Explainability: why did the model decide that?
Learn what AI explanations can and cannot tell you, and how to give people reasons they can use
A taste of a lesson
If a chatbot shows its reasoning steps, can I trust that is how it got the answer?
Not fully. The reasoning text is itself generated output. Research has found cases where models reach an answer influenced by something, such as a hint in the prompt, while the written reasoning never mentions it. So the steps are a useful, often sensible account, and they help you check the logic, but they are not a reliable window into the computation. Treat them like a colleague's explanation: check each step on its own merits. Exercise: take a reasoned answer and verify one step independently. Did it hold?
Written by the teacher as an example. In your lesson the tutor answers your own questions, and like any AI it can be wrong.
What you will be able to do
- Match explanation types to the needs of different audiences
- Describe feature importance, local, counterfactual and example based explanations
- Explain why post hoc explanations and model self explanations can mislead
- Write an explanation an affected person can understand and act on
Lesson plan
- 1 Explanation for whom? Identify what different audiences need from an explanation. Start
- 2 Models that explain themselves Understand interpretable by design models and when to prefer them. Start
- 3 Explanations added afterwards Learn the main post hoc methods in plain words. Start
- 4 When explanations mislead Recognise the limits and risks of explanation tools. Start
- 5 Writing explanations people can use Draft clear, honest explanations for affected individuals. Start
Try asking
About this tutor
For managers, analysts, compliance staff and curious learners who need to explain or question AI driven decisions. You learn the difference between inherently interpretable models and explanations added afterwards, and you meet the main explanation types in plain words: feature importance, local explanations for one decision, counterfactual explanations ('you would have been approved if'), and example based explanations. You learn why explanations can mislead, how a language model's own stated reasons may not reflect what drove its output, and how to write explanations that help an affected person act. No maths beyond simple tables.
Reviews
Students can review a tutor after a paid lesson. Nobody has yet.
About the teacher
I teach AI ethics as practical judgment: privacy, fairness and accountability you can act on
9 tutors 327 lessons taught Sample
I teach the ethics of AI as something you do, not something you recite. My lessons cover privacy and personal data, bias and fairness, explainability and the effects of AI on work and high stakes decisions. I use real cases and simple methods so learners can reason through a new situation on their own. My background combines data analysis with...
See Aisha's profile and tutorsMore like this
Other tutors on the same or nearby topics.