Reading path

How to judge a clinical AI tool

A path for reading the evidence behind a medical algorithm the way you would read a drug trial: what it was tested on, whether it holds up elsewhere, and how people actually use it.

The path, step by step

1. 1 A Buyer's Checklist for Deciding Whether Healthcare AI Deserves Your Trust

   6 min read

   The questions to ask before trusting any clinical model.

2. 2 External Validation: Why a Model Must Prove Itself Outside Its Training Data

   6 min read

   Why a model must be tested outside the data it learned from.

3. 3 Calibration vs Accuracy in Plain Terms, and Why Calibration Is the One That Keeps You Safe

   6 min read

   Why a confident-sounding score can still be poorly calibrated.

4. 4 Understanding Sensitivity and Specificity, and Why a Test's Real Usefulness Depends on Who Is Being Tested

   6 min read

   The two error rates every test and model trades off.

5. 5 How to Check That a Clinical Algorithm Serves Every Patient Group

   5 min read

   How a model can perform unevenly across groups of patients.

6. 6 What the Evidence Says About Trusting an AI Suggestion Too Much

   5 min read

   The human failure mode of trusting the tool too much.

7. 7 How to Judge an FDA Cleared Radiology AI Tool

   5 min read

   Reading what an FDA clearance does and does not certify.

Each step is a full article on the Reading the Evidence blog.

Back to all reading paths