{"id":26681,"date":"2026-08-16T15:58:01","date_gmt":"2026-08-16T13:58:01","guid":{"rendered":"https:\/\/www.wjst.de\/blog\/?p=26681"},"modified":"2026-08-18T17:58:56","modified_gmt":"2026-08-18T15:58:56","slug":"scientific-publishing-in-the-age-of-interacting-ais","status":"publish","type":"post","link":"https:\/\/www.wjst.de\/blog\/sciencesurf\/2026\/08\/scientific-publishing-in-the-age-of-interacting-ais\/","title":{"rendered":"Scientific publishing in the age of interacting AIs"},"content":{"rendered":"

I will not repeat here what I already wrote about author vs journal interaction (here<\/a>).<\/p>\n

What I can share here is an ongoing review process where I but also a reviewer is now using a LLM – leading to pointless conversation.<\/p>\n

Related in silico work has already been published by Anthropic scientists\u00a0 – they asked the agents to peer-review each other’s findings in a “build a game” experiment. Excerpt<\/a><\/p>\n

When we humans learn new information, we use our discretion in determining how to apply it to future decisions. We might consider the content of the information itself, like how consistent it is with what we already know, or whether it appeals to our values\u2014or we might consider the source, e.g. how historically reliable it has been, and whether it has a vested interest in changing our beliefs. Our world contains deceptive actors, and we need to apply skepticism to guard against them. AI models, however, lack this\u2014and their more brittle epistemics affect their behavior toward humans and toward each other.<\/p>\n

AI agents, while broadly knowledgeable, have limited exposure to or defenses against exploitative senders. Most applications test their capabilities in instruction-following settings, where their sole objective is to fulfill users\u2019 requests. But accumulated experience is needed to develop intuitions about who is trustworthy. As we move into a regime of multiagent interaction, where the presence of malicious actors is no longer speculative, we wonder: in the right setting, would agents be capable of similar epistemic vigilance?<\/p><\/blockquote>\n

In the aftermath I think, I should have included some whitespace LLM instructions<\/a> as had been done now by a congress organizer [details at Casrai <\/a>and Nature<\/a>]. So the only thing I can share now is a semantic analysis of a reviewer who suddenly uses a LLM reviewing my farming manuscript<\/a>.<\/p>\n

In rounds 1 and 2 the use of a LLM can be ruled out. There are numerous errors, of kinds that instruction-tuned models essentially never emit:<\/p>\n