Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-1607
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-10-16T18:14:29.783470Z
status: published
lens: trace
sector: health
headline: Application of deep learning and explainable AI-supported medical decision-making for facial phenotyping in genetic syndromes
dek: Objective To conduct a pilot study to assess whether saliency-based explainable artificial intelligence (XAI) affects the recognition of genetic conditions from facial images. Methods Forty-four medical geneticists, divided into AI-only and XAI-supported groups, assessed 18 images of individuals with or without genetic conditions. Diagnostic accuracy and confidence were recorded before and after viewing an AI classifier's prediction probability, with or without XAI explanations. Mediation analyses were conducted…
gain_title: When the AI classifier was correct, showing its prediction probability raised geneticists' diagnostic accuracy and confidence in recognizing genetic conditions from facial images.
problem_title: When the AI classifier was incorrect, geneticists' accuracy and confidence fell, and saliency-based explanations did not improve decisions, so the explanations were not integrated.
trace_subject: AI-assisted facial phenotyping of genetic syndromes by medical geneticists
gain_reading: When the AI classifier was correct, showing its prediction probability raised geneticists' diagnostic accuracy and confidence in recognizing genetic conditions from facial images.
gain_evidence: AI-only and XAI support improved accuracy for correct AI classifications
problem_reading: When the AI classifier was incorrect, geneticists' accuracy and confidence fell, and saliency-based explanations did not improve decisions, so the explanations were not integrated.
problem_evidence: incorrect AI classifications decreased accuracy | participants did not integrate saliency-based XAI into decisions
quick_read: Researchers ran a pilot experiment in which 44 medical geneticists, split into an AI-only group and an XAI-supported group, judged 18 facial images of people with or without genetic conditions. Accuracy and confidence were recorded before and after the geneticists saw a classifier's prediction probability, with or without saliency-based explanations.

The findings show that AI advice can help or harm depending on whether the classifier is right, which matters for clinical decision-making in genetics. The study is a pilot with a small sample, and the authors say the data may only help inform larger studies on how experts use AI and explanations.
limitation: The work is described as a pilot study with 44 geneticists and 18 images, and the authors state the data may only help inform larger studies.
tag: Dual reading
key_points: Forty-four medical geneticists assessed 18 facial images in AI-only and XAI-supported groups. | Accuracy rose with correct AI classifications and fell with incorrect ones; confidence moved in the same directions. | Geneticists found the AI prediction probability useful but viewed XAI explanations less favorably. | For incorrect AI classifications, accuracy improvement correlated negatively with perceived AI usefulness. | The authors concluded that prediction probability influenced decisions more than saliency-based explanations.
rundown: The design recorded each participant's diagnostic accuracy and confidence both before and after exposure to the AI output. This allowed the researchers to measure how much the AI prediction probability, with or without saliency-based explanations, shifted expert judgments.

Mediation analyses were used to interpret how geneticists interact with AI and XAI. When the AI was correct and no XAI was shown, the model's prediction probability mediated the link between user confidence and the decision to choose the same answer as the AI.

Participants reported that the prediction probability was useful, while the saliency-based explanations were viewed less favorably. The lack of accuracy or confidence gains from the explanations led the authors to conclude that participants did not integrate them into decisions.
sources:
- peer_reviewed | Clinical Dysmorphology | https://doi.org/10.1097/mcd.0000000000000581 | 2026-10-08
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

fd9089560e2160bc14f4ac7943b1fadc7f73f84b27b6fe370073efadb0a29811

previous

0000000000000000000000000000000000000000000000000000000000000000