Understanding how groups of individuals establish shared customs or professional guidelines has long been a subject of inquiry. Recent research indicates that this process isn't simply about mimicking others or sophisticated calculations. Rather, people navigate a two-phase journey: initially exploring various actions, then firmly adopting a specific behavior once it meets a predetermined mathematical benchmark. This critical threshold is defined by the Tolerance Principle, a concept originally conceived to elucidate how children master grammatical structures, and now demonstrated to influence the critical junctures of societal customs and cultural evolution.
This groundbreaking study, published in the Proceedings of the National Academy of Sciences (PNAS), addresses a fundamental question in the social sciences. Researchers from the CUNY Graduate Center, the University of Pennsylvania, and Stanford University found that individuals don't primarily learn by replicating observed actions or by always selecting the most rational option. Instead, they engage in an initial period of behavioral sampling, after which they commit to a pattern once sufficient supporting evidence has been accumulated. This transition is governed by the Tolerance Principle, a straightforward mathematical rule that predicts when people have observed enough regularity to accept a pattern as a rule, even in the presence of occasional deviations.
Spencer Caplan, a Linguistics professor at the CUNY Graduate Center and a co-lead author, emphasized that social learning is often mistakenly thought to involve imitation or careful optimization. However, their findings point to a more fundamental human process: exploration of options, followed by a commitment when a pattern is deemed 'good enough,' even if conflicting evidence arises later. This principle, which also explains how children acquire native language grammar—tolerating irregular verbs while still learning general rules—appears to be a universal cognitive mechanism for adopting shared behaviors, even dictating how competing norms can supersede one another.
To investigate the emergence of conventions, the researchers developed computational models representing different learning strategies. They tested these models against data from coordination experiments, including both existing studies and new experiments conducted by the team. In these experiments, participants were tasked with aligning on common choices, such as naming an unfamiliar face, within social networks. They received incentives for matching others' responses, allowing the researchers to track the evolution of their decisions over time.
The experimental results consistently showed that human behavior diverged from prevailing social learning theories. Participants did not merely imitate the most recent behavior they witnessed, nor did they invariably select the statistically optimal choice. Instead, their initial actions were probabilistic, reflecting uncertainty, until their cumulative experiences surpassed a mental threshold. This threshold is precisely described by the Tolerance Principle: a simple equation that elucidates when individuals deem a pattern consistent enough to be considered a rule, despite encountering some exceptions.
The proposed model not only more accurately reflected human learning patterns but also surpassed alternative approaches, including Bayesian models, in predicting behavior within controlled experimental settings. This suggests that a singular cognitive mechanism may underpin human learning across various domains, from linguistic development to the adoption of social norms. Just as children learn grammatical rules while accommodating exceptions, adults appear to integrate conventions in a rule-based manner once they gather adequate evidence.
This research also provides critical insights into the mechanisms of social change. By predicting when individuals commit to a convention, the model can estimate the minimum size of a dissenting minority required to challenge and potentially overturn an established norm. This offers a mathematical framework for understanding tipping points in collective behavior, with broad implications for public health initiatives, organizational dynamics, and the dissemination of new ideas within society. Future studies aim to explore these dynamics in more intricate real-world contexts, where factors like identity, status, and power also influence social conventions.