There are a number of posts here where people try to tell a story, to provide some sort of intuition pumping. Eliezer did it a lot and it worked for him. Assume you are not as good a writer, and stick to the business/academic writing format.
Start with a summary of your post (ideally, the post title should be an even shorter summary, e.g. "Alignment approach MOOSE fails on the dataset FORREST"), then a brief background and motivation, then your central point, then add a conclusion. Do not force the reader to go through a story ("you wake up in a windowless room, unsure of what is going on...") before they get to see your point, if any.
This is not a hard and fast rule, but a decent default.
This is what's great about the draft function on Lesswrong, and why I wish I used it more.
Given that you've come across something really good, that implies that it's worth a lot of people's time to read it. Which means it's worth your time to let it sit as a draft for a while, rearrange it, and spruce it up so that it's a little closer to the perfection they all deserve.
I'd really like to see content summaries for stories. I don't read much of the fiction posted here, because I don't know if it's going to be worthwhile.
I would like to shift the balance a little bit toward the more straightforward posts as well. I look forward to seeing if this advice accomplishes this.
For topics that have been discussed before, stories or dialogs can add some depth. For things that I don't have the right context for me, I tend to bounce off pretty hard.
I don't agree with this as a principle, although it may be a correct output. I think the notion of "a decent default" misses the mark compared to "think about your audience and the key elements of your message before deciding your form and tone."
To use a simple metaphor, if you need to anchor two pieces of wood together, a hammer and nails are usually going to be the quickest and cheapest way to do it. A drill and screws are often overkill. However I don't think that makes the hammer and nails the default; I think it makes them the correct tool in the majority of situations and the drill and screws the correct tool in a minority, but whichever one you end up using you should think about what type of stress your join will be under and use the right tool for the job.
There are a number of posts here where people try to tell a story, to provide some sort of intuition pumping. Eliezer did it a lot and it worked for him. Assume you are not as good a writer, and stick to the business/academic writing format.
Start with a summary of your post (ideally, the post title should be an even shorter summary, e.g. "Alignment approach MOOSE fails on the dataset FORREST"), then a brief background and motivation, then your central point, then add a conclusion. Do not force the reader to go through a story ("you wake up in a windowless room, unsure of what is going on...") before they get to see your point, if any.
This is not a hard and fast rule, but a decent default.