An Interdisciplinary Experience in the Study of School Shooters’ Written Communication to Identify Markers of Warning Behavior
The paper presents the results of an interdisciplinary study that aims to identify markers of warning behavior of school shooters’ based on analyzing their written communication using computer linguistics methods. Characteristic features of school shooters’ texts include the presence of signs of cognitive distortion – “personalization” – as well as such emotions as “anger” and “sadness”. In the texts of shooters, along with asocial and destructive intentions, intentions with a clear positive connotation also occupy a significant place. Provided in the study are the results of comparative analysis of significant markers of warning behavior of school shooters that are intrinsic to their written communication, as well as the results of processing open text data sets that are not directly related to the topic of school shooting, such as: text data from Reddit, Twitter; sets of texts from blogs devoted to criminal topics; collections of BBC news texts, articles from Medium’s website; a collection of essays of psychology students from 1993 to 1996; a collection of works of literature presented on the Gutenberg project, etc. Significant features of the texts and inherent differences are emphasized. Directions are proposed for further research as part of solving the problem of identifying markers of warning behavior, using pre-trained models of natural language analysis.
Authors
- Dmitri Tretyakov
- Aleksei Savelev (ORCID: https://orcid.org/0000-0002-7466-6142)
- Анна Юрьевна Карпова (ORCID: https://orcid.org/0000-0001-7854-1438)
Institutions
- National Research Tomsk State University (RU)
- Tomsk Polytechnic University (RU)
Publication Details
- Journal
- Sociological Journal
- Published
- 2025-09-29
- DOI
- https://doi.org/10.19181/socjour.2025.31.3.4
- Primary Topic
- Gun Ownership and Violence Research
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- Russian Science Foundation