Concern 114 · Working draft · editorial review in progress
Out-of-distribution fragility
Out-of-distribution fragility is poor performance on examples that differ from the kinds a model learned or was tested on.
Understanding the concern
A model may receive an unfamiliar combination without recognizing that its experience is limited. It can still produce a confident answer, which makes unfamiliarity hard for a user to see.
Not every new example is difficult, and not every difference causes failure. Teams should identify important unfamiliar cases, evaluate them, and provide a fallback when evidence is weak. Training-set performance alone cannot settle how a model behaves outside that setting.
An example
A plant identifier trained on garden flowers is asked to classify a damaged leaf from an unfamiliar habitat.
Illustrative unless identified as a documented case.
Words in plain English
- Model
- The learned part of an AI system. It uses patterns found during training to produce an answer or prediction.
- Training
- The process of using examples to shape how an AI model behaves. Learning patterns from material does not establish that the material is true or fair.
- Distribution
- The pattern of examples a system sees. A shift means new examples differ from the ones used in training or testing.
Evidence and its limits
Supports the distinction between performance on familiar data and generalization to changed settings.
Rabanser, Günnemann and Lipton · NeurIPS 2019 · Research study
Failing Loudly: An Empirical Study of Methods for Detecting Dataset Shift ↗ (opens in a new tab)Evaluates methods for detecting distribution changes; a detected shift is not automatically a harmful performance change.
Geirhos and colleagues · 2020 · Research review
Shortcut Learning in Deep Neural Networks ↗ (opens in a new tab)Explains how easy predictive cues can fail when used outside the training environment.
Broader frameworks and background
These resources provide context. They do not establish every claim in this concern.
A question to discuss
How does the system behave when the example is outside its tested experience?
Keep track of different views, what needs checking, and any next step in your own discussion notes.
Open discussion notes →Principles for the conversation
Before you finish · reflect together
Virtues and principles to consider
Virtues are qualities we practice in how we treat people and make choices. They are not scores for judging someone. Use these ideas to explore what a thoughtful response could look like.
Humility
Humility means recognizing the limits of our knowledge and being willing to learn from others. Expertise can be valuable without answering every question.
State where evidence is weak, invite criticism, and listen to people with different experience. Leave room to revise a decision when new information appears.
Ask together: What might we be missing, and who could help us see it?
Prudence
Prudence means using careful judgment before and during action. It balances worthwhile opportunities with the need to avoid harm that could have been anticipated.
Start with a limited trial, set stopping points, and prefer choices that can be undone when uncertainty is high. Review what happens rather than assuming the original plan remains sound.
Ask together: What is a careful next step, and when should we stop or change course?
Truthfulness
Truthfulness means making claims that match the evidence and being clear about what we do not know. It includes correcting mistakes rather than protecting a polished story.
Check sources, label guesses, and distinguish a trial result from a proven benefit. If AI helped produce a claim, someone still needs to take responsibility for checking it.
Ask together: What can we honestly say, and what still needs checking?
Editorial ratings
These are editorial judgments from the register, not measured probabilities or predictions about your situation.
About the register and scoring →More ways to explore
Search for more writing
For more specific or recent material, open a search below or ask for a web summary. AI-generated summaries can contain mistakes; open the original articles to check them.