Collective Bargaining for Data

[anoanoano]

RadicalxChange wrote a memo about how CC preference signals could relate to our work on collective bargaining for data. Here's a link: https://docs.google.com/document/d/15ONBx0kR4aaT9rty0jfkz0PqBKSQ7H9sefNKwVwJBzU/edit?usp=sharing

And here's another recent article on collective bargaining for data generally: https://arxiv.org/abs/2506.10272

I'm really curious what people think of this stuff. It seems, to me, like the main sentiment in these discussions is that some kind of creator-side collective power or lobby is what is needed to re: AI. I've been arguing along those lines for a long time. It seems to me that the preference signals framework could help make it happen if it evolves in the right way -- no?

[nickmvincent]

I am very keen to discuss this further here (I am one of the authors of the above arxiv preprint and following discussions on this topic in the RadicalxChange community and related orbits for a while).

small edit right after posting In short, I definitely think the preference signals framework could facilitate collective action, and could see a path towards preference signals as a starting point that eventually advances into more formal contractual bargaining.

Some quick thoughts (with more thorough thoughts in the paper above + other writing):

• I'm hopeful that a collective-focused approach (especially getting organizations like scholarly organizations that manage conference proceedings, other institutions that house data, etc on board) could alleviate some (but not all) of the concerns in other discussion threads here.
• To be more specific, I think getting some organizations on board as early adopters could make it easier to combine the use of mixed preference signals within a group, alongside no-scrape signals, increasing the chances that AI developers adhere, opening the door to move from "voluntary compensation based on preference signal" to "formal contract".

In terms of concrete changes to the current project, I think writing out some "user stories" for specific organizations (e.g. a scholarly organization with paper proceedings created by authors with varied preferences; an organization that runs a peer production-flavoured project, etc.), and then engaging those organizations directly to get feedback from them and their constituent members, could be a useful direction in the very short term.

I'm also curious if others, especially folks who are more pessimistic about the framework, have specific concerns about these signals being adopted by organizations first instead of "bottom up" via individuals (one thing that comes to mind: each org might have to make a choice about the default, which could lead to some orgs have opt out by default, others having opt in + some preference instead).

[ccjocelyn]

Thanks for pointing out this resource. This is exactly our intention! We’d love to see continued discussion here as well.