Jump to content

Talk:Right to explanation

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

Criticism?

[edit]

From Right to explanation#Criticism:

More fundamentally, many algorithms used in machine learning are not easily explainable. For example, the output of a deep neural network depends on many layers of computations, connected in a complex way, and no one input or computation may be a dominant factor. The field of Explainable AI seeks to provide better explanations from existing algorithms, and algorithms that are more easily explainable, but it is a young and active field.

How is this criticism? The whole notion of the right to explanation is that one cannot rely on opaque algorithms where the association between the inputs and the outputs cannot be easily explained to make decisions on people on an automated basis, so the fact that "many algorithms used in machine learning are not easily explainable" does not seem to constitute a criticism of the concept. Rather, the paragraph is simply describing it. First Comet (talk) 09:14, 19 August 2023 (UTC)[reply]