Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-1598
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-10-16T18:13:27.693633Z
status: published
lens: p_space
sector: labor
headline: Generative artificial intelligence in ergonomic risk assessment: A comparative analysis of REBA scores by human and AI raters
dek: BackgroundWork-related musculoskeletal disorders are a major occupational health concern, and REBA is widely used to evaluate postural risk. The role of generative artificial intelligence (AI) in REBA scoring remains insufficiently examined.ObjectiveTo explore REBA scoring differences between a certified human ergonomist and three generative AI systems under controlled multimodal input conditions.MethodsSeventy-seven anonymized posture images from four task categories were evaluated by a human reference evaluato…
gain_title: (none)
problem_title: Three generative AI systems produced systematic, body-region-specific REBA scoring differences from a certified human ergonomist, so unsupervised use in workplace ergonomic assessment could misstate postural risk.
trace_subject: (none)
gain_reading: (none)
gain_evidence: (none)
problem_reading: Three generative AI systems produced systematic, body-region-specific REBA scoring differences from a certified human ergonomist, so unsupervised use in workplace ergonomic assessment could misstate postural risk.
problem_evidence: Generative AI systems showed systematic body-region-specific differences in REBA scoring relative to the human reference evaluator.
quick_read: Researchers compared how a certified human ergonomist and three generative AI systems, ChatGPT, Gemini and Wrtn, scored 77 anonymized posture images using the Rapid Entire Body Assessment (REBA). The images covered four task categories, and the AI systems received identical images plus standardized posture descriptors.

The scores diverged in patterned ways rather than randomly, which matters because REBA results inform judgments about musculoskeletal injury risk at work. The authors describe the findings as exploratory and call for further validation and expert oversight, so it remains unknown how the tools would perform in real workplaces or on other tasks.
limitation: The authors label the findings exploratory and say further validation is needed before AI-assisted REBA assessment is used in workplace practice.
tag: Evidence-backed problem
key_points: The study used 77 anonymized posture images drawn from four task categories. | A certified human ergonomist served as the reference rater against ChatGPT, Gemini and Wrtn. | Wrtn gave comparatively higher trunk and wrist scores and lower lower-arm scores. | Gemini produced comparatively higher Final REBA scores, while ChatGPT scored lower across several body regions. | The authors call for further validation and expert oversight before AI-assisted REBA is used in practice.
rundown: Work-related musculoskeletal disorders are a major occupational health concern, and REBA is a widely used method for evaluating postural risk. The article states that the role of generative AI in REBA scoring has been insufficiently examined, which motivated this comparison.

Scores were compared using repeated-measures ANOVA with Bonferroni-adjusted post-hoc comparisons. The analysis found significant effects of rater, of body part, and of their interaction, meaning the AI-versus-human gap depended on which body region was scored.

The differences were not uniform across systems: each tool showed its own tendencies, with some over-scoring and others under-scoring particular regions compared with the human evaluator.
sources:
- peer_reviewed | WORK: A Journal of Prevention, Assessment & Rehabilitation | https://doi.org/10.1177/10519815261493106 | 2026-10-09
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

89b1c32bb7ec00a4d28e744740f802c9c9f551a0a3c5d15e5ab78812f4485ed6

previous

0000000000000000000000000000000000000000000000000000000000000000