Should we delete topics that end up not containing any useful information?

I have been thinking about this, too. Should we delete topics that end up not containing any useful information (similarly to those that have been posted by error)?

@pieper @jcfr @ihnorton @fedorov

I don’t like editing history as it leads to confusion for the people who remember seeing something and then it is gone (or changed).

1 Like

Should we delete topics that end up not containing any useful information (similarly to those that have been posted by error)?

vs

don’t like editing history as it leads to confusion for the people who remember seeing something and then it is gone (or changed).

Could go either way.

In that particular case, I suspect the user didn’t realize Slicer should be restarted.

Marking the topic as resolved is a good first step i think. It shows the questions was addressed in some way.

Such topic wastes time of all future forum readers that will try to find information on the site. If I search something on a forum then I read the first 5-10 topics that are found. Thus, any garbage topic left on the site reduces the chance of finding relevant answers.

Any deleted content is not actually deleted for a while (several weeks?), just indicated that it is deleted (it shows up with red borders or something like that), so people who get email notification and click on the link can still see what’s going on.

I agree that it may be difficult to decipher history when posts are edited/deleted, but this impacts magnitudes less number of people (only those signed up to get email notification about all topics; or those who set to follow a particular topic) than people who are trying to find information by searching on the site or using google. So, I think we should prioritize good final search results over accurate history.

I think removing noise is more beneficial overall. If we deem a thread not useful and remove it, then I think it means that whoever wants to find it, shouldn’t, because it won’t help them. So we should only remove topics that have no added value whatsoever.

I second @pieper, I would not edit history.

In addition to the reasons Steve mentioned, repeated questions on the same topic is an indication that users have troubles locating the answers, which may or may not be due to the noise in the forum. That information is important on its own. Also, evaluation of the post’s value is subjective. Once we start deleting posts, this sets us on a slippery slope. I would only remove spam messages.

I would also hope that most useful/comprehensive posts will naturally separate from noise and repeat posts due to larger number of likes, I would let that happen on its own.

What I would suggest is to improve FAQ to make the topics that come up again and again more prominent. Maybe featuring those in the ReadTheDocs version could help. We could also add guidelines for posting a question that would instruct users to check specific page with FAQ first before posting a new message.

What if we don’t even know the question based on the thread? For example in case of this one, I’m not sure what the actual problem was. Is keeping this one any useful?
(Edit: “this one” is the topic where this originally was, here)

I agree on the “slippery slope” notion, but we could have a brief checklist that makes this decision objective.

There is no such thing as a slippery slope. We can establish some guidelines that we follow and stick to it. If we want to change the rules then we do that.

For example, when a user essentially revokes a question (“sorry, my bad”) and there have been no useful answers, then I think it is safe to delete the post.

I agree that the fact of even receiving a question is a piece of valuable information, but we do not need to keep this information in the forum. For example, a note can be added to a related mantis issue.

@cpinter I think this post is in fact a great example that demonstrates why posts like this should not be deleted.

I may be wrong of course, but my interpretation of the problem that user had is that the module that PET-DICOM extension installs is a DICOM plugin, which is an “invisible” module and cannot be found in the list of modules. I can definitely see how this can be confusing to the users. This is also an example where for one developer this post could be noise, but in fact it can provide valuable insight for improving documentation. You cannot address something like this with a checklist. I also don’t quite see what why keeping this post would be particularly problematic.

In the situations where a post does not make sense (at least for the specific developer/maintainer at a given time), we could assign a descriptive tag to lump all such topics together. Developers/maintainers could on their spare time go over the topics in that category to try to get insight on the trends and interpret issues better.

About mantis, it contains interpretations and well-defined technical issues, while I think of the forum as “raw” user feedback. I am not convinced of the value of deleting such evidence documents.

This is what I mean, that we don’t know the question. We can guess around, but if we don’t know what the question was, the information is not useful.

At this point this is equiuvalent of thinking about design, that we do anyway. We already know that extensions containing DICOM plugins are not obvious to use. But this is not coming from a vague question where we cannot be sure what it refers to.

The slippery slope is that current simple guidelines (i.e., don’t delete unless spam) will need to be more complex and will be changing over time to address things as we discover them, and they will have to become subjective (right now spam vs non-spam can be pretty objective). We are dealing with humans, not machines.

What if you have a non-native speaker who has troubles explaining the issue? How would you feel if you posted a question on a forum, which was not understood, but then you got busy with other things and didn’t respond in time to clarify, just to find that your post is gone? Can you demonstrate a precedent in another forum where a similar policy is successfully implemented?

Those rules can also be interpreted as censorship, as users might feel passionate about their role in the community.

With all the discussion so far, I respectfully disagree with removing any posts, unless they are clearly abusive or can be attributed to spam.

As an aside, we are now in a situation where we have a dissent in opinion, and I guess we could start with establishing a checklist of how to resolve such situation. Or is this the case where “benevolent dictator rule” applies?

I agree with most of what you say. I’d feel very bad if somebody deleted my post although I’m trying to do my best. However there are extreme cases that we can cover. Here are two already, which are super easy to decide, and I don’t think there is any subjectivity involved.

  1. Person asking the question revokes the question
  2. Person asking the question replies that it was solved without giving any details about the method of solution

This second one even came up just a week ago at a conference, as an example how can stack exchange can be annoying. Even if it’s done with good intentions, it is just bad practice that could be a rule of our forum.

There are moderators on every forum, and if the forum has rules, then we can work to enforce them. In this case the goal is to avoid noise.

I agree posts in category 2 can be annoying, but I liked the gentle reminder you sent recently @cpinter and even if the original poser never comes to fix it, in a way, it’s exactly to @fedorov’s point that if someone finds that thread it will be a hint that they shouldn’t make posts like that (if it were deleted then nobody would ever find it!).

In any case, I wish it were easier to find old posts that are relevant. The search results aren’t as useful as they could be maybe in part because of content-free threads, but I think mostly because we aren’t tagging/liking or otherwise curating as much as we could to make good content come to the top.

1 Like

Yes the ultimate goal is to make search more usable. If there is a way to “star” threads we think are most useful in any given category, and which show up on the top when it fits the search category, then it would solve the problem. If all of us know that it’s possible and the way it can be done, then it would be a good way to facilitate finding relevant information. (Although this is also a subjective process! So in that sense not different from the trimming approach)

Good point about the wider utility of my reminder, but I don’t think I’ll remind everyone :slight_smile: I just don’t open some topics because they are out of my expertise… However if there is a way to skew search results in favor of the most useful posts, then I’m OK with leaving useless posts up.

It is already there:

If we can influence the default settings then it is useful. If you have to change something it is not, because nobody will do that.

It is subjective, but it is not lossy! :smiley:

The rating system is a self-regulating distributed mechanism that is familiar to all users. We should just keep encouraging folks to use it more. Even if just the core group of active users starts giving more “thumbs up” to threads that are good, this will quickly help reduce the visibility for threads like the ones you want to remove, since by definition they will not get any "thumbs up"s.

Plus, as was mentioned earlier, we should use topic tags more. This is again a place where some guidelines would help, and where we have better control of the outcome, since tags can be curated by core contributors (and they are also not going to lead to any content loss).

Here’s a post on (if I understand correctly!) how to configure discourse to sort by the number likes by default: https://meta.discourse.org/t/sort-filter-by-likes/26193

1 Like

I made an observation about the “Most Liked” option, which was not obvious to me, and which makes me wonder if it’s the good choice.

I liked two of Andras’ comments in the best thread about an issue that came up many times (because I didn’t want to like the question itself), and found that the search results point to any post/comment with the most likes, and not just the threads. See the multiple occurrences of “Save volume rendering as STL file” in this search.

In this form I don’t find this very useful. It would be much better if added all the likes together and showed the threads. I don’t want to dismiss anything, just an observarion so that we can make the most educated decision.