🗓️ The workshop features a keynote talk, a panel discussion on human perspectives in LLMs, poster presentations of accepted archival and non-archival work, and a PhD forum for doctoral students at any stage of their PhD to present their PhD topic and receive feedback.
(3/3)
Easter brought great news: Our paper was accepted to the #ACL2026 main conference!
💡 This year's edition of CPSS explores the special focus topic of human perspectives in LLMs. We also welcome research on computational social science, human-centered perspectives in language technologies, the societal implications of LLMs and AI, and many more topics.
(2/3)
While the responses for different cues are highly correlated, we find stronger biases for more explicit but less externally valid cues. We advise future work to focus on externally valid cues such as human written conversation histories or to base their findings on a mixture of cues (2/2)
Alignment with demographic subgroups can look good for single survey questions, yet miss the correlation structure of cultural values.
Tristan Williams, Sebastian Padó, Alan Akbik and I propose a 2-level eval framework and apply it to demographically aligned LLMs.
arxiv.org/abs/2601.15755