# When Consistency Becomes Bias: Interviewer Effects in Semi-Structured Clinical Interviews

arXiv: https://arxiv.org/abs/2603.24651 · HTML: https://mlsanity.com/paper/2603.24651/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 5 of its 23 references that are in the map.
- First posted on arXiv 25 Mar 2026.

## Builds on (most cited first, 5 of 5)

- [Robust Speech Recognition via Large-Scale Weak Supervision](https://mlsanity.com/paper/2212.04356/) (2022)
- [WhisperX: Time-Accurate Speech Transcription of Long-Form Audio](https://mlsanity.com/paper/2303.00747/) (2023)
- [Automatic Depression Detection: An Emotional Audio-Textual Corpus and a GRU/BiLSTM-based Model](https://mlsanity.com/paper/2202.08210/) (2022)
- [DAIC-WOZ: On the Validity of Using the Therapist's prompts in Automatic Depression Detection from Clinical Interviews](https://mlsanity.com/paper/2404.14463/) (2024)
- [Leveraging Audio and Text Modalities in Mental Health: A Study of LLMs Performance](https://mlsanity.com/paper/2412.10417/) (2024)

## Built on by (most cited first, 1 of 1)

- [A Multi-Probe Audit of Clinical-Interview Depression Detection Benchmarks](https://mlsanity.com/paper/2605.23977/) (2026)

## Abstract (arXiv)

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.