@inproceedings{nagy-farkas-2026-cognitive,
title = "How Cognitive Biases Propagate in {LLM} Clinical Differential Diagnosis",
author = "Nagy, Soma B{\'a}lint and
Farkas, Richard",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.2434/",
pages = "55252--55268",
ISBN = "979-8-89176-459-0",
abstract = "Clinical diagnostic reasoning is a sequential process: a working hypothesis is formed early in an encounter and updated as new evidence arrives, and biases at one stage can compound across the stages that follow. In human clinicians, anchoring and premature closure account for a large share of preventable diagnostic errors. As large language models are increasingly considered for clinical decision support, an analogous question arises: how do they handle wrong context that enters mid-encounter - whether from their own prior outputs, an earlier clinician{'}s note, or pushback from a user? Existing evaluations of LLM cognitive bias use single-shot question answering and cannot speak to this dynamic. We introduce a temporal-replay benchmark over MIMIC-IV in which 686 patient timelines are replayed step by step to three open-weight LLMs, with each step{'}s structured assessment carried forward as the prior for the next. At three clinically meaningful anchor points we inject the same wrong diagnosis through four structurally distinct channels. The most damaging is corruption of the model{'}s own previous assessment, with effects roughly twice those of any other channel. When this manipulation derails the correct answer, the model lands specifically on the planted diagnosis about half of the time."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="nagy-farkas-2026-cognitive">
<titleInfo>
<title>How Cognitive Biases Propagate in LLM Clinical Differential Diagnosis</title>
</titleInfo>
<name type="personal">
<namePart type="given">Soma</namePart>
<namePart type="given">Bálint</namePart>
<namePart type="family">Nagy</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Richard</namePart>
<namePart type="family">Farkas</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Clinical diagnostic reasoning is a sequential process: a working hypothesis is formed early in an encounter and updated as new evidence arrives, and biases at one stage can compound across the stages that follow. In human clinicians, anchoring and premature closure account for a large share of preventable diagnostic errors. As large language models are increasingly considered for clinical decision support, an analogous question arises: how do they handle wrong context that enters mid-encounter - whether from their own prior outputs, an earlier clinician’s note, or pushback from a user? Existing evaluations of LLM cognitive bias use single-shot question answering and cannot speak to this dynamic. We introduce a temporal-replay benchmark over MIMIC-IV in which 686 patient timelines are replayed step by step to three open-weight LLMs, with each step’s structured assessment carried forward as the prior for the next. At three clinically meaningful anchor points we inject the same wrong diagnosis through four structurally distinct channels. The most damaging is corruption of the model’s own previous assessment, with effects roughly twice those of any other channel. When this manipulation derails the correct answer, the model lands specifically on the planted diagnosis about half of the time.</abstract>
<identifier type="citekey">nagy-farkas-2026-cognitive</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.2434/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>55252</start>
<end>55268</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T How Cognitive Biases Propagate in LLM Clinical Differential Diagnosis
%A Nagy, Soma Bálint
%A Farkas, Richard
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F nagy-farkas-2026-cognitive
%X Clinical diagnostic reasoning is a sequential process: a working hypothesis is formed early in an encounter and updated as new evidence arrives, and biases at one stage can compound across the stages that follow. In human clinicians, anchoring and premature closure account for a large share of preventable diagnostic errors. As large language models are increasingly considered for clinical decision support, an analogous question arises: how do they handle wrong context that enters mid-encounter - whether from their own prior outputs, an earlier clinician’s note, or pushback from a user? Existing evaluations of LLM cognitive bias use single-shot question answering and cannot speak to this dynamic. We introduce a temporal-replay benchmark over MIMIC-IV in which 686 patient timelines are replayed step by step to three open-weight LLMs, with each step’s structured assessment carried forward as the prior for the next. At three clinically meaningful anchor points we inject the same wrong diagnosis through four structurally distinct channels. The most damaging is corruption of the model’s own previous assessment, with effects roughly twice those of any other channel. When this manipulation derails the correct answer, the model lands specifically on the planted diagnosis about half of the time.
%U https://aclanthology.org/2026.emnlp-main.2434/
%P 55252-55268
Markdown (Informal)
[How Cognitive Biases Propagate in LLM Clinical Differential Diagnosis](https://aclanthology.org/2026.emnlp-main.2434/) (Nagy & Farkas, EMNLP 2026)
ACL