Norms emerge through iterated learning

Abstract

Injunctive norms are universal: Every culture has rules that specify what actions are forbidden, obligatory, or permitted. Where do all of these norms come from? In this paper, we identify a mechanism of cultural transmission that can explain the emergence of injunctive norms. In particular, we develop an iterated learning account that shows how weak biases in pedagogy and inference can drive norm emergence through cultural transmission. Using transmission chain studies, we confirm the core predictions of the iterated learning account. In five studies (N = 3,688), an initial generation of participants learn about an action that is merely inadvisable and teach a new generation of participants about that action. After this learning process repeats iteratively, participants in later generations reliably judge that the action is impermissible and subject to punishment. We find this inadvisable-to-impermissible effect is elicited across a wide range of initial conditions. Overall, then, our results support the idea that iterated learning can drive the emergence of injunctive norms across a wide range of contexts.

Keywords: cultural evolution; iterated learning; moral psychology; norms.

Fig. 1.

Study 1: Proportion of Bare Directive (blue), Directive+Descriptive (pink), and Bare Descriptive (yellow) responses (y-axis) across generations (x-axis). The thick lines with large dots correspond to the mean proportion across all five studies, with shaded regions corresponding to 95% CIs. The thin lines correspond to the mean proportion from the individual scenarios.

Fig. 2.

Study 1: Average scale rating (y-axis) across generations (x-axis) for each judgment measure (panels, from left to right: Rule, Forbidden, Trouble, Punish, Allowed, Permitted, Unsafe). Thick lines with large dots correspond to the mean rating across all five studies, with shaded regions corresponding to 95% CIs. Thin lines correspond to the mean rating from the individual scenarios. The horizontal line indicates the scale midpoint.

Fig. 3.

Study 2: Average scale rating (y-axis) by response type received (x-axis; Bare Directive = “d,”Directive+Descriptive = “d+,” Bare Descriptive = “ds”) for each judgment measure. Error bars correspond to 95% CIs. The horizontal line indicates the scale midpoint.

Fig. 4.

Studies 1–5: Across generations (x-axis), average scale rating (y-axis) for each judgment (panels). Black lines and dots correspond to results from Study 4. Colored lines and dots correspond to average ratings from all other studies. The shaded region corresponds to 95% CIs. The horizontal line indicates the scale mid-point.