This page is about the belief notices (or tags or markers) that are displayed at the top of pages of this site. The belief or epistemic status of a page efficiently communicates to the reader my overall confidence in the veracity of the claims made on the page.

In practice, belief notices are most useful when I want to publish something on here that I find interesting but I’m not especially confident about. In other words, it’s often a shorthand for something like, “I’ve only briefly thought about this topic, and haven’t really spent much effort working on this page, so even though I think it’s worth making public, you shouldn’t take this page very seriously, nor should you assume that I believe the things I’m writing”.

More abstractly, publishing content on one’s personal site is not just a statement (i.e. the content on the page), but also a meta-statement—that one believes in what is said strongly enough to publish it on one’s site. I think this probably tends to make people more averse to publishing content that might come across as unconventional. Of course, there is still a meta-statement even with belief tags; one can always attempt to pry into the psychology of the writer, e.g. speculating along the lines of “Well, the fact that they’re willing to make this public—even while marking it as ‘unlikely’—still shows that they’re a heretic”. While I recognize that this might be a problem, I still think belief notices solve most of the initial problem without being much of a burden.

The “standard” scale of belief that I use is as follows:

(I think these are self-explanatory; see the history section for the details.)

In addition to the above possibilities, I also use the following:

For fictional work, which may include work that ironically resembles non-fictional work.
For rants and other writing that contain strong emotions. Pages tagged with this are probably especially epistemically unreliable.
[not displayed]
For pages where the notion of “belief” does not make sense (e.g. the front page of this site). This replaces gwern’s and Muflax’s “log” (see history).


I took the idea of belief tags from gwern, who in turn got it from Muflax.

Muflax explains:

I didn’t want to throw away old texts just because I changed my mind, but I also couldn’t let them sit around without an annotation or someone might think I still endorsed them.

Here is gwern’s section on belief tags:

Most of the metadata in each page is self-explanatory: the date is the last time the file was modified, the tags are categorization, etc. The “status” tag describes the state of completion: whether it’s a pile of links & snippets & “notes”, or whether it is a “draft” which at least has some structure and conveys a coherent thesis, or it’s a well-developed draft which could be described as “in progress”, and finally when a page is done - in lieu of additional material turning up - it is simply “finished”.

The “belief” tag is a little more unusual. I stole the idea from Muflax’s “epistemic state” tags; I use the same meaning for “log” for collections of data or links (“log entries that simply describe what happened without any judgment or reflection”) personal or reflective writing can be tagged “emotional” (“some cluster of ideas that got itself entangled with a complex emotional state, and I needed to externalize it to even look at it; in no way endorsed, but occasionally necessary (similar to fiction)”), and “fiction” needs no explanation (every author has some reason for writing the story or poem they do, but not even they always know whether it is an expression of their deepest fears, desires, history, or simply random thoughts). I drop his other tags in favor of giving my subjective probability using the “Kesselman List of Estimative Words”:

  1. “certain”
  2. “highly likely”
  3. “likely”
  4. “possible” (my preference over Kesselman’s “Chances a Little Better [or Less]”)
  5. “unlikely”
  6. “highly unlikely”
  7. “remote”
  8. “impossible”

These are used to express my feeling about how well-supported the essay is, or how likely it is the overall ideas are right. (Of course, an interesting idea may be worth writing about even if very wrong, and even a long shot may be profitable to examine if the potential payoff is large enough.)


Here is a comment I once placed (at a private location):

Can you elaborate on this? Would you prefer that these articles and blog posts not be published (since if the authors can’t add the epistemic status headers, they would be unwilling to publish them as if they believe it)? Maybe you want them to publish less “egregious content”, period, or disclaim each claim in their article by qualifying them, but that takes a lot more effort (at least for some people). If we think of writing as being on a “sloppiness scale”, then these epistemically “egregious” articles might be at one end, with higher quality articles in the middle, and maybe academic journal articles and published books at the other end. Relative to journal articles and books, even high-quality blog posts might seem lacking because e.g. they might not cite every claim, might contain grammatical errors, etc., but not meeting some arbitrary quality threshold doesn’t subtract from the value it provides. Similarly even articles with epistemic status headers that seem of poor quality occasionally contain valuable insight, and if adding these headers means more of these valuable insights are made public, then I think that’s a net benefit for all of us.

Somewhat related (to my last point that “epistemically egregious content” can still be valuable) is Ben Hoffman’s post on philosophers:

Someone once described Michael Vassar to me as paying attention not to the conclusions but to the intellectual processes of people he’s talking to, not in order to conform to them, but out of curiosity, in case they are using some valuable heuristic he should add to his toolbox. I often find that I get something different than usual from philosophers because I read them this way. I find it jarring when people casually refer to Nietzsche’s philosophy (to name one example) to reference some proposition or other he’s famous for asserting. It seems natural to me to be referring to his methods, but his specific conclusions seem like almost a weird irrelevancy. I think this is one of the most valuable things I got out of my St. John’s education – the ability to read thinkers to figure out what their project was, rather than a bunch of specific propositions they were advancing. This is my attempt to share this way of reading, by working through an example.


But now, given these belief tags, one might ask what the frequency of each belief on a site such as this one should be. Having everything be “certain” to “likely” might be how a typical site or blog operates; i.e. the writer only publishes what they believe. But this sounds limiting—and this is exactly why belief tags are useful. But what of the other extreme?—In other words, what if one were to only write what one believed to be false? A successful such execution would display the writer’s creativity, to be sure, but is unlikely to contain much useful information.

One might also ask what the “natural frequency” of each belief might be, if a writer were to log all of their thoughts …

See also