Dynamics of Task-oriented Communication: Tests of Computer Models of Affect Control Theory

本文基于一种新的理论方法，结合了地位（顺从）、权力（支配）和表达性三个维度，研究了对话互动中身份维持的动力学，重点研究了对话中的轮流和中断。两项研究将测试这个新模型。研究1使用了53个模拟陪审团的数据。在讨论开始之前，对小组参与者进行身份测量。事件历史分析将模拟以正常过渡或中断结束回合的风险。来自当前讲话者和下一位讲话者的身份信息将用于预测转场的长度和类型。研究二将使用一个实验设计来操纵人们在谈话中带来的身份。通过将人们分配到具有不同评价和效力内涵的角色，然后让他们参与实例化这些角色的活动，实验将随机分配期望在会话互动中保持的身份。与研究1平行的事件历史分析将用于检验理论预测。这个项目是向先进计算领域重要的新发展迈出的一步。它采用了情感控制理论，这是一个关于身份、行为、情感和归因相互关系的正式数学模型。该理论体现在一个名为INTERACT的计算机模拟中，该模拟可以预测人类面对各种复杂情况时的情绪和评估反应。这项研究将测试计算机模拟的预测，如果这些假设得到支持，将使INTERACT的阐述能够预测小团体的行为。最重要的是，INTERACT软件可以扩展到包含更大的语言术语集，在更快更大的机器上运行，这将允许计算机模拟情感行为，并对人类的情感行为做出更适当的反应。情绪控制理论是当代社会心理学的一股重要的智力力量，而情绪社会学是其中最活跃的新兴分支学科之一，因此这一研究将引起社会和行为科学家的广泛兴趣。社会互动中情感模拟模型的改进将对人工智能和人机交流产生重大影响。

This is a study of dynamics of identity maintenance in conversational interaction, focusing on turn-taking and interruption in conversation, based on a new theoretical approach that combines the three dimensions of status (deference), power (dominance), and expressivity. Two studies will test this new model. Study 1 uses data from 53 mock juries. Identity measurements are taken from group participants before the discussions began. Event history analysis will model the risk of ending a turn with either a normal transition or an interruption. Identity information from both the current speaker and the next speaker will be used to predict turn length and type of transition. Study 2 will use an experimental design to manipulate the identities that people bring to conversations. By assigning people to roles with varying evaluative and potency connotations, then having them engage in activities to instantiate those roles, the experiment will randomly assign the identities expected to be maintained in conversational interaction. Event history analyses parallel to Study 1 will be used to test the theoretical predictions. %%% This project is a step toward important new developments in advanced computing. It employs affect control theory, a formal, mathematical model of the interrelationship of identity, behavior, emotion and attribution. The theory is embodied in a computer simulation called INTERACT that predicts the emotional and evaluative responses of human beings confronted with various complex situations. This research will test the computer simulation's predictions, and if the hypotheses are supported will enable the elaboration of INTERACT to allow prediction of small group behavior. Most importantly, INTERACT software can then be expanded to include a much larger set of linguistic terms, running on faster and larger machines, that would permit computers to simulate emotional behavior and to respond more appropriately to the emotional behavior of human beings. %%% Affect c ontrol theory is a major intellectual force in contemporary social psychology, and the sociology of emotions is among the most active new subdisciplines, so this research will be of wide interest to social and behavioral scientists. The improvement of simulation models of emotion in social interaction will have significant implications for artificial intelligence and for human-computer communication.