This paper is amazing. I don’t think I’ve ever seen such a scathing critique in an academic context as is presented here.
There is now a vast and confusing literature on some combination of interpretability and ex-
plainability. Much literature on explainability confounds it with interpretability/comprehensibility,
thus obscuring the arguments, detracting from their precision, and failing to convey the relative
importance and use-cases of the two topics in practice. Some of the literature discusses topics in
such generality that its lessons have little bearing on any specific problem. Some of it aims to
design taxonomies that miss vast topics within interpretable ML. Some of it provides definitions
that we disagree with. Some of it even provides guidance that could perpetuate bad practice. Most
of it assumes that one would explain a black box without consideration of whether there is an
interpretable model of the same accuracy.
[...]
XAI surveys have (thus far) universally failed to acknowledge the important point that inter-
pretability begets accuracy when considering the full data science process, and not the other way
around. [...]
[...]
In this survey, we do not aim to provide yet another dull taxonomy of “explainability” termi-
nology. The ideas of interpretable ML can be stated in just one sentence: [...]
As far as I can tell, this is all pretty on point. (And I know I’ve conflated explanability and interpretability before.)
I think I like this because it makes up update downward on how restricted you actually are in what you can publish, as soon as you have some reasonable amount of reputation. I used to find the idea of diving into the publishing world paralyzing because you have to adhere to the process, but nowadays that seems like much less of a big deal.
This paper is amazing. I don’t think I’ve ever seen such a scathing critique in an academic context as is presented here.
As far as I can tell, this is all pretty on point. (And I know I’ve conflated explanability and interpretability before.)
I think I like this because it makes up update downward on how restricted you actually are in what you can publish, as soon as you have some reasonable amount of reputation. I used to find the idea of diving into the publishing world paralyzing because you have to adhere to the process, but nowadays that seems like much less of a big deal.