Where health-care AI ethics becomes public: documentation as ethics infrastructure in open-source repositories

Purpose This paper aims to examine how ethical concerns about open-source artificial intelligence (AI) artefacts become publicly sustained and incorporated into documentation, with implications for health-care organisations procuring and governing components they did not build. Design/methodology/approach Comparative case study of four GitHub repositories (January 2020 to January 2025): Fairlearn, AI Fairness 360 (AIF360), Clinical Bidirectional encoder representations from transformers (BERT) and BioBERT. The corpus comprises 450 keyword-retained issue threads and 2,136 within-window comments, analysed through Big Q reflexive thematic analysis, a transparent marker audit and cross-case pattern matching. Findings Fairlearn repeatedly converts data set, terminology, intended-use and metric-interpretation debates into documentation. AIF360 raises comparable data set-provenance and metric-interpretation concerns, but they receive brief or no replies and are rarely carried into linked guidance. The clinical repositories are not silent: model-validity and intended-use concerns appear, including sustained BioBERT discussion, but they are weakly classified and unevenly carried into durable guidance. The pattern is associated with maintenance capacity, contributor pathways, issue classification and cross-channel coordination; the design does not isolate resourcing from topic or repository genre. Practical implications Health-care AI procurement should distinguish an artefact’s ethical labels from the upstream processes that produced them. Provisional audit prompts indicate where ethical reasoning is publicly inspectable; they are not a validated procurement instrument. Originality/value Documentation as ethics infrastructure names the practices through which ethical reasoning becomes durable and publicly inspectable. The translation gap is advanced as a downstream proposition requiring procurement-side testing, rather than as an outcome directly observed in these repositories.

Authors

Publication Details

Journal
Digital Policy Regulation and Governance
Published
2026-11-13
DOI
https://doi.org/10.1108/dprg-08-2026-0623
Primary Topic
Ethics and Social Impacts of AI
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Where health-care AI ethics becomes public: documentation as ethics infrastructure in open-source repositories

George Gheverghese Joseph
2 links
Digital Policy Regulation and Governance
Ethics and Social Impacts of AI
article

Where health-care AI ethics becomes public: documentation as ethics infrastructure in open-source repositories

George Gheverghese Joseph
article en

Abstract

Purpose This paper aims to examine how ethical concerns about open-source artificial intelligence (AI) artefacts become publicly sustained and incorporated into documentation, with implications for health-care organisations procuring and governing components they did not build. Design/methodology/approach Comparative case study of four GitHub repositories (January 2020 to January 2025): Fairlearn, AI Fairness 360 (AIF360), Clinical Bidirectional encoder representations from transformers (BERT) and BioBERT. The corpus comprises 450 keyword-retained issue threads and 2,136 within-window comments, analysed through Big Q reflexive thematic analysis, a transparent marker audit and cross-case pattern matching. Findings Fairlearn repeatedly converts data set, terminology, intended-use and metric-interpretation debates into documentation. AIF360 raises comparable data set-provenance and metric-interpretation concerns, but they receive brief or no replies and are rarely carried into linked guidance. The clinical repositories are not silent: model-validity and intended-use concerns appear, including sustained BioBERT discussion, but they are weakly classified and unevenly carried into durable guidance. The pattern is associated with maintenance capacity, contributor pathways, issue classification and cross-channel coordination; the design does not isolate resourcing from topic or repository genre. Practical implications Health-care AI procurement should distinguish an artefact’s ethical labels from the upstream processes that produced them. Provisional audit prompts indicate where ethical reasoning is publicly inspectable; they are not a validated procurement instrument. Originality/value Documentation as ethics infrastructure names the practices through which ethical reasoning becomes durable and publicly inspectable. The translation gap is advanced as a downstream proposition requiring procurement-side testing, rather than as an outcome directly observed in these repositories.

Digital Policy Regulation and Governance
Ethics and Social Impacts of AI
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon