CKO-016Validation & EvaluationStrong evidence

What is robustness?

Robustness is the ability of an AI system to maintain acceptable performance when faced with unusual, complex or unexpected inputs.

In more detail

Real-world evidence synthesis rarely involves ideal datasets. AI systems encounter poorly reported studies, unusual formats, contradictory information and diverse contexts. Robustness testing assesses how well a system handles these challenges without significant performance degradation.

Why it matters

A tool that performs well only under ideal conditions may fail in practice.

Decision rule

Evaluate performance on difficult and realistic cases.

Common misconception

  • “High accuracy automatically means robustness.”

At a glance

Evidence strength
Strong

Related concepts

Stability Validation Generalisability
Key takeaway

Robust tools continue performing under challenging conditions.

More on Validation & Evaluation