-
Notifications
You must be signed in to change notification settings - Fork 20
Define a set of standard questions to ask of ML #10
Comments
This is great! It would be nice to brainstorm questions and discuss in class too. Would this make sense as a wiki page, additional README file in the repo, or. . .? |
Seems like tomorrow is the time to discuss this a bit? Would it make sense to fold some of these points into @lydiajessup's Data Wrangling write-up in #38. |
I think that would be great! Let's discuss tomorrow morning before class the best way to add them in. In particular, I think that some of these questions could be a great way to discuss the historical context of the way gender is classified in the Titanic dataset. |
I hesitate to close this as it's such an excellent reference and I didn't get every bit of language from the original suggestion. @joeyklee if you review the materials and feel it meets the spirit of what you suggested I'll let you click the button below. We can also leave it open as a reminder to ask these questions for the rest of the semester and in future iterations of the course. |
For the archive! Datasheets for Datasets suggests questions to ask from these categories: motivation, composition, collection process, preprocessing/cleaning/labeling, uses, distribution, and maintenance |
This isn't a set of questions exactly, but ITP student Emma Norton (@doodybrains on github) just published this article that I think does an amazing job of explaining why these kinds of questions are important in the larger context of computing: https://reallifemag.com/bad-metaphors-close-to-the-metal/ I wasn't sure exactly where to put this but wanted to be sure to share! |
It would be great/helpful to define a set of critical questions for students to define when approaching ML models and data:
e.g.
As a learning outcome, it would be great if students start to always ask or address these questions in their work, reflections, and future projects.
The text was updated successfully, but these errors were encountered: