Written Dialog and Social Power: Manifestations of Different Types of Power in Dialog Behavior

International Joint Conference on Natural Language Processing, pages 216–224, Nagoya, Japan, 14-18 October 2013. Written Dialog and Social Power: Manifestations of Different Types of Power in Dialog Behavior Vinodkumar Prabhakaran Department of Computer Science Columbia University New York, NY vinod@cs.columbia.edu Owen Rambow Center for Computational Learning Systems Columbia University New York, NY rambow@ccls.columbia.edu Abstract Dialog behavior is affected by power re- lations among the discourse participants. We show that four different types of power relations (hierarchical power, situational power, influence, and power over commu- nication) affect written dialog behavior in different ways. We also present a system that can identify power relations given a written dialog. 1 Introduction The recent increase in online social interactions has triggered great interest in computationally an- alyzing such interactions to gain insights about the discourse participants (DPs). Within the field of analyzing online interactions, there is a grow- ing interest in finding how social power relations between participants are reflected in the various facets of interactions, and whether the power rela- tions can be detected using computational means (Rowe et al., 2007; Bramsen et al., 2011). More recent work has shown that an analysis of the di- alog structure (and not just the message content) helps detecting power relations (Biran et al., 2012; Danescu-Niculescu-Mizil et al., 2012). Understanding the relation between dialog and power may help in various applications. For ex- ample, if a dialog system is engineered to behave appropriately given the user’s expectation of rela- tive power (for different types of power), then the user may experience the interaction with the sys- tem as more natural. Turning to dialog analysis rather than generation, we can build a computa- tional system to analyze power relations between participants in an interaction. Such a system could have various applications. Power analysis in on- line forums and communities could be useful in determining relevance to a user searching the fo- rum. For example, a user may want to limit his search to posts authored by the DPs with higher power. Power analysis may also aid law enforce- ment agencies to detect leaders and influencers in suspicious online communities. This is especially useful since the real identities of the members of such communities are often not revealed and their hierarchies may not be available to the law en- forcement agencies. The power differential between the DPs may be based on a multitude of factors such as status, au- thority, role, knowledge and so on. Early com- putational approaches to analyzing power in in- teractions relied solely on static power structures such as corporate hierarchies as the source of the power differential (Rowe et al., 2007; Bramsen et al., 2011). More recent studies have looked into dynamic notions of power as well, such as influ- ence (Biran et al., 2012). However, not much work has been done to understand how different types of power differ in the ways they affect how people interact in dialog. In this paper, we study four different types of power — hierarchical power, …

Content truncated to 500 words