Add title uniqueness to the heuristics for detecting low-quality questions

Question

This feature request is a response to Anyone want to clean up some bad titles?

Two days ago, Björn suggested that we warn users when they post a new question with an identical title to an existing question. However, given that there are 10,881 question pairs with identical titles on Stack Overflow (hat tip M. Tibbits) as well as ...

• 262 duplicate titles on Server Fault
• 103 duplicate titles on Super User

... I suggest that we take it a step further.

Let's extend the heuristics for detecting—and blocking—low-quality questions to enforce unique titles, or at least flag new questions with duplicate titles automatically.

We can leave the precise heuristics of detecting a duplicate title up to the Stack Overflow development team. I have no delusions about the fact that a few questions will "slip through the cracks." The goal here is to be proactive about minimizing the number of duplicates, and I think that the best place to start is the existing heuristics for detecting low-quality questions.

After all, duplicate question titles are bad for a number of reasons: They often indicate a question that is itself a duplicate, but perhaps just as importantly, they decrease the signal-to-noise ratio when searching for questions. I honestly can't think of any reason that duplicate question titles are helpful, unless you are asking for a solution to the same problem in a different programming language. Even then, I think that it would be acceptable to rephrase the question to say "How do I do X in (insert programming language here)?"

Frédéric makes a good point about the potential problems with this solution:

I have a hunch that if we do that, people will find ways to circumvent it (changing case or punctuation, adding random spam), and we'll end up stuck in a an arms race.

We may get stuck in an arms race with some users, but I honestly think that many people new to Stack Overflow just don't understand the etiquette of asking questions here. If we could even convince people to rephrase their question more clearly 20% of the time, I think that it would be a tremendous boon to the community. Furthermore, remember that heuristics for low-quality questions already exist, arms race or no.

Answer 1

As noted, the data shows that there's a good chance of duplication of titles, despite the existing UI that offers suggestions:

Basically, "we agree", so there will^* be a strong check on adding new duplicates, including a link to (one of the) pre-existing questions (so you can go and see if that solves the problem), or (much more likely, IMO) suggest clarification of the title:

Since titles are often edited after the fact, this is also checked during edit:

You can of course work around this with trivial changes to the question, but anything that makes the user:

1. consider whether a pre-existing question answers their question
2. make the title more precise about their problem (rather than "C# problem")

is a good thing.

^{*=as usual, this refers to the next build; please don't rush off testing it right now...}

Answer 2

I don't think this will help. What's wrong with my code? will just turn into What's wrong with this code?, What's wrong with my code, What's wrong with my Java?, and so on. Forcing unique titles is not the same as forcing good titles.

Obviously I agree with Frédéric in the comments above. Fewer dupe titles isn't a better situation unless the new titles are better. If they aren't fully changed to meaningful titles then you're only making crappy titles harder to find. Assuming users who post with bad titles will magically improve instead of doing the minimum to work around an obstacle requires more blind faith than I possess, personally.

In particular I don't think your example is any better. Tags take care of identifying the relevant language; putting it in the question title doesn't make it more informative.

Answer 3

When something like this is done, it needs a good UI.

My first idea would be to show the duplicate question, and say:

Could it be that you wanted to re-ask this question below?

If not, please select a title which clearly shows that your question is about something else than the question below.

This could be shown already when only the title is given, before one types the actual text.

On the other hand, in most cases of duplicate titles, both questions need to get better titles. So maybe simply silently (or after a warning) flagging both questions as "possible duplicate" or "bad title" for review could be actually more useful.

Answer 4

Statistically, this is bound to fail, especially going to the point of completely disallowing duplicate titles. There are only so many words used in Programming and in questions in general, and with a few million questions, it will become a huge task to create a new title everytime you make a question or edit one.

It's ok to throw some ideas, and present the dupes in the list as shown in Marc's answer, but there needs to be some slack on actually being able to use duplicate titles.