Artificial Intelligence and the Implementation Challenge
Bibliographic record
Abstract
BACKGROUND: Applications of artificial intelligence (AI) in health care have garnered much attention in recent years, but the implementation issues posed by AI have not been substantially addressed. OBJECTIVE: In this paper, we have focused on machine learning (ML) as a form of AI and have provided a framework for thinking about use cases of ML in health care. We have structured our discussion of challenges in the implementation of ML in comparison with other technologies using the framework of Nonadoption, Abandonment, and Challenges to the Scale-Up, Spread, and Sustainability of Health and Care Technologies (NASSS). METHODS: After providing an overview of AI technology, we describe use cases of ML as falling into the categories of decision support and automation. We suggest these use cases apply to clinical, operational, and epidemiological tasks and that the primary function of ML in health care in the near term will be decision support. We then outline unique implementation issues posed by ML initiatives in the categories addressed by the NASSS framework, specifically including meaningful decision support, explainability, privacy, consent, algorithmic bias, security, scalability, the role of corporations, and the changing nature of health care work. RESULTS: Ultimately, we suggest that the future of ML in health care remains positive but uncertain, as support from patients, the public, and a wide range of health care stakeholders is necessary to enable its meaningful implementation. CONCLUSIONS: If the implementation science community is to facilitate the adoption of ML in ways that stand to generate widespread benefits, the issues raised in this paper will require substantial attention in the coming years.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.111 | 0.139 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.007 | 0.043 |
| Scholarly communication | 0.019 | 0.028 |
| Open science | 0.008 | 0.015 |
| Research integrity | 0.015 | 0.018 |
| Insufficient payload (model declined to judge) | 0.009 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".