A demographic theory of similarity-biased social learning

Engineering notes

Key topics: autonomous driving. See the paper for implementation details and experimental results.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为端到端自动驾驶、BEV感知、3D目标检测、轨迹预测、路径规划、LiDAR感知等高价值论文补充中文说明。

Original abstract

We develop a demographic theory of similarity-biased social learning that formalizes our understanding of when and why individuals should preferentially copy others that look or act like them. We build an evolutionary model in which individuals can either learn on their own or copy others from a demonstrator pool that contains varying proportions of in-group and out-group members, and where group tags can be more or less informative about local knowledge. We find that where social learning becomes common, selection favors copying biases that track the direction of informational advantage-toward the group that tends to be better adapted to local conditions, including an antisimilarity bias when tags are negatively associated with local correctness (as may be the case for some immigrant communities). We also find conditions in which a similarity bias can stabilize social learning when such learners are already common, but not when they are rare, with implications for the role of group identities in cultural evolution. We discuss implications for understanding parochialism as risk aversion, majority-minority dynamics, the sociology of immigration, and the lasting impacts of colonialism.

Links and sources

• PDF from original source

Comments