Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Evolutionary Computation and Explainable AI: A Roadmap to Understandable Intelligent Systems

Ryan Zhou, Jaume Bacardit, Alexander Brownlee, Stefano Cagnoni, Martin Fyvie, Giovanni Iacca, John McCall, Niki van Stein, David Walker, Ting Hu

TL;DR

The goal is to demonstrate EC’s suitability for addressing current explainability challenges and to encourage further exploration of these methods, ultimately contributing to the development of more understandable and trustworthy ML models and EC algorithms.

Abstract

Artificial intelligence methods are being increasingly applied across various domains, but their often opaque nature has raised concerns about accountability and trust. In response, the field of explainable AI (XAI) has emerged to address the need for human-understandable AI systems. Evolutionary computation (EC), a family of powerful optimization and learning algorithms, offers significant potential to contribute to XAI, and vice versa. This paper provides an introduction to XAI and reviews current techniques for explaining machine learning models. We then explore how EC can be leveraged in XAI and examine existing XAI approaches that incorporate EC techniques. Furthermore, we discuss the application of XAI principles within EC itself, investigating how these principles can illuminate the behavior and outcomes of EC algorithms, their (automatic) configuration, and the underlying problem landscapes they optimize. Finally, we discuss open challenges in XAI and highlight opportunities for future research at the intersection of XAI and EC. Our goal is to demonstrate EC's suitability for addressing current explainability challenges and to encourage further exploration of these methods, ultimately contributing to the development of more understandable and trustworthy ML models and EC algorithms.

Evolutionary Computation and Explainable AI: A Roadmap to Understandable Intelligent Systems

TL;DR

The goal is to demonstrate EC’s suitability for addressing current explainability challenges and to encourage further exploration of these methods, ultimately contributing to the development of more understandable and trustworthy ML models and EC algorithms.

Abstract

Artificial intelligence methods are being increasingly applied across various domains, but their often opaque nature has raised concerns about accountability and trust. In response, the field of explainable AI (XAI) has emerged to address the need for human-understandable AI systems. Evolutionary computation (EC), a family of powerful optimization and learning algorithms, offers significant potential to contribute to XAI, and vice versa. This paper provides an introduction to XAI and reviews current techniques for explaining machine learning models. We then explore how EC can be leveraged in XAI and examine existing XAI approaches that incorporate EC techniques. Furthermore, we discuss the application of XAI principles within EC itself, investigating how these principles can illuminate the behavior and outcomes of EC algorithms, their (automatic) configuration, and the underlying problem landscapes they optimize. Finally, we discuss open challenges in XAI and highlight opportunities for future research at the intersection of XAI and EC. Our goal is to demonstrate EC's suitability for addressing current explainability challenges and to encourage further exploration of these methods, ultimately contributing to the development of more understandable and trustworthy ML models and EC algorithms.
Paper Structure (36 sections, 4 figures)

This paper contains 36 sections, 4 figures.

Table of Contents

  1. Introduction
  2. Explainable AI
  3. What is an explanation?
  4. Explainability and interpretability
  5. Why EC and XAI?
  6. EC for XAI
  7. Explainability and complexity
  8. Types of explanations
  9. Interpretability by design
  10. Explaining data and preprocessing
  11. Dimensionality reduction
  12. Feature selection and feature engineering
  13. Explaining model behavior
  14. Feature importance
  15. Global model approximations
  16. ...and 21 more sections

Figures (4)

  • Figure 1: As models and solutions become more difficult to understand, the amount of explanation required increases. Simple solutions (left point) may not require any explanation at all, and are intrinsically interpretable. Others (middle two points) may lie beyond the ability of a human to grasp easily, but can be understood with explanation. Finally, some models (right point) may remain incomprehensible even with the current best efforts at explanation.
  • Figure 2: The interaction between problem complexity and model complexity. Simple models are inherently interpretable and suitable for simple problems, but may not capture the full behavior of complex problems. When complex models are needed for complex problems, explainability becomes essential to understand the model's behavior.
  • Figure 3: Overview of the process of building an ML model, showing areas where explanations (magnifying glasses) are often applied. Also shown is the intrinsic interpretability approach (cogwheel), where models are designed to be interpretable from the start. All these methods can be used together to form a complete picture of a model's behavior.
  • Figure 4: Overview of the process of using an optimization algorithm, showing areas where explanations (magnifying glasses) can be applied.