TRV-2026-3061Version 1 · Published

Saved · see the current record

Why this version exists

Certified into the record

The exact text we saved

This is the version exactly as it was saved, character for character. The SHA-256 hash below is a 64-character code worked out from that text. Run the same text through SHA-256 and you get the same code; change a single letter and the code comes out completely different. Each version also stores the previous version’s hash, which chains them together so no version can be swapped out quietly.

TRUVACE RECORD VERSION
record: TRV-2026-3061
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-11-19T18:02:42.426200Z
status: published
lens: p_space
sector: health
headline: Physician AI disclosure, condition severity, and public trust in digital medicine
dek: Abstract Public trust is central to the implementation of artificial intelligence (AI) in digital medicine, yet little is known about how patients respond to different forms of physician AI disclosure across clinical risk contexts. In a 5 × 2 factorial experiment in mainland China, 1574 adults were randomly assigned to physician profiles varying by AI use type and condition severity. Administrative AI was rated more favorably than diagnostic and therapeutic AI across physician perception and care-seeking intenti…
gain_title: (none)
problem_title: In a randomized experiment with 1,574 adults in mainland China, physicians who disclosed diagnostic or therapeutic AI use were rated less favorably than those disclosing administrative AI, and high-severity conditions amplified negative responses to therapeutic AI.
trace_subject: (none)
gain_reading: (none)
gain_evidence: (none)
problem_reading: In a randomized experiment with 1,574 adults in mainland China, physicians who disclosed diagnostic or therapeutic AI use were rated less favorably than those disclosing administrative AI, and high-severity conditions amplified negative responses to therapeutic AI.
problem_evidence: Administrative AI was rated more favorably than diagnostic and therapeutic AI across physician perception and care-seeking intention outcomes. | High-severity conditions amplified negative responses to therapeutic AI
quick_read: A randomized experiment in mainland China assigned 1,574 adults to physician profiles that differed by the type of AI use disclosed and by how severe the patient's condition was. Doctors who disclosed administrative AI were rated more favorably than those disclosing diagnostic or therapeutic AI, on both perceptions of the physician and willingness to seek care.

Severity changed the picture. When the condition was high-severity, people reacted more negatively to therapeutic AI, while an explicit statement that no AI was used was linked to greater willingness to book an appointment. Exploratory analyses suggested competence, trust and empathy accounted for different response patterns.

The authors conclude that AI disclosure should separate administrative from clinical uses and be calibrated to severity. The experiment measured stated intentions rather than real behavior, and the authors call for preregistered replication with behavioral outcomes.
limitation: The authors say the findings need preregistered replication using behavioral outcomes, and the mediation analyses were exploratory.
tag: Evidence-backed problem
key_points: The study was a 5 × 2 factorial experiment: five types of physician AI use crossed with two levels of condition severity. | 1,574 adults in mainland China were randomly assigned to physician profiles. | Administrative AI was rated more favorably than diagnostic and therapeutic AI on physician perception and care-seeking intention. | High-severity conditions amplified negative responses to therapeutic AI. | Explicit non-AI disclosure was associated with greater appointment willingness under high severity. | Exploratory mediation analyses suggested competence, trust and empathy explained different disclosure-related response patterns. | A supplementary offline community sample in Macao (n = 128) showed directionally consistent main effects.
rundown: Participants in mainland China viewed physician profiles that varied by the kind of AI use disclosed and the severity of the condition. Outcomes included perceptions of the physician and intention to seek care from that doctor.

Administrative AI fared best across outcomes. Diagnostic and therapeutic AI were rated less favorably, and the penalty for therapeutic AI grew when the condition was severe. Under high severity, explicitly saying no AI was used was associated with greater willingness to make an appointment.

Exploratory mediation analyses pointed to competence, trust and empathy as different routes behind the response patterns. A smaller offline sample of 128 people in Macao showed main effects in the same direction.
sources:
- peer_reviewed | npj Digital Medicine | https://doi.org/10.1038/s41746-026-03484-7 | 2026-11-19
prev: 0000000000000000000000000000000000000000000000000000000000000000
SHA-256 hash
6681c2c9c4c7cd097dbf84bc2f412ed23d4ad99ce64ad7070521afc2eb953beb
Previous version’s hash
0000000000000000000000000000000000000000000000000000000000000000
Check this record hasn’t been altered
How to check it yourself, without trusting this page

Download any version from /api/record/TRV-2026-3061 and hash its canonical field yourself, for example with shasum -a 256. You should get its content_hash. Each version’s text also ends with prev:and the hash of the version before it (version 1 points to 64 zeros). If anyone had changed a single character of any version after it was saved, those hashes wouldn’t line up.