%0 Conference Proceedings %T Spurious Correlations in Cross-Topic Argument Mining %A Thorn Jakobsen, Terne Sasha %A Barrett, Maria %A Søgaard, Anders %Y Ku, Lun-Wei %Y Nastase, Vivi %Y Vulić, Ivan %S Proceedings of *SEM 2021: The Tenth Joint Conference on Lexical and Computational Semantics %D 2021 %8 August %I Association for Computational Linguistics %C Online %F thorn-jakobsen-etal-2021-spurious %X Recent work in cross-topic argument mining attempts to learn models that generalise across topics rather than merely relying on within-topic spurious correlations. We examine the effectiveness of this approach by analysing the output of single-task and multi-task models for cross-topic argument mining, through a combination of linear approximations of their decision boundaries, manual feature grouping, challenge examples, and ablations across the input vocabulary. Surprisingly, we show that cross-topic models still rely mostly on spurious correlations and only generalise within closely related topics, e.g., a model trained only on closed-class words and a few common open-class words outperforms a state-of-the-art cross-topic model on distant target topics. %R 10.18653/v1/2021.starsem-1.25 %U https://aclanthology.org/2021.starsem-1.25 %U https://doi.org/10.18653/v1/2021.starsem-1.25 %P 263-277