Technology-Enabled, Evidence-Driven, and Patient-Centered: The Way Forward for Regulating Software as a Medical Device
Bibliographic record
Abstract
Artificial intelligence (AI) is a broad discipline that aims to understand and design systems that display properties of intelligence. Machine learning (ML) is a subset of AI that describes how algorithms and models can assist computer systems in progressively improving their performance. In health care, an increasingly common application of AI/ML is software as a medical device (SaMD), which has the intention to diagnose, treat, cure, mitigate, or prevent disease. AI/ML includes either "locked" or "continuous learning" algorithms. Locked algorithms consistently provide the same output for a particular input. Conversely, continuous learning algorithms, in their infancy in terms of SaMD, modify in real-time based on incoming real-world data, without controlled software version releases. This continuous learning has the potential to better handle local population characteristics, but with the risk of reinforcing existing structural biases. Continuous learning algorithms pose the greatest regulatory complexity, requiring seemingly continuous oversight in the form of special controls to ensure ongoing safety and effectiveness. We describe the challenges of continuous learning algorithms, then highlight the new evidence standards and frameworks under development, and discuss the need for stakeholder engagement. The paper concludes with 2 key steps that regulators need to address in order to optimize and realize the benefits of SaMD: first, international standards and guiding principles addressing the uniqueness of SaMD with a continuous learning algorithm are required and second, throughout the product life cycle and appropriate to the SaMD risk classification, there needs to be continuous communication between regulators, developers, and SaMD end users to ensure vigilance and an accurate understanding of the technology.
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 distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.005 | 0.095 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.005 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
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 teacher head, 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".