Performance of a large language model in the informed consent process for participation in a clinical trial
Rebeca Moscatel, Komal Aryal, David Chen, Allan S Detsky, Kieran L Quinn · 2026
Read Performance of a large language model in the informed consent process for participation in a clinical trial by Rebeca Moscatel, Komal Aryal, David Chen, Allan S Detsky, Kieran L Quinn (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/PMC13612517.txt.
Description
Rebeca Moscatel leads a five-author team testing whether an AI chatbot can handle the consent conversation for a clinical trial in 2026. Prospective participants ask hard questions, and research coordinators are stretched thin. The team built a language model fenced in by the trial's own documents and checked its answers two ways: human raters, and a second model acting as moderator. Both graded accuracy, readability, and consistency.
Scores looked strong overall. Humans rated accuracy at 4.8 out of five, the moderator model at 4.7. Both found reading levels near grade seven and near-identical consistency. The catch surfaced in agreement: humans barely agreed with each other, a kappa of 0.1, while the moderator agreed with itself at 0.8. Raters marked down technically correct but toneless, context-blind answers. The authors frame the system as a design tool for trial teams. Their rater split warns that consent is a human interaction, not a quiz.
Catalog record
| CREATOR | Rebeca Moscatel, Komal Aryal, David Chen, Allan S Detsky, Kieran L Quinn |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 38464 |
| LICENSE | https://creativecommons.org/licenses/by/4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |