Pro-Social on Agentic Feedback on Bksy: A Recap
A conversation with Fay Johnson, founder of CLR:SKY, a Bsky middleware tool
When PDN discovered CLR:SKY earlier this year, we were like kids in a candy shop; this Bluesky add-on (or, technically, "overlay") combined a bagful of prosocial design principles and opportunities that get us excited. It leverages friction, by popping up when a comment is getting toxic. It then harnesses LLMs to suggest how a comment can be not only less offensive, but more effective. It makes use of Middleware, third party tools that can mediate our experience on digital spaces. And it bolsters agency both by being a voluntary add-on and by letting its users make the final call on what to comment.
It was a great pleasure, then, that CLR:SKY's founder, Fay Johnson, was able to join us for a Pro-Social to talk more about the work that went into building CLR:SKY and what her team learned from the roll-out. (If you're on Bluesky, you can try CLR:SKY out here.)
Fay was inspired to build CLR:SKY, which she developed with collaborators as part of a fellowship at Berkman-Klein's Applied Social Media Lab, as a chance to apply and expand some of the successful approaches to design she had seen while working in industry, for example, giving users kindness reminders as they enter heated discussions or letting users know when their own comment may be offensive.
CLR:SKY takes those approaches - but puts them in the hands of the users themselves, letting users in essence be their own content moderators. It also assists users on how to communicate more effectively by offering suggested rewrites from an LLM that, we learned, was trained on practices of nonviolent communication; reflecting back the perspective of the other person in the thread, finding common ground, acknowledging others' feelings, and asking questions to invite dialogue. In the conversation we were also pleased to learn that part of CLR:SKY's theory of change is that by modeling healthier communication, the LLM can help users acquire the tools of effective dialogue - in essence "boosting" their capacity to communicate effectively, another one of our favorite principles of prosocial design.
When we turned the conversation to what Fay and her team learned, perhaps unsurprisingly a big takeaway was the challenge of adoption, especially when - as Fay put it - the "frequent fliers" of offensive comments are probably least likely to voluntarily install tools like CLR:SKY. In order to get wider use, you'd need a lot more awareness or platforms would need to buy in by integrating CLR:SKY or similar tools. Still we were impressed with the 40k+ who have used CLR:SKY and see it as a sign that people see value in AI being used to foster healthier conversations.
If any are interested in seeing CLR:SKY take off - by raising awareness, porting it to other platforms - let us know; we'll put you in the CLR:SKY team who are open to working with collaborators.
Catch the full conversation with Fay below.
About the Prosocial Design Network
The Prosocial Design Network researches and promotes prosocial design: evidence-based design practices that bring out the best in human nature online. Learn more at prosocialdesign.org.