9

Some sites on Stack Exchange have decided to unpin the asker-accepted answer in the default sort order. The official rationale for this (at least initially) was to promote fresh, up-to-date answers over outdated ones; on Retrocomputing, answers being out of date as such is not usually a problem (although dated history is not exactly unheard of either)… but the other problems with pinning the accepted answer remain.

The most pertinent one is that pinning exacerbates the Fastest Gun problem: askers tend to hastily mark answers as accepted, often choosing the first answer they receive, which may only superficially seem good. Pinning unfairly gives the accepted answer more visibility and therefore makes it more susceptible to drive-by upvotes (which we get a lot of, thanks to appearing in HNQ often), exacerbating the FGITW problem.

And the ‘better’ answer doesn’t even have to be posted seconds after; the Fastest Gun can also fire in slow motion. Not to blow my own trumpet, but it does make me sad that my answer about Windows 3.1 dithering, which I posted five years after the misleading accepted answer, will never appear first in the default sort order, even if it ever miraculously gets seven times as many upvotes as it has now: +27 / −0 compared to the accepted answer’s +171 / −1. (The only downvote being, of course, my own.) It’s probably the most egregious example of the FGITW problem, and how pinning makes the problem worse. Not even my vote-weighing script makes it any better, not without making some questionable adjustments to the weighing factors.

Given that, do we want to keep pinning accepted answers to the top?

1
  • I upvoted your answer about Windows 3.1 dithering. :)
    – ArrayBolt3
    Sep 25, 2022 at 4:54

3 Answers 3

7

If pinning accepted answers were useful for “problem”-type questions, SE sites focused on such questions would in all likelihood have kept them pinned when they were offered the choice; that’s not the case, and most (but not all) such SE sites have unpinned accepted answers. In many cases, the asking user is asking because they have a problem to solve, often in a hurry; it’s not unusual for them to accept the first answer which happens to (appear to) work for them, and then ignore further updates to their post. So sub-optimal answers do get accepted, even when they are not (yet) outdated. In some cases the accepted answers are actually incorrect; see this Unix.SE answer for example (the answer has been unaccepted since it was edited). I have another Unix.SE answer in mind where the answer is destructive, and was edited to warn readers, but I can’t find it now (charitably, one might assume that the asker accepted the answer before trying it, and couldn’t connect to their account after trying it).

RC.SE doesn’t see many questions where the poster is looking for an answer in a hurry, which is to be expected given the topic at hand (although we do have one user who periodically posts questions, and deletes them if they don’t get an acceptable answer within a few hours). That doesn’t mean that posters don’t accept answers in a hurry, and we do see questions where the poster accepts the first answer which looks vaguely sensible (and my hunch is that that really boils down to “is well-written” and/or “has lots of footnotes”).

RC.SE does however see lots of questions where the asker isn’t the best person to choose the “right” answer, especially not in the first few hours/days after the question is posted, before many of our more experienced users have even had a chance to see it, let alone write up a good answer. Even widely-acknowledged “retro experts” get things wrong sometimes, which isn’t surprising given the fallibility of human memory and record.

Ultimately, the question of whether or not to unpin is really how much influence do we want to give the question’s author? (See related research into, among other things, the impact of showing a post first on the page.)

There’s no way to satisfy all cases, so this really has to be a compromise. Some questions are “decidable” only by their author; perhaps the best example of these here are identification questions, e.g. . One might think these are particularly at risk of having incorrect answers voted higher than the correct answer (assuming there is one): it can take a while for the correct answer to be found, and in the mean-time, plausible answers can attract upvotes. But in practice that’s not much of a problem here; most of the correct answers are accepted and most-upvoted (as of this writing, there are a few cases where the accepted answer isn’t the most-upvoted, and one where the original asker wrote up their own answer but accepted another; most of the questions don’t have an accepted answer anyway).

It seems to me that for most questions here, the community is better-equipped to choose the best answer than the question’s author. What’s more, even though the focus of the site is on technology which stopped changing a long time ago, answers can become obsolete for a variety of reasons — e.g. they are link-mostly answers and the link dies, or they are historical questions where the correct answer isn’t obvious (as illustrated in the question here), or they are “how do I do this now”-type questions and the available solutions change (which is especially true for any “hobbyist” hardware to interface with older systems).

(This discussion is only relevant for the main site; I think that the meta site should pin accepted answers, but that would be a separate discussion. I’m also ignoring the effects of HNQ but that’s difficult to take into account.)

0
4

Pinning accepted answers is useful for "problem"-type questions, where the OP has the ability to verify the answer better than anybody else. At a guess, I think these make up more than half of our questions.

For history questions, yes, it's not ideal. However, that's less than a third of our questions, and most of those don't have this accepted answer issue; nearly a third of questions tagged don't even have accepted answers!

The most pertinent one is that pinning exacerbates the Fastest Gun problem: askers tend to hastily mark answers as accepted, often choosing the first answer they receive, which may only superficially seem good.

Is that true on this site? I made a hackish database query to verify that, but doing so frazzled my brain enough that I can no longer interpret its output. (Pretty sure it's a really inefficient query, but fortunately Retrocomputing isn't a very big site.)

TODO: run the query a few times, make a table of results, and decide whether I think this proposal is good for our site.

0
0

Since the long form may be too detailed, here a quick TL;DR:

  • Keep it as it is now.
  • The accepted should be pinned.
  • Accepting an answer is (usually) a quality statement
    • This is true for 'Problem' questions as Wizzwizz4 mentiones
    • But as well for any other, as the OP can/must decide which answer really hits his question.
  • Votes in contrast are quite often just likes for whatever reason.
  • Likes are not a quality statement
  • Changing that solves no problem and serves no purpose.
  • Changing will only worsen the experience of (new) users looking for a solution/answer to the same issue.

The most pertinent one is that pinning exacerbates the Fastest Gun problem

Somehow it seems as if you're obsessed with this assumption, having it used several times for attempted accusations.

Bringing it up here seems rather frivolous as the mechanic of sorting accepted answers to the top is exactly what prevents any benefit of answering first (see). After all, it's the questions author acting as filter with (some) knowledge about the topic - even if it's only newly acquired one.

askers tend to hastily mark answers as accepted, often choosing the first answer they receive, which may only superficially seem good.

Accepting an answer is, in my experience, usually a rather slow process on RC.SE. Unless it's an obvious case - like a pointer to a missing source. It didn't happen often that I have seen an answer accepted early one. I noticed this only a few (very few) times with obvious first time users, making it in itself notable.

It's quite common that askers do wait several days (or for better answers) until they decide to mark one as accepted. Sure, this may be different on other sites, but RC.SE is not the average ask for code site.

Using Wizzwizz4's script with a 'history' tag shows that for many questions the accepted is not the first given. Roughly half picks a later one. Given, that script my benefit from more refinement and especially providing summary results, but it gives a good first indication about an assumption being worth to follow up.

And the ‘better’ answer doesn’t even have to be posted seconds after; the Fastest Gun can also fire in slow motion.

Fixation on FGITW won't get more useful by adding sub-theories.

Not to blow my own trumpet, but it does make me sad that my answer about Windows 3.1 dithering,

That question possess a general problem for complex technical issues - they can't be easy followed without deep knowledge or a considerable effort to validate the claims. So people may rather act by 'gut-vaue'.

which I posted five years after the misleading accepted answer,

That adds of course as well.

It’s probably the most egregious example of the FGITW problem,

Sorry, but arguing with an answer made five years later and then calling anything before FGITW is by no standard a serious contribution.

and how pinning makes the problem worse.

But it will go up to second. Just the example doesn't help advance the issue, as the post happened so late, that it might, even without an accepted answer, never catch up.

(And I guess we all can agree that 'last post first' would be the worst of all sorting orders, making late coming 'something' the top answer)

Not even my vote-weighing script makes it any better, not without making some questionable adjustments to the weighing factors.

I would say any vote weighting factor is questionable. Who tells that a down-vote is in any way more telling than an up-vote? There are quite numerous reasons for down-votes. May it be malicious, like for not liking the one answering, over cultural(*1), all the way to simply not liking the content given/expecting a different answer.

The cultural part is for example easy to see with answers that state facts which do not go well with fan based beliefs - e.g. pointing out that the C64 is just a pimped up game console (max machine). Such answer receive an above average down-votes.

One of my highest ranking answers is as well a nice example, here the OP accepts a different answer. Both were given within 12 hours. Still did my answer receive about 3 times as many votes (151 vs. 53). Looking thru my answers shows several similar examples. Notably more than one from the same asker with the same result, so being a nice example for an askers bias.


Also, when talking about this, we should not forget people who post questions and then play moving target with the explicit goal of invalidating existing Answers they may not like for either reason. This as well screws answer ranking.


Many more examples where better answers are available can be made - as well as other where askers simply don't accept an answer despite being obvious, but I guess flaws of over detailed argumentation will outwight adding more.

Long story Short:

While there are some questions better left without an 'accepted' answer, I do believe that accepting is a very valuable tool where an asker can use his (new found) judgement to mark out the one he found more useful,which helped him to understand the issue he asked about.

Never forget, this is a social media site - we are people with all their rough edges and weird ways to interact. Any too 'fine' regulation will create even more trouble than they solve.

7
  • I'm not sure what you're trying to say about unpinning accepted answers. Should we, or shouldn't we?
    – wizzwizz4 Mod
    Sep 20, 2022 at 13:55
  • @wizzwizz4 I thought the conclusion (highlighted at the end) was clar. In any case I added a short version on top.
    – Raffzahn
    Sep 20, 2022 at 14:49
  • "Likes are not a quality statement" represents a fundamental misunderstanding of how Stack Exchange sites work. Move your mouse over the upvote icon, and leave it there for a few seconds; a tooltip will appear, explaining how you're meant to use the upvote button. Is that not how you've been using it?
    – wizzwizz4 Mod
    Sep 20, 2022 at 14:51
  • True, and in a perfect world Apple and Amazon would pay taxes while no one would ever vote for populists, right? But in a real world, people are used to up- and downvote on a like-base, from Twitter to Facebook and whatever else is out there. Just take Meta, an area where usually only people act that should be aware of the semantics - still there are downvotes. So tell me how can any part of a discussion ever be 'unhelpful' - unless it's complete off topic? Or take this recent answer which solved the issue the OP had, which I ...
    – Raffzahn
    Sep 20, 2022 at 15:15
  • ... would consider helpful above everything else - still it got downvotes (just not enough to give a badge - if there is any). Over the years I noticed quite often that downvotes are seem to be given on a like base - for example when stating something less charming about a usually hyped machine like the Amiga or C64, it will receive downvotes, despite being backed by hard facts. There is noting good in keeping up an idea of pure rational and good intending users. SE is a social media with all its pros and cons. (Also, no need to turn this personal).
    – Raffzahn
    Sep 20, 2022 at 15:15
  • On meta, upvotes are actually to signify agree and disagree. That's why they don't affect your rep. And sorry for turning it personal; you're one of the most active users on the site, and I haven't seen anything wrong with the way you were voting, so I was surprised when you said voting was about something it isn't.
    – wizzwizz4 Mod
    Sep 20, 2022 at 15:23
  • @wizzwizz4 Well, in practice it's visible that real world voting on RC.SE and even more meta, is not just as mentioned in some tooltip. And as you say, UP-votes are about agree, but then there people also use DOWN-votes seemingly to disagree. But that comes with the downside of not being symmetric - or do you know any serious voting systems with up and downvotes - and multiple thereof?
    – Raffzahn
    Sep 21, 2022 at 13:14

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .