On the evolution of social ties as an instrumental tool for resource competition in resource patch networks

The study asks whether diversification of human skills and knowledge favors the emergence of social ties as instruments for acquiring resources. Framed as a resource competition problem, skills/knowledge are modeled as a network of resource patches where similarity is reflected by links: dense networks represent low diversification; sparse, ring-like networks represent high diversification. Two behavioral strategies compete: solitary actors move locally between linked patches to exploit resources; social actors sometimes leverage social ties to access resources near others. The motivation draws on human cultural evolution, where mastering all skills individually is costly, making dependence on others plausible. The research question is how the structure of the skills/knowledge network (diversification) conditions the evolutionary success of social versus solitary strategies.

The paper situates its question within several literatures: (1) Social foraging and producer–scrounger theory, which studies conditions under which individuals benefit from group foraging, resource sharing, scrounging versus producing, and patch exploitation (e.g., Clark and Mangel 1986; Vickery et al. 1991; Giraldeau and Caraco 2000). It highlights that prior models often omit explicit relationships among resource patches. (2) Social learning theory, distinguishing individual and social learning, with predictions about environmental stability and competition influencing adoption of social learning (Boyd and Richerson 1985; Hoppitt and Laland 2013). (3) The social brain hypothesis, proposing ecological conditions that favor complex social behavior and larger social networks alongside larger brains (Dunbar 1992, 1998, 2011). (4) Sociological theories: social exchange theory, social resource theory, and the weak tie hypothesis (Homans 1958; Emerson 1976; Lin 2001; Granovetter 1973), which posit advantages of ties for accessing dissimilar, spatially distant resources. The paper’s novelty is modeling skills/knowledge as a patch network and limiting social information use by social network range, to examine how diversification (network sparsity) affects the competitive dynamics of social versus solitary strategies.

Agent-based simulation with two actor types competing for resources on a patch network representing skills/knowledge similarity. Key components and parameters: - Patch network: M patches, each initially containing Q resources. Network structure controlled by link density D, constructed by starting from a ring (each patch linked to two nearest neighbors) and incrementally adding equally many nearest unconnected links per node until complete graph at D=1. Lower D implies localized, diversified knowledge structure; higher D implies homogeneous skills/knowledge. - Actors and behavior: N actors. Two strategies: (i) Solitary: moves from current patch to a random neighbor (neighborhood includes current patch and its linked neighbors), taking one resource if available. (ii) Social: behaves like solitary except with probability p at a time step, initiates a social action: selects another actor j with probability proportional to inverse shortest-path distance in the current social network Y (if unreachable, uses N as effective distance); adds a social tie if absent; actor j randomly selects a patch in j’s neighborhood and “peeks”; if a resource exists, actor i takes one unit and shares it with j, with fraction S to i and 1−S to j; the patch’s resource count decreases by one. - Social network: Y among actors starts empty each generation and evolves as ties are created by social actions; ties persist within the generation. - Time and generations: Each simulation experiment runs multiple generations; each generation lasts T time steps. At each time step, one actor is chosen uniformly at random to act. - Initialization of strategies: At the start of the first generation, each actor is assigned social type with probability q_initial=0.1; otherwise solitary. - Reproduction/update rule: At generation end, compute total resources collected by solitary actors r_solitary and by social actors r_social. Set q = r_social / (r_solitary + r_social) as the probability an actor adopts the social type in the next generation. All actors are replaced by N new actors, each assigned social with probability q. The patch network structure (D) remains fixed across generations; patch resources reset each generation; the social network resets each generation. - Termination/Outcome: Repeat generations until fixation of one strategy: q=1 (social-type fixation) or q=0 (solitary-type fixation). - Experimental design: For each (p, T, D) combination, run E replicate experiments and count s, the number of runs ending in social fixation. Sensitivity analyses vary M, N, Q, S, and q_initial. - Null model and statistical testing: Construct a null model identical except social actors behave like solitary (set p=0), producing s_Null over E replicates. s_Null follows Binomial(E, q_initial). Repeat to form a null distribution and compute p-values: α = Pr(s_Null ≥ s) tests whether social outcompetes solitary; β = Pr(s_Null ≤ s) tests whether solitary outcompetes social. Use 0.05 significance threshold. Implementation: Codes written in Delphi/Pascal (provided as supplementary).

- Network structure (D): Social strategy is favored primarily in sparsely connected, localized patch networks (low D), representing high diversification of skills/knowledge. In densely connected networks (high D), social ties are detrimental; solitary strategy frequently outcompetes social. - Reliance on social action (p): Purely social strategies (high p approaching 1) are inferior across network structures. Mixed strategies (moderate p) can outperform solitary in sparse networks, especially with longer T. - Generation time (T): Longer generations favor social strategy; short T amplifies disadvantages of social behavior (setup costs and opportunity costs) and favors solitary. - System size: More patches (larger M) favors social strategy (greater opportunity to leverage ties before exhausting local resources). More actors (larger N) favors solitary (increased crowding and competition reduces the effectiveness of social acquisition and depletes local patches). - Resource abundance (Q): Social strategy is advantageous when resources per patch are scarce (low Q). When Q is high, solitary actors benefit and outcompete social. - Sharing rule (S): Social behavior requires a high share for the initiator; as S decreases from 1, social becomes more disadvantageous; high S (close to 1) is needed for social strategy to be viable. - Initial conditions (q_initial): Results are robust across different initial social proportions; qualitative outcomes persist. - Emergent social network: As D decreases (more localized resource network), the average number of social ties per actor increases to a plateau; isolated individuals (no social groups) at high D give way to rapid emergence of a single large connected social group at lower D. - Overall: Diversification of skills/knowledge (sparse resource networks), long time horizons, scarce resources, fewer competitors, larger opportunity spaces, and favorable sharing make social ties an effective instrument; otherwise, solitary search dominates.

The findings address the central question by showing that diversification in skills/knowledge, modeled as sparse, localized resource patch networks, creates ecological conditions under which social ties confer access advantages that solitary movement cannot match. Social action is beneficial when it bridges otherwise hard-to-reach areas of the patch network, especially over longer timeframes and when local resources are limited. Conversely, in dense networks where movement grants broad access, the time and sharing costs of social actions reduce competitiveness of social strategies. These results align with and extend theories in social foraging (producer–scrounger), predicting that heterogeneities and access constraints favor scrounging-like behaviors; with social learning models, indicating that stable environments and weak competition favor social information use; and with the social brain hypothesis, suggesting that diversification in knowledge domains may underpin the emergence of larger social circles and cohesive groups. Sociological parallels include social exchange theory and the weak tie hypothesis, where dissimilar resources and bridging ties facilitate access to valuable, distant opportunities. The model illustrates how structural features of resource/knowledge landscapes shape the evolution of sociality, and under what constraints social ties become instrumental tools.

The paper contributes a simple yet insightful agent-based framework linking the structure of a skills/knowledge landscape to the evolution of social ties as instruments for resource acquisition. It demonstrates that social ties are favored under high diversification (sparse, localized patch networks), long time horizons, scarce resources, few competitors, and favorable sharing, while dense networks, short horizons, abundant resources, and crowded conditions favor solitary strategies. Purely social strategies are consistently inferior to mixed or solitary approaches. The model also predicts emergence of cohesive social networks as resource networks localize. Future research directions include: incorporating multi-resource sharing among helpers within patches; generating social networks with realistic features (e.g., triadic closure); empirically constructing and parameterizing skills/knowledge networks over time; and allowing behavioral mutation and dynamic strategy updating within lifetimes. Applications are suggested for academic collaboration, technology/patent landscapes, and business competition contexts.

- Resource allocation assumption: When a social actor utilizes ties, helpers do not acquire the same resource unless via defined sharing; real contexts may allow multiple credits per patch or more complex co-authorship/credit rules. - Social network structure: Within-generation social networks are random and lack realistic features such as triadic closure or degree heterogeneity. - Patch network simplicity: Skills/knowledge are modeled via ring-to-complete constructions; real knowledge networks may be modular, hierarchical, or scale-free. - Parameterization and empirical mapping: Constructing and validating real skills/knowledge networks and calibrating parameters remain challenging. - Strategy rigidity: Actors are fixed as social or solitary within a lifetime; no mutation or within-life behavioral adaptation is modeled. - Environmental reset: Patches reset each generation, approximating stable environments within generations, potentially simplifying real-world turnover dynamics.