Metraipolis › archive › dc_edd0fab111c0

When Patients Cut In: Extending Clinical Conversational AI Safety to Interruptions — arXiv:2608.29241

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.29241
License (as archived)
CC0-1.0
Archived
2026-10-17T08:06:54.865Z
SHA-256 of the archived text
57ff561353411ae835bc82b2f62d0c48f4489fa545e35b517b7a1e32fa1d3558
Stable URL of this page
https://metraipolis.com/documento/dc_edd0fab111c0

How to cite

When Patients Cut In: Extending Clinical Conversational AI Safety to Interruptions — arXiv:2608.29241 — arXiv (en), https://arxiv.org/abs/2608.29241 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_edd0fab111c0 (Metraipolis archive dc_edd0fab111c0, sha256 57ff561353411ae835bc82b2f62d0c48f4489fa545e35b517b7a1e32fa1d3558)
BibTeX
@misc{dcedd0fab111c0,
  title = {When Patients Cut In: Extending Clinical Conversational AI Safety to Interruptions — arXiv:2608.29241},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.29241},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_edd0fab111c0; sha256 57ff561353411ae835bc82b2f62d0c48f4489fa545e35b517b7a1e32fa1d3558},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.29241. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Clinical voice agents are now deployed in routine care, where real patients do not wait their turn: they interrupt. These systems typically use a cascaded architecture (speech-to-text -> LLM -> text-to-speech), so when a patient cuts the agent off mid-utterance, clinically required content can be lost even when the model handles cooperative transcripts well. Yet clinical conversational-AI benchmarks almost universally assume patients wait for the agent to finish, missing interruption-induced loss of required content. We present a transcript-based evaluation of interruption recovery, adapting conversation-analytic overlap categories into three operational types (recognitional, competitive, transitional sub-unit) and testing four deployment-oriented, non-reasoning LLM configurations across four cells spanning history-taking (information gathering) and FAQ (information provision), scored on whether the agent preserves the clinically required content. In the gathering cells, target-question failure varied across models; in the provision cells, where arms are directly comparable, failure rose for every model. Rankings differ across cells, and competitive FAQ interruption produced 30/30 provision-coverage failures for all four models (Wilson 95% CI: 88.6-100.0%; baseline 0/30 for three, 4/30 for Llama). A brief apology marker ("sorry to interrupt") shifts recovery by tens of percentage points, inconsistently across models, and for one it reduces recovery. Interruption robustness therefore cannot be a single score: evaluation must be content-grounded, reported per cell, and matched to the deployment's interruption profile.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-01 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2608.29241 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-17T07:47:52.682Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).