Travis Lloyd

We offer a novel taxonomy of AI rule types and plan to make our datasets public after peer-review. Check out the preprint (w/ co-authors @Mor@hci.social, Jennah and Tung) for more details about our findings and their implications: https://arxiv.org/abs/2410.11698 /fin

The number of subreddits with AI rules has nearly doubled over the last 12 months. Larger subreddits, as well as those devoted to art and celebrity topics, are the most likely to have such rules. Rules about AI most often focus on regulating the use of AI images and raise as justification concerns about quality and authenticity. /2

Our participants perceived both good and bad-faith motivations behind AIGC use. Some felt that it could increase participation, but others saw it as a serious threat to the health of their community. Without a foolproof way to identify AIGC, our participants rely on heuristics that will likely become less effective as generative AI technology improves. /2

Check out the preprint (w/ co-authors @Mor@hci.social and @reagle@ohai.social) for details of these (and other) findings and a discussion of the implications for the future of online communities: https://arxiv.org/abs/2311.12702 /fin

@ericpsb@hci.social thank you! yes, the pre-print is currently under peer review. we'll update it when it completes the process. in the meantime, feel free to cite the arXiv version!