Advanced level · Module 2: Interpretability, Fairness and Documentation · Lesson 1 of 4
Explaining Individual Predictions
Explain why a model made a particular prediction for one person or case.
In this lesson you will
- Describe local explanation methods such as SHAP
- Produce and communicate an explanation for a single prediction
Key concepts
local explanationShapley valuesfeature contributionsexplanation limits
The lesson
Global importance says what matters on average; local explanations say why this applicant was scored high risk. SHAP assigns each feature a contribution that, added to a base value, gives the prediction.
Explanations help people check decisions, spot errors and give reasons to those affected. They must be written in plain language: 'The main factors were a short repayment history and a high loan-to-income ratio.'
Explanations describe the model, not the world, and can be unstable when features are correlated. Do not present them as the true cause of an outcome.
Step by step
- Compute local explanations for three applicants.
- Plot the contributions for one of them.
- Write a plain-language reason statement.
- Find a case where the explanation reveals a data error.
Practice activity
Write explanation statements for five decisions made by the credit model.
Expected result: Accurate, plain-language explanations with appropriate caveats.
Check your understanding
-
What does a local explanation describe?
Show answer
Why the model made a specific prediction for one case.
-
Why should explanations not be presented as true causes?
Show answer
They describe the model's behaviour, which reflects associations in the data.
Taught live by an instructor — online or in person. Free course lessons are for self-study and do not lead to a certificate — see registration options.