Brainstorming new datasets

[kevinrobinson]

In response to the awesome post https://dalelane.co.uk/blog/?p=3723, here's some thoughts on data sets. It'd be super powerful to connect these to applications that either 1) have come up in public debate or media or 2) are things that impact young people directly. Especially for questions related to bias and fairness, I think there's a lot of potential in situating the problems in scenarios with real stakes.

Here's two specific brainstorming ideas.

There's great work by folks at AI now on existing AI systems with complicated ethics and fairness questions, here's some that relate to young people:
via AI Now. It would be amazing to see what young people had equal power in the design (or non-design) of these systems themselves.

Making a text classifier for recommending YouTube videos as being about different topics, based on the text descriptions of previous videos. Then kids can deploy that into their Scratch app, and use it to make live prediction on searches through the YouTube API and see how well their models generalize to new data (eg, they make their own YouTube search that classifies each search result based on how much it fits categories they make like "about cats" or "awesome" or "inappropriate").

I'll hopefully have more time in August or the fall to work on this, and will be excited to check out what your'e working on and brainstorm more! Thanks again for sharing all your awesome work! 👍

[kevinrobinson]

In this spirit, two other AI projects that might impact students directly are: https://www.collegeai.com/ and https://www.emotuit.com/ Both seem like applications where AI education could help young people become aware of other ways AI impacts their lives, and help them gain the power to express their own perspectives and critiques.

[kevinrobinson]

https://twitter.com/cfiesler/status/1163475515165966336 has some other great places to look, and if it's okay with you @dalelane, I could share this issue on the twitters and see if other folks there have helpful ideas here or for how to be more concretely helpful in #228. I think there are lots of other folks that might have helpful suggestions about this too!

I also understand that ethical things are complicated! :) and that you might want to frame any kinds of invitations for help or collaboration in a productive way that fits your vision for the project, so let me know what you think would be best.

[dalelane]

@kevinrobinson Thanks for the pointer - that looks like a useful thread. And any other contributions you could help get would be fantastic too, thank you.

I think it's best to keep #228 separate, but other than that I welcome pretty much any ideas!