"From Preference Learning to Language Shifts: Quantifying the Bidirectional Interactions of Humans and Machines"
Thomas Stephan Juzek
Department of Modern Languages and Linguistics
Florida State University (FSU)
Wednesday, Mar 11, 2026
- Colloquium - 499 DSL Seminar Room
- 03:30 to 04:30 PM Eastern Time (US and Canada)
Click Here to Join via Zoom
Meeting # 942 7359 5552
Zoom Meeting # 942 7359 5552
Abstract:
This talk explores how human language and language produced by large language models diverge, and how they interact. First, model-agnostic diagnostics for identifying AI language behaviour are presented, by comparing paired human texts with matched AI outputs under the same conditions. Large-scale datasets are used, spanning 30+ languages. Second, analyses tracking change over time test the extent to which these AI-associated signals propagate into human communication, including news, academic writing, and unscripted speech. Cross-lingual homogenisation pressures are discussed. Finally, results comparing base and instruction-tuned models are used to assess how preference-based training may amplify word-choice biases, with validation from behavioural tasks. The talk concludes by outlining a planned ablation study to establish causal mechanisms linking preference learning to model behaviour.
| File | Description | File size |
|---|---|---|
 Tom Juzek.jpg | Tom Juzek | 102 kB |
