Explainable AI
Approaches that make an AI system’s results or operation understandable to the intended recipients.
What should an explanation make understandable?
Explainable AI covers approaches that make an AI system’s results or operation understandable. An explanation must fit its recipients’ question and prior knowledge and faithfully describe the actual system behaviour. A development team often needs different details from the person reviewing a service schedule. Whether the explanation is accurate and whether the prediction is correct for the business task are separate questions. Understanding the explanation also includes its limits.
In a fictional planning system, a model estimates the duration of a service visit. A displayed explanation names the type of work and a planned part replacement as influential inputs. Reviewers investigate whether these statements come from a suitable explanation method for this exact prediction and match the model version used. Fluent justification text alone does not establish that. Service planning also considers whether the estimated duration makes operational sense and whether inputs are uncertain or missing. The explanation shows where a review can focus; it does not replace that review.
Sources
Related terms
- Machine learningModels learn patterns from data and apply them to new inputs.
- LLM evaluationLLM evaluation tests model outputs against defined tasks and criteria. One good example does not establish overall quality.
- Forecasting modelA model that uses available information to estimate a future quantity for a defined period.
- AI biasSystematic distortions in AI systems that can disadvantage particular cases or groups of people.
- Model driftA decline in a model’s predictive quality as data or relationships change during use.
- Re-rankingRe-ranking reassesses search candidates already retrieved. Relevance to the query and the limits of the initial selection both matter.
Your project with ASES
Describe your project to us. We will get back to you with an initial assessment.