Welcome
My name is Luc, and I enjoy machine learning, especially when it comes to natural language. I recently graduated with a Master of Computer Science from the University of Melbourne.
If you would like to get in touch with me, I am reachable at lraszewski at protonmail.com.
Publications
Luc Raszewski & Christine De Kock, 2025.
ACL 2025 Main.
Malicious sockpuppet detection on Wikipedia is critical to preserving access to reliable information on the internet and preventing the spread of disinformation. Prior machine learning approaches rely on stylistic and meta-data features, but do not prioritise adaptability to author-specific behaviours. As a result, they struggle to effectively model the behaviour of specific sockpuppet-groups, especially when text data is limited. To address this, we propose the application of meta-learning, a machine learning technique designed to improve performance in data-scarce settings by training models across multiple tasks. Meta-learning optimises a model for rapid adaptation to the writing style of a new sockpuppet-group. Our results show that meta-learning significantly enhances the precision of predictions compared to pre-trained models, marking an advancement in combating sockpuppetry on open editing platforms. We release an updated dataset of sockpuppet investigations to foster future research in both sockpuppetry and meta-learning fields.
Naomi Baes, Luc Raszewski, Ekaterina Vylomova, Nick Haslam & Christine de Kock 2026.
ICWSM 2026 Main.
Dehumanizing language is a core feature of hostile online communities, often serving to reinforce ideology and justify harm. Despite widespread claims that i**** discourse uniquely dehumanizes women, it remains unclear whether dehumanization systematically differs across gendered targets. This study provides the first theory-informed, multidimensional analysis of gendered dehumanization in i**** discourse, operationalizing five dimensions (negative evaluation, moral disgust, animalistic framing, mind denial, and agency denial) and applying them to 10.3 million posts from i**** forums (2018–2024). We extend an existing framework by adding a mind denial dimension and refining the operationalization of animality and negative evaluation, thereby strengthening its theoretical grounding and empirical coverage. Across analyses, women-associated terms demonstrate a small but statistically significant overall increase in dehumanization. However, none of the individual dimensions show reliable women–men differences, and these patterns remain stable over time. Taken together, the results indicate that dehumanization functions as an enduring representational baseline in this community: both women- and men-associated terms are evaluated negatively and embedded in dehumanizing contexts. These findings illustrate how theory-informed computational measures can characterize the multidimensional ways ideological harm is expressed in online communities.
Naomi Baes, Jemima Kang, Nick Haslam, Chris Groot, Alsa Wu, Luc Raszewski & Yulia Otmakhova, 2026.
AACL-IJCNLP 2026 Main.
Mental health stigma has profoundly harmful impacts but its complexity makes it difficult to evaluate. Stigma may involve explicit derogation, but also subtler forms of blame, fear, paternalistic pity, social distancing, structural exclusion, and discrimination. We introduce a theory-grounded benchmark for automatic evaluation of mental health stigma in online communication, consisting of naturally occurring online news and social media text annotated with a fine-grained taxonomy of stigma across multiple mental health conditions. Our annotation framework comprises a binary stigma-detection task and a multi-level taxonomy covering (i) stigma mode, (ii) domain, and (iii) specific components of certain forms of stigma. We apply this framework to texts mentioning six mental health conditions and evaluate large language models alongside stigma-related classifiers for detecting sentiment, toxicity, and hate speech. Results show that mental health stigma is not well captured by models trained to detect these neighboring constructs, and that LLMs often overpredict stigma unless given explicit operational rules - mirroring the importance of decision rules in human annotation.
Datasets
A meta-learning compatible dataset of Wikipedia Sockpuppet Investigations.