In this work we focus on the dynamics of the conflict that often arises in a police interview between suspects and police officers. Police interviews are a special type of social encounter, primarily because of the authority role of the police interviewer and the often uncooperative stance that the suspect takes: a conflict situation. The skill to resolve or reduce the conflict, to make an uncooperative suspect more cooperative, requires training of the police officer. Leary's interactional circumplex  is used in police interview training as a theoretical framework to understand how suspects take stance during an interview and how this is related to the stance that the interviewer takes. The circumplex consists of two axes, power (dominance-submission) and affiliation (opposed-together) and is divided in stances. Leary predicts the dynamics between the stances of interactants which he calls “interpersonal reflexes‿. Acts on the power dimension are complementary (dominant invokes submissive and vice versa) and acts on the affiliation dimension are symmetric (together invokes together, and opposed invokes opposed). Currently, officers practice applying this theory with expensive actors that are sparsely available. Artificial conversational characters that play the role of a suspect in a police interrogation game, a game where policemen can practice applying Leary's theory, would allow for cheaper training and fewer restriction in time and location of the training. Building artificial suspects requires explicit models of strategies and tactics that policemen apply and explicit models of the relevant internal psycho-social mechanisms that underlie the behaviors of a suspect in a police interview. Therefore, we annotated (practice) police interviews on the stance the suspect (professional actors) and police officer take towards each other. Depending on the part, up to nine independent annotators labeled the stance of the speech contributions in three police interviews (using audio and video). In the interviews, one or two officers interviewed one suspect. The result was a small corpus of 50 minutes and 1300 contributions annotated on stance. First, we investigated whether different observers (annotators) agree on the type of stance that suspects and policemen take by having all annotators annotate a small part of the corpus. Labeling stance on the level of speech segments is difficult. Even when the annotators were allowed to discuss, they were often unable to come to an unanimous agreement of the stance displayed. We found that although inter-annotator agreement on stance labeling is low (Krippendorf's a = 0:24), a majority voting “meta-annotator‿ was able to reveal the important dynamics and trends in stance taking in a police interview with relative high “inter-meta-annotator" accuracy (Cohen's = 0:55) . The results of the meta-annotator showed that police officers generally take a dominant-together stance. This is part of their taught strategy. According to Leary's theory this stance would make the suspect move to a submissive-together stance, resulting in a cooperative dialogue. Indeed, our meta-annotator showed that a suspect goes from a typical opposed stance at the start of the interview to a more cooperative stance later. This shows the correctness of Leary's theory in the special type of conversations, police interviews, where conflict is abundant and interactants are engaged in uncooperative dialogue. It also shows the applicability of the theory in modeling an artificial suspect. Annotations showed that the trend towards cooperation in suspects is not always visible and sometimes destroyed. This occurs when suspects felt disrespected or threatened by the interviewer, showing that Leary's theory alone is insufficient to model a police interview or a convincing artificial suspect. Other psycho-social theories (e.g. face threats ) should be taken into account in future models for artificial suspects and perhaps be made explicit in the police training.
|Number of pages||2|
|Publication status||Published - Oct 2013|
|Event||SSPNet Workshop on Conflict and Communication 2013 - Aula Magna Universita' Roma Tre, Roma, Italy|
Duration: 29 Oct 2013 → 31 Oct 2013
|Conference||SSPNet Workshop on Conflict and Communication 2013|
|Period||29/10/13 → 31/10/13|