IRHAI White Paper: Medical AI Literacy & Clinical AI Governability

Institute for Responsible
Healthcare AI

Medical AI Literacy & Clinical AI Governability

A Longitudinal Educational Framework for Undergraduate Medical Education.
"The physician of the future is not merely an AI user. The physician of the future is an AI-literate clinical governor."

Author: Dr. Sharad Maheshwari MD

imagingsimplifed@gmail.com

The Core Paradigm Shift

This section outlines the foundational philosophy of the proposed curriculum. The traditional approach attempts to teach doctors how algorithms are built. This curriculum argues that doctors think in terms of patients, decisions, risk, and evidence. Therefore, education must shift from "building" to "governing."

✖

The Engineering Misconception

Most AI courses are built by engineers and follow an engineering mental model:

Algorithms → Models → Applications

The goal is NOT to create "Doctors who can code."

✔

The Clinical Reality

Doctors must map AI to their existing clinical mental model:

Patient → Problem → Decision → Outcome

"Doctors need to learn how to safely use, evaluate, supervise, and govern AI."

The 4-Year Progression

Rather than a single intensive course, this interactive module details a longitudinal approach—dedicating one month per year to AI literacy, integrated naturally with how medical students mature clinically. Select a year below to explore the themes, modules, and learning outcomes.

The Centerpiece: Six Functions of AI

This interactive visualization represents Year 3 of the curriculum. The most important realization for a medical student is identifying what AI actually does. Click on any segment of the chart to understand the specific AI function, the clinical question it answers, and real-world medical examples.

Select a function

Interactive Exploration

Click on any section of the Doughnut chart to the left to explore the six foundational functions of Medical AI.

The Clinical Execution Architecture

This is the single most important diagram a graduate must internalize. It visualizes the RATSE governance model and IRHAI philosophy, demonstrating that AI prediction is merely the middle step. Hover over each layer of the architecture to reveal the critical questions a doctor must ask at that stage.

1. Clinical Problem

Goal: Identify repetitive work, error-prone work, or a prediction opportunity.
"What problem are we actually trying to solve?"

↓

2. Data Capture

Goal: Understand the evidence and its limits.
"What data is needed? What is the bias? Correlation or causation?"

↓

3. AI Function (Prediction)

Goal: Regression, Classification, Clustering, Recommendation, or Generation.
"What is the system outputting? Where can it fail?"

↓

4. Human Validation

Goal: Benchmarking, evaluating hallucinations, and sanity checks.
"How is the output validated against ground truth?"

↓

5. Clinical Judgment & Governance

Goal: Accountability, safe deployment, human oversight.
"Who remains accountable? Can we safely act on this?"

↓

6. Patient Outcome

Goal: Better, safer patient care.
"Did this safely convert prediction into improved care?"

Target Graduate Competencies

The radar chart below illustrates the shift in competency expectations. We do not assess coding or mathematics. Assessment focuses on the ability to integrate AI into medical practice safely. By graduation, every doctor must be able to answer the 7 core questions listed.

The 7 Essential Questions

• ✓ What problem is being solved?
• ✓ What data is being used?
• ✓ Which AI function is being applied?
• ✓ What are the failure modes?
• ✓ How is the output validated?
• ✓ Who is accountable?
• ✓ How does this improve patient care?

A unified educational architecture. Medical AI Literacy Curriculum © 2026