Sweden, City of Malmö, Swedish Center for Preventing Violent Extremism (CVE) Nordic Safe Cities, Common Consultancy, Analyse & Tal (2022), A safe and secure digital city - Malmö, October 2022.
Country
Sweden
Title
Year
Publication type
Geographical coverage
Local
Area/location of interest
Type of Institution
Public authority
Institution
Thematic focus
Hate crime Racism & xenophobia Hate speech: Public incitement to hatred or violence
Groups
General population
Key findings
Methodology (Qualitative/Quantitative and exact type used, questionnaires etc)
The analysis in the project is based on certain methodological foundations, and the results are derived from several considerations. The project utilizes a language algorithm that has been trained to identify and classify hate speech using artificial intelligence (AI). The technique and methodology developed are open-source, making them available to other actors who wish to use and/or further develop it within their local context. The platforms selected for the project were Facebook and Flashback. Malmö is geographically tagged on Facebook, meaning that the approximately 700 pages and groups initially identified through various searches and subsequently analyzed can be definitively tied to Malmö geographically. However, Malmö on Flashback is not geographically tagged, which means that the approximately 400 threads containing 'Malmö' in their title, which were identified and analyzed, cannot be fully tied to Malmö geographically, but instead represent the narrative surrounding Malmö.
Sample details and representativeness
Facebook is the largest discussion platform in Malmö, with 74 percent of all Swedes using it. It hosts around 700 pages and groups discussing Malmö. Flashback is the platform where the most openly hateful discourse occurs in and about Malmö, with 32 percent of Swedes using the forum. It features over 450 threads discussing Malmö. The project’s so-called hate algorithm is built on machine learning and is based on 31,500 comments from Facebook and Flashback, which have been manually coded by a group of native Swedish speakers. A Swedish language model was used, meaning the algorithm can only identify hate expressed in Swedish. The data used in the project was collected between 1 January 2019 and 28 February 2022. The data was gathered both through Facebook’s GRAPH API and through scraping. On Facebook, the data is based on anonymised posts, comments, and reactions. On Flashback, the data consists of published posts by anonymous user identities in different threads about Malmö.