The technicalities of XAI

Explainable Artificial Intelligence (XAI) Concepts

Folkert Stijnman

ML Engineer

Explainability

• Persons from different backgrounds will interpret results differently

Explainability

• Persons from different backgrounds will interpret results differently

Embracing XAI's limitations

 

• More sophisticated models are inherently complex: black box models
• Black box models: type of model where the decision-making process is not easily understood by humans
• Trade-off between performance and explainability

Balance between complexity and interpretability

• Not all powerful models sacrifice interpretability
• Certain techniques have slightly less accuracy but greater interpretability

Techniques in XAI

• Interpretable models by design: decision trees, linear regression
• Explainability techniques: provide an estimation of what happened in the decision-making process

Let's practice!

Explainable Artificial Intelligence (XAI) Concepts