Search
Program Calendar
Browse By Day
Browse By Time
Browse By Person
Browse By Room
Browse By Division
Browse By Session Type
Search Tips
Personal Schedule
Sign In
Session Submission Type: Panel
The panel is aimed to encourage discussion of research procedures and behind-the-curtains difficulties that communication scholars encounter when applying machine learning bid-data-based approaches to specific social science goals. Rather than reporting results, panelists will focus on how they have achieved them, what obstacles have been overcome, and what goals could not be achieved and why. The specific goal addressed in the panel is detection of hateful and problematic content in online media, notably in user-generated messages. This goal is relatively broad to cover a variety of methodological challenges associated with it, and relatively narrow to bring together researchers who have many methodological problems in common and may benefit from collective reflection on them. Issues to discuss will include problems of text sampling and relevant text detection, concept and class content definition, quality of human coding and assessment, feature engineering for communication studies, criteria for the choice of algorithms and their parameters, and interpretation of algorithms’ output in terms of social science. All these issues will be firmly related to a concrete type of hateful content. Conclusions will be made based on the panelists’ experience from specific projects and will not be limited to common knowledge rules of machine learning research,
The list of panelists is made so as to embrace different methods – both supervised and unsupervised machine learning – and different types of hateful and problematic messages. The latter are represented by content containing incivility, political slant, polarized political value manifestations, ethnic hostility and issue problematization. The panel brings together researchers who study this content in different social and political contexts, including USA elections, Brexit, Ukrainian crisis, and ethnically divided Russian social media landscape. Age, gender, and “status” balance has been also considered when forming the panel.
The panel will contain five presentations each of which should be short enough to leave time for audience participation. The audience will be encouraged to produce comments and to contribute to the discussion rather than to ask questions. The panel will be chaired by the last presenter as the person who is most interested in timing (panel organizer).
Methodological Challenges for Detecting Incivility on Social Media - Jennifer Stromer-Galley, Syracuse U; Patricia G. C. Rossini, School of Information Studies, Syracuse U; Feifei Zhang, Syracuse U; Erin Bartolo, Syracuse U
Identifying Partisan Slant in News Articles and Twitter during the Ukraine-Russia Crisis - Dmytro Karamshuk, Skyscanner / King's College London; Tetyana Lokot, Dublin City University; Olexandr Pryymak, Facebook; Nishanth Sastry, King's College London
Detecting Social Problems From User-Generated Content: A Methodological Reflection - Oleg Nagornyy, National Research University Higher School of Economics
The Brexit Classifier: How Helpful is Text Vectorization for Mapping the Political Value Space on Twitter - Marco Bastos, City, University of London; Dan Mercea, City U London; Andrea Baronchelli, City, U of London
Methodological Challenges for Detecting Interethnic Hostility on Social Media - Olessia Koltsova, National Research U Higher School of Economics; Sergei Koltcov
Dreaming the Machine: Monitoring and Datafication of Hate Speech With Machine Learning - Salla-Maaria Laaksonen, U of Helsinki; Reeta Pöyhtäri, University of Tampere; Matti Nelimarkka, Helsinki Institute for Information Technology; Jesse Haapoja, U of Helsinki; Teemu Kinnunen, Futurice Oy