TRV-2026-3012Version 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-2026-3012
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-11-18T00:03:00.218266Z
status: published
lens: g_space
sector: health
headline: Beyond the Black Box: A 30-Minute Artificial Intelligence Critical Appraisal Workshop for Internal Medicine Residents
dek: Introduction: Artificial intelligence (AI) and large language models (LLMs) are rapidly entering resident workflows. However, formal training on appropriate LLM use is limited, and AI outputs may be inaccurate, outdated, or biased. This creates a curricular gap: residents may use AI before receiving guidance on evaluating AI-generated recommendations.
Methods: We developed a 30-minute AI critical appraisal workshop for postgraduate year 1 to 3 (PGY-1-PGY-3) internal medicine residents. The workshop replaced 1 ev…
gain_title: In an anonymous before-and-after survey at one internal medicine program, residents' confidence in critically appraising AI output rose from 60.4% before a 30-minute workshop to 79.5% after it.
problem_title: (none)
trace_subject: (none)
gain_reading: In an anonymous before-and-after survey at one internal medicine program, residents' confidence in critically appraising AI output rose from 60.4% before a 30-minute workshop to 79.5% after it.
gain_evidence: Confidence critically appraising AI output increased from 60.4% preworkshop to 79.5% postworkshop.
problem_reading: (none)
problem_evidence: (none)
quick_read: A 30-minute workshop on critically appraising AI output replaced one evidence-based medicine journal club session for postgraduate year 1 to 3 internal medicine residents. Six sessions ran from July to December 2025. Residents prompted a large language model with a fictional patient scenario, compared outputs in small groups and learned a structured appraisal framework.

Of 72 eligible residents, 53 completed a pre-workshop survey and 44 a post-workshop survey. Confidence in appraising AI output rose from 60.4% to 79.5%, and 90.5% rated the session 4 or 5 out of 5.

Concern about inaccurate AI predictions rose from 69.8% to 88.6%, and concern about bias from 77.4% to 88.6%. Belief that AI improves health outcomes stayed high at 94.3% before and 97.7% after. Surveys were anonymous and unpaired, so the results were analyzed descriptively only.
limitation: Responses were anonymous and unpaired, with 53 pre-workshop and 44 post-workshop respondents out of 72 eligible residents, so the results were analyzed descriptively and cannot show change in individual residents.
tag: Evidence-backed gain
key_points: The workshop lasted 30 minutes and replaced one evidence-based medicine journal club session during ambulatory continuity clinic blocks. | Six workshops were delivered between July and December 2025 to PGY-1 to PGY-3 internal medicine residents. | Residents prompted an LLM with a fictional patient scenario, compared outputs in small groups and learned a structured appraisal framework. | Of 72 eligible residents, 53 completed the pre-workshop survey and 44 the post-workshop survey. | Self-reported confidence in appraising AI output rose from 60.4% to 79.5%. | Concern about inaccurate AI predictions rose from 69.8% to 88.6%, and concern about AI bias from 77.4% to 88.6%. | Belief that AI improves health outcomes stayed high, at 94.3% before and 97.7% after, and 90.5% rated the session 4 or 5 out of 5.
rundown: The authors describe a curricular gap: residents may use AI tools before getting guidance on evaluating AI-generated recommendations, even though outputs can be inaccurate, outdated or biased. The workshop was designed to fill that gap within existing teaching time.

Surveys were collected anonymously through QR codes immediately before and after each session. Because responses were unpaired, the changes compare two groups of respondents rather than the same residents over time.

Residents became more worried about inaccurate predictions and bias after the session while their belief that AI improves health outcomes remained high. The authors conclude the required workshop was feasible and well received, and that it is adaptable, needs minimal technology and fits into existing evidence-based medicine teaching.
sources:
- peer_reviewed | MedEdPORTAL | https://doi.org/10.15766/mep_2374-8265.11654 | 2026-11-16
prev: 0000000000000000000000000000000000000000000000000000000000000000
SHA-256 hash
3945409cfad8161ecfede528632ed4bc084bc23660850a0122fb35732e4367cb
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-2026-3012 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.