local explanation
In plain wordsAn explanation of why an AI model gave one particular answer for one particular case — this patient, this loan — rather than how it behaves in general.
An explanation of a single prediction, stating which features or factors drove the model's output for one input and, often, how that output would change if they changed. Most feature attributionfeature attribution, n.: Giving each input to an AI model — each pixel, word or data column — a score for how much it pushed a particular output up or down. methods (SHAPSHAP, abbr.: A popular method that explains a single prediction by fairly dividing it among the inputs, using a rule from game theory for splitting a team's winnings among its players., LIMELIME, abbr.: A method that explains one prediction of any AI model by testing slightly altered versions of the input and fitting a very simple model that copies the AI's behavior near that example., integrated gradientsintegrated gradients, n.: A way to score how much each input feature contributed to a neural network's output by gradually morphing a blank 'baseline' input into the real one and adding up the network's sensitivity along the way., saliency mapssaliency map, n.: A heat map laid over an image or text that highlights which parts most influenced an AI model's output — or, in vision science, which parts of a scene would most catch a viewer's eye.), counterfactual explanationscounterfactual explanation, n.: An explanation that tells someone the smallest change that would have produced a different outcome — for example, that the application would have been accepted if it had listed two more years of work experience. and anchor rulesAnchors, prop. n.: A method that explains a prediction with a simple if-then rule that locks in the result — for example, if an email contains both 'invoice' and 'urgent', the filter almost always flags it. are local. Local explanations answer the questions that affected individuals and auditors of particular decisions ask, and are what laws on explaining automated decisions typically require. Because a model can behave very differently in different regions of its input space, a local explanation need not generalize: what drove one applicant's denial may not matter for another's. Aggregating many local explanations, as in SHAP summary plots, is one route to a global explanationglobal explanation, n.: An explanation of how an AI model behaves overall — which factors matter most across all its decisions, and in what way — rather than why it made one particular decision..
The local explanation for this customer cited three late payments; the next customer's cited none of them.
Origin
Explaining individual classification decisions was formalized by David Baehrens and colleagues (Journal of Machine Learning Research, 2010); the local/global contrast became standard after LIME (2016) put 'local' in its name.
Usage note
'Local' refers to the neighborhood of one input in feature space, not to a part of the model.
Related
- global explanationglobal explanation, n.: An explanation of how an AI model behaves overall — which factors matter most across all its decisions, and in what way — rather than why it made one particular decision.
- feature attributionfeature attribution, n.: Giving each input to an AI model — each pixel, word or data column — a score for how much it pushed a particular output up or down.
- LIMELIME, abbr.: A method that explains one prediction of any AI model by testing slightly altered versions of the input and fitting a very simple model that copies the AI's behavior near that example.
- SHAPSHAP, abbr.: A popular method that explains a single prediction by fairly dividing it among the inputs, using a rule from game theory for splitting a team's winnings among its players.
- counterfactual explanationcounterfactual explanation, n.: An explanation that tells someone the smallest change that would have produced a different outcome — for example, that the application would have been accepted if it had listed two more years of work experience.
- right to explanationright to explanation, n.: The idea that a person affected by a decision made or shaped by an algorithm — a refused loan, a rejected job application — is entitled to a meaningful account of why.
- global explanationglobal explanation, n.: An explanation of how an AI model behaves overall — which factors matter most across all its decisions, and in what way — rather than why it made one particular decision.
Sources
- Baehrens, D., Schroeter, T., Harmeling, S., Kawanabe, M., Hansen, K. & Müller, K.-R. (2010). How to Explain Individual Classification Decisions. Journal of Machine Learning Research 11, 1803–1831.
Mentioned in
explainable AIexplainable AI, n.: The field and set of methods that let people understand why an AI system produced a particular output — for example, which factors led it to reject a loan application., counterfactual explanationcounterfactual explanation, n.: An explanation that tells someone the smallest change that would have produced a different outcome — for example, that the application would have been accepted if it had listed two more years of work experience., feature attributionfeature attribution, n.: Giving each input to an AI model — each pixel, word or data column — a score for how much it pushed a particular output up or down., global explanationglobal explanation, n.: An explanation of how an AI model behaves overall — which factors matter most across all its decisions, and in what way — rather than why it made one particular decision., LIMELIME, abbr.: A method that explains one prediction of any AI model by testing slightly altered versions of the input and fitting a very simple model that copies the AI's behavior near that example., post hoc explanationpost hoc explanation, n.: An explanation produced after a model has been built, by a separate tool that examines the finished model, rather than one that comes from a model designed to be understandable., SHAPSHAP, abbr.: A popular method that explains a single prediction by fairly dividing it among the inputs, using a rule from game theory for splitting a team's winnings among its players., algorithmic recoursealgorithmic recourse, n.: Giving a person who received an unfavorable automated decision a realistic set of actions they could take to get a better result next time, such as paying down a particular debt., AnchorsAnchors, prop. n.: A method that explains a prediction with a simple if-then rule that locks in the result — for example, if an email contains both 'invoice' and 'urgent', the filter almost always flags it., individual conditional expectationindividual conditional expectation, n.: A chart that traces, for each example separately, what a model would predict as one input is swept across its range, exposing differences between examples that an overall average would hide.