Emotion. Escalation. Violence.Development of a Video-Based Procedure for an Early Detection of Emotion Processes in Crowd Events

Whenever people come together in large numbers in public, be it at football matches, rock concerts or demonstrations, group emotions (positive as well as discordant) are often generated. Sometimes, conflicting group emotions can escalate and may in some cases result in major acts of violence by individuals or groups. Whether and when these group emotions will arise has not been possible to predict, recognize at an early stage, or explain in socioscientific terms to date.

Only by observing the emotions and dynamics of the emotions in large groups from the very beginning are experienced observers able, on the strength of the (mostly implicit) knowledge they have gained in practice, to recognize escalation processes early on and understand, explain and to some extent control them.

The theoretical question at the centre of this project is whether emotional processes (that is, emotional escalation processes) can also be detected automatically by an observing camera and depicted in a corresponding imaging technique, with the aim of enabling escalating emotions at crowd events to be recognized as early as possible.

To answer this question, the proposed research project sets out to initially conduct the following basic research: (1) capture, analysis and conceptualization using hermeneutic sociology of knowledge of the implicit knowledge among experienced observers (police officers) of escalating group emotions by means of participant observation and interviews (field work accompanying the Federal police action force (BFHu)). The process of emotional excitement among groups will then be differentiated, validated, described systematically and conceptualized using video analysis (2) and further data material (video archive and training material). In interdisciplinary collaboration (socioscientific emotion research, sociology of technology, informatics) (3) video-based algorithms for recognition of group emotions will be developed and mapped onto emotional processes by means of machine learning. Results should be presented in real-time by a video-based visualization (e.g. colour representation of emotions in crowds of people or relevant regions of interest) that can be used for the basic research into emotions.