Integrating trust into artificial intelligence for medicine: using diabetes as the exemplar disease

Published in Journal of Translational Medicine on 2026-02-25.

Mandy M. Shao; Agatha F. Scheideman; David Kerr; Tien Y. Wong; Juan Espinoza; Shahid N. Shah; Mohammed E. Al-Sofiani; Ashley N. Beecy; Dieter Bruno; Elizabeth Healey; Nestoras Mathioudakis; Bin Sheng; Michael P. Snyder; Yih Chung Tham; David C. Klonoff

Read the paper at the publisher (DOI 10.1186/s12967-026-07774-2)

Abstract

Artificial Intelligence (AI) has the potential to impact healthcare across multiple domains. In diabetes, a complex chronic disease affecting 600 million people globally, AI is already being used from primary care to tertiary specialist care to reduce patient and clinician burden. However, for medical AI to be widely implemented and applied specifically to diabetes, such stakeholders as patients, clinicians, healthcare administrators, regulators, and AI developers will need to establish trust in this technology.

Building trust is a balancing act depending on individual priorities of stakeholders which may not necessarily align. Both probabilistic outputs and top-choice only outputs are used in medical AI. To achieve trust in AI for diabetes care, it will be necessary to move beyond expecting only single, deterministic outputs and to establish clear standards for medical AI provenance and performance. This article presents priorities for each of the various stakeholders if they are to develop trust in medical AI and their responsibilities for contributing to the establishment of trust in medical AI.

Put this to work

Finding application record

Test whether the published finding applies in another setting.

For: Practitioners, researchers, founders, and operating leaders.
What you keep: A finding application record you can review, revise, and send.
What counts as sound: Answers a named decision
Separates evidence from assumptions
Includes the strongest contrary case
Names missing evidence
Ends with a test, owner, and date

Read the published paper and its limits before acting on the result.

Nothing entered here is stored or sent. Review the prompt before sharing confidential, personal, patient, or privileged information.

The caseName the organization, setting, product, or piece of work without including confidential information.The decisionWhat must someone decide, and who owns that decision?Evidence in handList records, observations, measurements, rights, or prior work that can be checked.What is still unknownName missing evidence, disputed assumptions, and practical constraints.

Review the prompt

Finding application record 1.0

PURPOSE
Test whether the published finding applies in another setting.

CONTEXT
The case: Not provided
The decision: Not provided
Evidence in hand: Not provided
What is still unknown: Not provided

METHOD
1. State the finding being used.
2. Compare the original setting with yours.
3. List evidence that transfers and evidence that does not.
4. Name alternative explanations.
5. Define a replication or application test.

TASK
Use the method below to produce a Finding application record.

RULES
Separate facts, inferences, assumptions, estimates, hypotheses, and unknowns. Do not invent evidence, performance, rights, customers, economics, or regulatory standing. Quote or link the source for material facts. State uncertainty. Name what would disprove the conclusion. End with the smallest credible test, an owner, a decision date, and a stop condition.

A SOUND ANSWER
- Answers a named decision
- Separates evidence from assumptions
- Includes the strongest contrary case
- Names missing evidence
- Ends with a test, owner, and date

CAUTION
Read the published paper and its limits before acting on the result.

You can leave any field blank. The prompt will mark it as not provided.

If the record survives your review, send the question, evidence, unknowns, and requested next step.

Discuss a replication or application Intellectual Frontiers on LinkedIn