Concern 192 · Working draft · editorial review in progress
Post-hoc explanation unreliability
Post-hoc explanation unreliability occurs when an explanation produced after a prediction gives a misleading picture of what influenced it.
Understanding the concern
Some tools highlight parts of an image or assign importance to input features. These can be useful, but an attractive chart is not proof that the tool has captured the model's actual behavior. Certain methods have failed tests that change the model or its training data.
This differs from a model offering a verbal excuse: the explanation may come from a separate mathematical tool. Reliability depends on the method's assumptions and the question it is meant to answer. Explanations need their own tests rather than borrowing credibility from the model's accuracy.
An example
A map highlights a scenic route even when the vehicle actually traveled somewhere else.
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.
Evidence and its limits
The studies demonstrate failures in particular methods and settings; they do not establish that every feature-importance explanation is useless.
Adebayo and colleagues · NeurIPS 2018 · Research study
Sanity Checks for Saliency Maps ↗ (opens in a new tab)Tests whether selected visual explanation methods depend on the model and training data; not a rejection of all interpretability methods.
Slack and colleagues · 2020 · Research study
Fooling LIME and SHAP: Adversarial Attacks on Post hoc Explanation Methods ↗ (opens in a new tab)Demonstrates manipulated explanations in tested perturbation-based methods and settings.
Broader frameworks and background
These resources provide context. They do not establish every claim in this concern.
A question to discuss
Has the explanation method been tested against changes that should alter the explanation?
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.
Trustworthiness
Trustworthiness means giving people good reasons to rely on you. It grows when promises, actions, and evidence line up. With AI, a friendly voice or a famous brand is not enough: people need to know what was checked and what could still go wrong.
In practice, keep promises small enough to test, record mistakes, and tell people when the tool's limits affect them. If a service promises human review, make sure a person really has time and authority to review it.
Ask together: What would give the people affected a sound reason to trust this use?
Accountability
Accountability means being answerable for decisions and taking responsibility for repair. It requires more than saying that the software made the choice.
Name the person or team responsible, keep useful records, and give them authority to correct problems. Make the route to a complaint clear to the people affected.
Ask together: Who will answer for the outcome and help put things right?
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.