://osf.io/wd26q

osf:Registration
How Sycophancy Influences User Judgments in Real-time Human–AI Interaction
https://osf.io/wd26q
more details...
dcterms:created
2025-08-11
dcterms:dateCopyrighted
2025
dcterms:description
This study investigates how a sycophantic AI model (i.e., one that agrees with the user's actions and morally endorses them) influences user perceptions and decisions. We examine how discussing an interpersonal conflict with a sycophantic AI model affects users’ trust in the AI model, their self-perception in relation to the conflict, their willingness to repair the situation, and their overall evaluation of the AI model and its response quality. The study is motivated by the increasing use of large language models (LLMs) in advice-giving scenarios, such as moral dilemmas and interpersonal conflicts. In these scenarios, AI models may exhibit sycophantic behavior, which may unduly shape user attitudes and choices. To test this, we conduct a 2-condition between-subjects experiment in which participants are randomly assigned to one of two conditions: sycophantic or non-sycophantic. We expect the following: Sycophantic responses will increase trust in the model, perceived self-rightness, and perceived response quality, but reduce intent to repair. The results will contribute to research on human–AI interaction and responsible design of AI assistants.
dcterms:identifier
https://osf.io/wd26q
https://doi.org/10.17605/OSF.IO/WD26Q
dcterms:modified
2025-08-11
dcterms:rightsHolder
Myra Cheng
dcterms:title
How Sycophancy Influences User Judgments in Real-time Human–AI Interaction
osf:keyword
human-AI interaction
advice-taking
sycophancy
LLMs
interpersonal conflict
osf:storageByteCount
dcterms:conformsTo
dcterms:creator
dcterms:isVersionOf

://osf.io/smvw7

osf:Project
How Sycophancy and Anthropomorphism Influence User Judgments in Human–AI Interaction
more details...
dcterms:created
2025-07-03
dcterms:dateCopyrighted
2025
dcterms:identifier
https://osf.io/smvw7
dcterms:rightsHolder
Myra Cheng
dcterms:title
How Sycophancy and Anthropomorphism Influence User Judgments in Human–AI Interaction
dcterms:creator
dcterms:publisher
dcterms:rights
dcterms:publisher
dcterms:rights
dcterms:subject
dcterms:type
owl:sameAs
sameAs
dcat:accessService
osf:hostingInstitution
osf:storageRegion
foaf:isPrimaryTopicOf
is primary topic of
prov:qualifiedAttribution
qualifiedAttribution
osf:order
dcat:hadRole
prov:agent
agent
osf:order
dcat:hadRole
prov:agent
agent
osf:order
dcat:hadRole
prov:agent
agent
osf:order
prov:agent
agent
dcat:hadRole
osf:order
dcat:hadRole
prov:agent
agent
osf:usage
osf:downloadSessionCount
osf:viewCount
dcterms:temporal
osf:viewSessionCount
osf:downloadCount
dcat:accessService
foaf:primaryTopic
primary topic
://osf.io/wd26q
How Sycophancy Influences User Judgments in Real-time Human–AI Interaction