How Do We Write Science? - by CasualPhysicsEnjoyer

Casual Physics Enjoyer

SubscribeSign in

How Do We Write Science?<br>Or... why to invest in brain transmission technology....<br>CasualPhysicsEnjoyer<br>Jul 22, 2026

Share

With Arina Berezovsky​, in preparation for a dinner with the Independent Science Society.<br>Presumably the first job of the scientist is to experiment with the natural world and then figure out some sort of general law that describes it. This doesn't seem controversial, and even Karl Popper uses this rough definition at the beginning of 'The Logic Of Scientific Discovery'. Drop an orange a hundred times, then time it to see how fast it falls down. You've now discovered a physical law of oranges!<br>​But suppose that scientists did all this and didn't write down what they were doing. If they told no one, the discovered law would probably just unceremoniously die with the scientist. If that happens, the law may as well have never been discovered in the first place. This is a shame, and in this case it's wasted effort. If they tried to tell people purely by word of mouth, this could work, but it's inefficient and the information transfer would be lossy in some way, like a game of Chinese whispers.<br>Because of this, there are likely to be considerations in the crucial second part of the job of the scientist.<br>This second job is broadcasting their proposed laws, along with why they think the evidence points to this law. Or in simpler words, communicate the science to other people. But what is the best way to do this? And the use of the word broadcasting, and not necessarily writing, is an intentional choice here.<br>​One classic way of broadcasting science is through academic papers and peer review. This came at a time when the written word was pretty much the only way to transmit information easily over long distances. But there is no guarantee that the classic style of papers remains the optimal way of doing the second step of science. In fact, we might actually be worse off looking at the written word: because so much is being generated by LLMs, there is a lot of slop out there, and the reviewing process from publishers is unclear.<br>And so if the standard way of writing papers is suspect, what are we meant to do instead?<br>The main question to ask is this: what are we really trying to do when we communicate science? Well if we are trying to get to the truth of the state of the natural world through an experiment, then communicating this well means transferring all of the information about an experiment well .<br>In the limit of infinite technology, the best way to communicate an experiment would be to record your conscious experience in doing something, all in infinite detail, sight, smells, internal monologue and all, and then share this cognitive recording with others. That way people could see exactly what you're doing, why you did it, and if you had any bad motivations in trying to measure something.<br>But unfortunately for us, we don't have the technology to do this yet, and so we have to settle for lower fidelity mediums of sharing experience. This means that we have some tradeoffs and questions to make when deciding how to communicate things.<br>But in any case, whilst we figure out the details, all the decisions made in communicating experiments should be towards giving the most detailed reflection of the scientist's reality at the time .<br>To do this, we have some hard questions to answer. What medium do we use? How can we trade off clarity and detail? Are we meant to write stuff that goes wrong? How do we decide what's important?<br>Answering these questions has been a long running side quest of this blog. But for now my rough conclusion from thinking about writing science is to show all my thinking when doing something, including intermediate steps. And also to give as much information as possible about my process.<br>For some reason, until recently, culture didn't seem to value thinking out loud, or showing intermediate steps. Not sure why. Maybe this is because of the general vibe about caring about finished products? Previously it would have been hard to show all your thinking, because it didn't fit the vibe of a fancy literary work you needed to get into a learned society? Maybe it was hard because we didn't have cameras?<br>But this seems to be changing now. People like reading about the process, like the unfinished sculptures of Michelangelo in the Academia Gallery in Florence. Or for example Faraday's diary has loads of cool musings about different shapes of wires that might generate electricity.<br>But now more scientists are writing their results or their thinking on blogs, like Terence Tao (and now Twitter, where people are posting their math results). People are even posting their LLM chats.<br>We also have other mediums. We have cameras now, so why not just send videos of experimental procedures instead of writing up protocols? One thing I've tried to do on the blog is film everything I do, then get Claude...