https://web.archive.org/web/20240725080148/https://fossacademic.tech/2024/02/11/Move-Slowy-Preview.html [[Move Slowly and Build Bridges by Robert Gehl]] is a forthcoming book on 'Mastodon, the Fediverse, and the Struggle for Ethical Social Media'. This post gives summaries per chapter of the draft. Ch1 focuses on Xodus after Musk only. Odd, there are many examples where costs of leaving socmed platforms played a role, which may well be more informative than just n=1. Ch 2 on AP as protocol Ch 3 CoC as a social layer on networked tech (no regard here it seems for the fact that human networks exist outside of tech and span multiple tech platforms simultaneously, and themselves have social norms that guid behaviour regardless whether codified in CoC or expressed in federation choices) Ch 4 on blocking and defederation as a needed safety tool. Socially I think the default might need to be the other way around, federating is the choice, defed the default, as it is how we do it socially irl. We are not unwelcoming to newcomers in a group but we are wary. Ch 5. Who pays for the fediverse infra. Short answer is we all do/many of us do. I pay my own instance, and also contribute hours to the governance of the largest Dutch instance. Good point about people forgetting there are other bizz models for digital media than what centralised adtech kraken do. Ch 6. on eco impact of socmed, and need of awareness what running this stuff costs ecologically. Seems to then pivot to how degrowth and solarpunk people using fediverse tech to interact, which is not the same thing. (It says mitigate, but compared to what, X? ) Ch 7. Threads , or the corp reaction to a growing fediverse. Conclusion, this is where the ethics will be discussed finally.

Forthcoming w Oxford Univ Press. Not sure this is for me, reads like a snapshot with a limited time window in which it might be informative. Perhaps of interest for [[Stichting ActivityClub Bestuur Hoofdnote]].

This paper introduces the DDPG algorithm which builds on the existing DPG algorithm from classic RL theory. The main idea is to define a deterministic policy, or nearly deterministic, for situations where the environment is very sensitive to suboptimal actions, and one action setting usually dominates in each state. This showed good performance, but could not beat algorithms such as PPO until the additions of SAC were added. SAC adds an entropy penalty which essentially penalizes uncertainty in any states. Using this, the deterministic policy gradient approach performs well.