Expand transparency and explainability mitigations - Githubissues

webmachinelearning / webmachinelearning-ethics

😇 Ethical Principles for Web Machine Learning

https://www.w3.org/TR/webmachinelearning-ethics/

Other

16 stars 3 forks source link

Expand transparency and explainability mitigations #32

Closed anssiko closed 8 months ago

anssiko commented 8 months ago

Link to WebSHAP as another example https://poloclub.github.io/webshap/

anssiko commented 8 months ago

(Please note WebSHAP already links to this Ethical Principles document. Here we create a backlink.)

anssiko commented 8 months ago

@dontcallmedom CI deploy failed https://github.com/webmachinelearning/webmachinelearning-ethics/actions/runs/7445084552/job/20252666527, can you look into this? Builds clean AFAICT. Thanks!

dontcallmedom commented 8 months ago

fixed, see https://www.w3.org/TR/2024/DNOTE-webmachinelearning-ethics-20240108/

anssiko commented 8 months ago

Much thanks! Is https://www.w3.org/pubrules/ still the best tool to test for TR compliance?

Note to self: Roll up a CLI command to that takes a .bs file as an argument to test against the Specberus REST API. Unless someone has made that already.

dontcallmedom commented 8 months ago

At least in this case, the deploy failure wasn't linked to pubrules per se and had been detected by the CI at the Pull Request level https://github.com/webmachinelearning/webmachinelearning-ethics/actions/runs/7444503961/job/20251010838

That being said, it looks like spec-prod can be configured to do more pubrules checking pre-deployment if you want to make sure more errors get caught.