooperation in response to shared ecological challenges. We also predicted that starvation would decrease alloparental care, due to prohibitive energetic costs. Using Bayesian phylogenetic multilevel models, we tested these predictions while accounting for potential confounds as well as for population history. Consistent with our hypotheses, we found increased alloparental infant care in regions characterized by both reduced climate predictability and relatively lower average temperatures and precipitation. We also observed reduced alloparental care under conditions of high starvation. These results provide evidence of plasticity in human alloparenting in response to ecological contexts, comparable to previously observed patterns across avian and mammalian cooperative breeders. This suggests convergent social evolutionary processes may underlie both inter- and intraspecific variation in alloparental care.
Issues of labor mobilization for monumental construction are central to inquiries based in anthropological archaeology in particular as well as collective action theory in general. Societies at all scales built monumental constructions, yet the mechanisms of labor mobilization in small-scale societies remain poorly understood. Many scholars point to kin networks as sources of labor and ritual as a means of attracting laborers. These explanations fail to point to specific mechanisms that encourage cooperative labor among large groups. In this paper, I address this issue by outlining the importance of ritualized monument construction for creating and maintaining cooperative labor parties in small-scale societies. Theoretical and ethnographic evidence illustrates the importance of the four Rs of reciprocity, reputation, retribution, and reward in ritual cooperation. This in turn provides the backdrop for interpreting the scale of ritual and monumental construction during the Woodland period (500 BCE–600 CE) in the Middle Ohio River Valley to illustrate the importance of ritualized monumental construction in small-scale societies.
The understanding of cooperative behavior in social systems has been the subject of intense research over the past few decades. In this regard, the theoretical models used to explain cooperation in human societies have been complemented with a growing interest in experimental studies to validate the proposed mechanisms. In this work, we rely on previous experimental findings to build a theoretical model based on two cooperation driving mechanisms: second-order reputation and memory. Specifically, taking the donation game as a starting point, the agents are distributed among three strategies, namely, unconditional cooperators, unconditional defectors, and discriminators, where the latter follow a second-order assessment rule: shunning, stern judging, image scoring, or simple standing. A discriminator will cooperate if the evaluation of the recipient's last actions contained in his memory is above a threshold of (in)tolerance. In addition to the dynamics inherent to the game, another imitation dynamics, involving much longer times (generations), is introduced. The model is approached through a mean-field approximation that predicts the macroscopic behavior observed in Monte Carlo simulations. We found that, while in most second-order assessment rules, intolerance hinders cooperation, it has the opposite (positive) effect under the simple standing rule. Furthermore, we show that, when considering memory, the stern judging rule shows the lowest values of cooperation, while stricter rules show higher cooperation levels.