Towards Simulating Social Media Users with LLMs: Evaluating the Operational Validity of Conditioned Comment Prediction
Nils Schwager, Simon Münker, Alistair Plum, Achim Rettinger · 2026
Read Towards Simulating Social Media Users with LLMs: Evaluating the Operational Validity of Conditioned Comment Prediction by Nils Schwager, Simon Münker, Alistair Plum, Achim Rettinger (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-22752.txt.
Description
Nils Schwager and three co-authors test whether language models can stand in for social media users. The shift from exploratory tools to silicon subjects in social science needs validation of operational validity, the match between simulated and real behaviour. Their task, Conditioned Comment Prediction, asks a model to predict how a user would comment, judged against authentic digital traces.
Tests on open-weight 8-billion-parameter models in English, German, and Luxembourgish reveal a form-versus-content split in low-resource settings. Fine-tuning aligns surface form, length and syntax, yet weakens semantic grounding. Explicit conditioning from generated biographies also turns redundant under fine-tuning, as models infer latently from behavioural histories. The guidance: favour authentic behavioural traces over descriptive personas.
Catalog record
| CREATOR | Nils Schwager, Simon Münker, Alistair Plum, Achim Rettinger |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-22752