Usually, "echo chamber" is used to criticize political polarization and avoiding news that challenge one's ideology. This seems like it wouldn't be a concern on a Q&A site, where we discuss not political opinions and current events, but factual questions about politically neutral questions. What sort of echo chamber could possibly develop among people discussing the Django API, for example?

Seems to me that the whole point of this site is to allow people to filter relevant posts via tags and site sections.

I don't know that we have a community on Django specifically, but there are a lot of "ideologies" in seemingly "factual" and "politically neutral" topics. For example, for programming and mathematics we have: choice of programming language, static vs dynamic typing, SQL vs NoSQL, desired level of abstraction, math foundations, pro- vs anti-category theory, levels of formality. I could even see valuing different trade-offs leading to splits even for the Django example. In my experience, primarily on Math.SE and CS.SE, people don't really vote based on these values, at least not directly.

Essentially what you are suggesting is a basic recommendation algorithm for the express purpose of forming filter bubbles. The difference with filtering based on tags is that it's an explicit decision made by the user with explicit criteria that the user can easily change. One of the problems with filter bubbles is people don't necessarily know that they are in one and can't easily get out of it.

I do think there's a difference between content-level "ideologies", e.g. political ones, and meta-level "ideologies", e.g. whether homework questions should be allowed. I think the latter is more relevant to what you're considering, and filter bubbles and echo chambers have different consequences for them. One problem with both cases is that you have multiple communities sharing the same virtual space while being ghosts to each other. Why not simply have separate communities then? For meta issues, the problem with this is the communities have conflicting policies. A new "homework" question comes in. One user downvotes and votes to close commenting that such questions aren't welcome based on that user seeing tons of such questions heavily downvoted. Another user comments that they are welcome because they see those same questions as heavily upvoted. You also have the issue of reputation in one community being shared across multiple.

choice of programming language, static vs dynamic typing, SQL vs NoSQL, desired level of abstraction

Are any of these on-topic for https://software.codidact.com/? They all seem irrelevant in that asking those things is against the rules to begin with, and I can't imagine anybody voting based on these. (can't comment on the math ones due to lack of experience)

meta-level "ideologies", e.g. whether homework questions should be allowed

Thank you for bringing this up - this is exactly what I'm trying to get at. To put it in your terminology, I'm saying let's create a filter bubble so that you see only the types of question people like you want to see. Except I'd prefer to not use the term "filter bubble", because it's a pejorative about political discussion.

while being ghosts to each other

So, no - because I'm not talking about filtering users. That would indeed not work well. In your example, people who upvote homework questions would see homework questions. People who don't upvote them wouldn't. Both groups would see non-homework questions, and interact on those, which they presumably both upvote. So really only some questions would be ghosts, not users. Does this not seem like a reasonable vision for a Q&A site? (I'm assuming the answer is no, as I'm becoming increasingly pessimistic about my proposal)

A question doesn't have to be explicitly about a topic for people's "ideological" views to be relevant. If someone asks how to implement Dijkstra's algorithm a person could downvote an answer because it's in a programming language they dislike. I'm sure people do do stuff like this, but, as I said, I don't think it happens to a meaningful amount.

A core aspect of your proposal is that the only problem with content you downvote is that you see it. Things like rep and policies are ones I've mentioned. Another is the new user experience. A lot of people don't like homework questions, because they don't want the site to devolve into a homework help site because of the reputation such sites have and the kind of people they attract. Those people will downvote the homework questions they see, but new users will see that they are upvoted overall (because more people who like homework questions will see it).

Your proposal also makes voting much more of a like/dislike than it is. While there is definitely some degree of like/dislike in voting, currently, the only thing voting does is communicate to others. It changes nothing about your experience with the site. In your scheme, voting directly impacts what you see, so you are incentivized to actually vote your preferences rather than vote on the merits of the question.

People's voting will be for a mixture of reasons, some of them political, and this proposal will not distinguish the reasons. This means that filtering based on prejudice will implicitly happen, which would be much easier to detect with tags than with automatically adjusted vote values.

I like the idea behind the proposal - a way of automatically filtering to reduce the amount of manual filtering with tags. However, in practice I would expect this to separate more than just question types. A homework tag could be used to filter based on whether you want to see homework questions. If you do want to see homework questions, you can still downvote bad homework questions (for whatever your personal definition of bad is). A vote adjustment will instead not only show you homework questions, it will show you only the homework questions you would upvote. This means the bad questions get only a positive reception. It also means they only get answers from people with certain biases.