ExplainabilityTransparencyTrustworthinesscs.AI
Axe the X in XAI: A Plea for Understandable AI
Andrés Páez
March 1, 2024
arXiv.org
6 citations
1 influential
Key Takeaways
-
The term "explanation" in XAI is misleading: The paper argues that what researchers call "explainable AI" doesn't actually match how explanation works in traditional science, and that adopting philosophical models of scientific explanation (Deductive Nomological, Inductive Statistical, Causal Mechanical, or New Mechanist) doesn't solve this problem for deep neural networks.
-
"Understandable AI" is a better goal than "explainable AI": The author proposes replacing the term "explainable AI" with "understandable AI" to more accurately reflect what these systems actually do and to reduce confusion about what XAI is trying to achieve.
-
Understanding should be measured by practical success, not scientific rigor: Rather than seeking explanations that meet scientific standards, the paper suggests that we should define understanding an AI system by whether users can successfully use it and draw correct inferences from it—a pragmatic approach based on actual performance rather than theoretical purity.
-
Current XAI approaches may be solving the wrong problem: By framing AI interpretability through the lens of scientific explanation, the field may be pursuing an inappropriate standard that doesn't match the real-world need for people to work effectively with AI systems.
Abstract
In a recent paper, Erasmus et al. (2021) defend the idea that the ambiguity of the term "explanation" in explainable AI (XAI) can be solved by adopting any of four different extant accounts of explanation in the philosophy of science: the Deductive Nomological, Inductive Statistical, Causal Mechanical, and New Mechanist models. In this chapter, I show that the authors' claim that these accounts can be applied to deep neural networks as they would to any natural phenomenon is mistaken. I also provide a more general argument as to why the notion of explainability as it is currently used in the XAI literature bears little resemblance to the traditional concept of scientific explanation. It would be more fruitful to use the label "understandable AI" to avoid the confusion that surrounds the goal and purposes of XAI. In the second half of the chapter, I argue for a pragmatic conception of understanding that is better suited to play the central role attributed to explanation in XAI. Following Kuorikoski & Ylikoski (2015), the conditions of satisfaction for understanding an ML system are fleshed out in terms of an agent's success in using the system, in drawing correct inferences from it.