the title. It is always the title. But if there would be topic preview too then it can be a reason too.
Skimming? AI-explaining took way too long, but is it same as browsing? If yes, then everytime on new forum, but here never. Latest, or searching
I never use those views, anywhere. Just latest. And hot/top are practically same and unseen is same as latest on new place, but here it gives too many ancient hits
Possibly stating the obvious, but I’m not looking through the topic list when I’m looking for the answer to a question. In that case, I’m either using Discourse search, or my browser’s search.
Apparently a question mark in the title will get me to click on a topic
In general, I’m looking for either questions I can answer, or topics I’m interested in reading about. Essentially, I’m trying to summarize the topic based on its title.
On Meta, there’s usually enough information in the title for me to make a decision about what to click on. That’s because the titles are clear, and I’ve got a fairly deep knowledge of the subject.
Taking another forum as an example, titles like “Issue with assignment 2” don’t give a lot to go by. AI generated summaries would be useful for cases like that. I’m actually on a forum now where I’m interested in the subject, but find the latest topics page completely uninteresting. There’s not enough information to grab onto.
On forums that I visit regularly, I rarely skim past the “last visit”. On forums that I don’t visit regularly, I rarely skim past the first page of results.
I always use the latest view, either for the forum as a whole, or category latest views for forums where I’m only interested in specific categories. Hot/top feel a bit like hype to me. Unseen feels overly complicated. I think that’s a separate topic though.
On Meta and other communities, I tend to hang around in limited categories about things I understand. So for example, in a new topic list I might only click on 8 of 30 new topics.
Another factor that means I don’t click on a new topic is if it has been solved. If I want to find solved topics I go to the search.
And finally, the title clarity. However, sometimes I’m just curious in the title wording even if its not clear to see what the topic entails. I look for topics where I think I could be useful which on Meta is mainly support.
I sometimes skim through to have a look at other topics I might find interesting. So although I do have a quick skim I don’t just “keep scrolling”.
I just stick to ‘Latest’. Its just what I’m most comfortable with. The other views aren’t too relevant. For example, just because a topic is ‘Top’ doesn’t mean it will interest me.
Here’s something I don’t like, and which AI could possibly help with: titles which say nothing, like “A new problem” or “Help me” or “Oh no, not again”
(I don’t run any AI, but I do run with the topic list previews component, which includes a good sized snippet from the head post, and thumbnail if there happens to be an image in there)
If I’m browsing on Meta or Intfiction discourse forum to catch up on forums that i haven’t checked in on for a while, I like to go to Top/Hot and see which topics have the most replies (if i could see amount of hearts, i would click on the ones with most hearts too). This is to help me find what are most interesting conversations right now.
In communities where I’m a regular user, I mostly judge by title and category.
In my “home community” where I’m a moderator, I read, or at least skim through, probably about 60-80% of the topics. When I skip topics, it is usually because a) I have checked the topic recently, but it’s neither particularly interesting to me nor an overly heated discussion, or b) it’s from one of the categories with a lot of small talk content.
Unfortunately, I don’t think AI summaries would help a lot with either case. For me, they basically only come into play, if I am starting to read a very long topic with a lot of posts, or if the posts themselves are quite long.
It looks like everyone has their ways of navigating topics - sticking to specific categories/tags/view, searching for specific ones, clicking around, using Summarize, evaluating titles etc.
I think the distinction of whether one is visiting a community they are familiar with vs a brand new, one does switch things up in terms of how all the content is viewed (or valued I should say)
Hopefully the short summaries in the topic view to help give this exact context, especially useful for bad titles in our testing.
We are hoping with more context, better LLMs and prompt tweaking - this gets improved over time.
Yeah, I could see a combination of sentiment analysis and summaries being beneficial to moderators. But it would need to be pretty accurate, support languages other than English and work with a self-hosted LLM (assuming we get to a point where the average root server has enough AI capabilities for inference on a larger scale).
