When Consistency Becomes Bias: Interviewer Effects in Semi-Structured Clinical Interviews
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866917362120261632 |
|---|---|
| author | Watawana, Hasindri Burdisso, Sergio Moreno-Galván, Diego A. Sánchez-Vega, Fernando López-Monroy, A. Pastor Motlicek, Petr Villatoro-Tello, Esaú |
| author_facet | Watawana, Hasindri Burdisso, Sergio Moreno-Galván, Diego A. Sánchez-Vega, Fernando López-Monroy, A. Pastor Motlicek, Petr Villatoro-Tello, Esaú |
| contents | Automatic depression detection from doctor-patient conversations has gained momentum thanks to the availability of public corpora and advances in language modeling. However, interpretability remains limited: strong performance is often reported without revealing what drives predictions. We analyze three datasets: ANDROIDS, DAIC-WOZ, E-DAIC and identify a systematic bias from interviewer prompts in semi-structured interviews. Models trained on interviewer turns exploit fixed prompts and positions to distinguish depressed from control subjects, often achieving high classification scores without using participant language. Restricting models to participant utterances distributes decision evidence more broadly and reflects genuine linguistic cues. While semi-structured protocols ensure consistency, including interviewer prompts inflates performance by leveraging script artifacts. Our results highlight a cross-dataset, architecture-agnostic bias and emphasize the need for analyses that localize decision evidence by time and speaker to ensure models learn from participants' language. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_24651 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | When Consistency Becomes Bias: Interviewer Effects in Semi-Structured Clinical Interviews Watawana, Hasindri Burdisso, Sergio Moreno-Galván, Diego A. Sánchez-Vega, Fernando López-Monroy, A. Pastor Motlicek, Petr Villatoro-Tello, Esaú Computation and Language Artificial Intelligence Sound Audio and Speech Processing Automatic depression detection from doctor-patient conversations has gained momentum thanks to the availability of public corpora and advances in language modeling. However, interpretability remains limited: strong performance is often reported without revealing what drives predictions. We analyze three datasets: ANDROIDS, DAIC-WOZ, E-DAIC and identify a systematic bias from interviewer prompts in semi-structured interviews. Models trained on interviewer turns exploit fixed prompts and positions to distinguish depressed from control subjects, often achieving high classification scores without using participant language. Restricting models to participant utterances distributes decision evidence more broadly and reflects genuine linguistic cues. While semi-structured protocols ensure consistency, including interviewer prompts inflates performance by leveraging script artifacts. Our results highlight a cross-dataset, architecture-agnostic bias and emphasize the need for analyses that localize decision evidence by time and speaker to ensure models learn from participants' language. |
| title | When Consistency Becomes Bias: Interviewer Effects in Semi-Structured Clinical Interviews |
| topic | Computation and Language Artificial Intelligence Sound Audio and Speech Processing |
| url | https://arxiv.org/abs/2603.24651 |