EID: A computational model of emergent sociopolitical identity.

摘要

Identity is one of the most widely researched concepts in social analysis, yet remains one of the most poorly understood, especially in terms of causal modeling and prediction. This dissertation claims that the intractability of identity under current analytical approaches stems from a failure to acknowledge the emergent nature of human identity and the mathematical complexity of the systems that generate it. It further argues that only a theoretical and methodological framework that accounts for the inherent complexity of identity dynamics will sustain cumulative scholarly progress. Using a complex-systems-analytic framework proposed by John Holland, it documents and provides proof-of-concept empirical tests for an agent-based model of emergent sociopolitical identity, EID, developed for the causal modeling of emergent identity phenomena.;EID is grounded in the interaction between micro-cognitive processes and interpersonal and institutional information flows. Using a Connectionist discourse comprehension model (Walter Kintsch's C-I theory), EID agents consume real-world newspaper and television content, and talk with each other and with researchers, as they seek to learn more about their surroundings and themselves. Their activities are self-governing by means of an affect-based feedback system that helps them to form appraisals of others and to develop durable preferences for particular information sources and content. EID models social and political institutions as well as other potential constraints on individuals' activities, and is language and culture-agnostic.;In micro-level empirical tests, EID replicates and integrates findings from Connectionist research into identity-salient cognitive processes, and demonstrates an ability to serve as a common framework for evaluating a wide range of extant identity theories that have until now been empirically irreconcilable. In macro-level tests, EID agents show abilities to generate human-realistic patterns of social network organization; in particular, the "small worlds" pattern. The empirical testing is limited by the computational intensiveness of EID and the availability of suitable data for empirical comparison, but the results support both the promise of EID as a platform for further causal modeling of identity and the validity of the underlying approach for shedding new light on a wide range of social-analysis problems whose intractability may be related to the presence of emergence.