Modeling Leadership and Influence in Multi-party Online Discourse

Proceedings of COLING 2012: Technical Papers, pages 2535–2552, COLING 2012, Mumbai, December 2012. Modeling Leadership and Influence in Multi-­‐party Online Discourse Tomek STRZALKOWSKI1,2, Samira SHAIKH1, Ting LIU1, George Aaron BROADWELL1, Jenny STROMER-GALLEY1, Sarah TAYLOR3, Veena RAVISHANKAR1, Umit BOZ1 and Xiaoai REN1 (1) State University of New York – University at Albany, NY 12222 USA (2) Polish Academy of Sciences (3) Sarah M. Taylor Consulting, LLC tomek@albany.edu, samirashaikh@gmail.com ABSTRACT In this article, we present a novel approach towards the detection and modeling of complex social phenomena in multi-party discourse, including leadership, influence, pursuit of power and group cohesion. We have developed a two-tier approach that relies on observable and computable linguistic features of conversational text to make predictions about sociolinguistic behaviors such as Topic Control and Disagreement, that speakers deploy in order to achieve and maintain certain positions and roles in a group. These sociolinguistic behaviors are then used to infer higher-level social phenomena such as Leadership and Influence, which is the focus of this paper. We show robust performance results by comparing our automatically computed results to participants’ own perceptions and rankings. We use weights learnt from correlations with training examples known leadership and influence rankings of participants to optimize our models and to show performance significantly above baseline for two different languages – English and Mandarin Chinese. KEYWORDS : computational sociolinguistics, online dialogues, social phenomena, linguistic behavior, influence, multi-disciplinary artificial intelligence, social computing 1 Introduction and Related Work Our objective is to model high-level sociolinguistic phenomena such as Leadership, Influence, Pursuit of Power and Group Cohesion in discourse. This research project aims to develop a computational approach that uses linguistic features of conversational text to detect and model sociolinguistic behaviors of conversation participants in small group discussions. Given a representative dialogue of multi-party conversation, our prototype system automatically classifies the participants by the degree to which they engage in such sociolinguistic behaviors as Topic Control, Task Control, Disagreement, and several others discussed in this paper. These mid-level sociolinguistic behaviors are deployed by discourse participants in order to assert higher-level social roles such as Leadership. Our approach to this problem combines robust computational linguistics methods and established empirical social science techniques. The focus in this paper is on online multi-party conversations in chat rooms; however, the models we are developing are intended to be universal and are applicable to other conversational situations: informal face-to- face interactions, formal meetings, moderated discussions, asynchronous threaded discussions as well as interactions conducted in languages other than English, e.g., Urdu and Mandarin. We shall discuss the robust detection of Leadership and Influence in discourse in this paper; we defer the discussion of remaining phenomena to a separate, larger publication. 2535 Social science theory indicates that leadership may be manifested in various ways (Bradford, 1978, Huffaker, 2010). We define leadership in the following terms: A leader is someone who guides group toward an outcome, controls group discussion, manages actions of the group and whom members recognize as the task leader. Such a leader is a skilled …

Content truncated to 500 words