6

This is a wide-reaching question not just about Software Recommendations (SR herein) but also about the future of sites you would hope will feed into it (the sites like Stack Overflow, Super User, Ask Ubuntu and Think Different).

I am not experienced here but I have tried my best to look around at the existing definition and beta questions and answers. Tim Post's "ground rules" post seems to share some of my concerns but I don't see nearly enough solutions to the problems I'll outline below.

Please be patient with me, even if that means spelling out what should be obvious.


As a moderator on Ask Ubuntu and an fairly experienced user elsewhere on SE I can see that it's always going to be natural for people to ask for suggestions. What laptop should I use? What speakers go best with my amp? What database engine should I use? What cars should I consider? What detergent should I use? The list is near-infinite. People want answers and they're going to ask them.

So I can see why somebody thought SR was a good idea —there is certainly existing demand on the SE network for this type of question— but my first question is, as an Ubuntu user, why is it more logical or better to come here over Ask Ubuntu to get recommendations? As somebody looking for answers for Ubuntu, why would I come here?

On Ask Ubuntu, we have classed the software-recommendation tag as an occasionally necessary evil. It's meta and not the "bad tag"-misnomer-meta but the actual "it describes the type of question"-meta. By extension, isn't SR a meta site? An entire site dedicated to a type of question with only the vaguest idea of "software" as the glue. This seems like a pretty weak raison d'être.

Conversely, why just software? Why not handle all recommendations? We get plenty of people asking what hardware they should buy for Ubuntu and I know Bicycles.SE gets a fair amount of shopping-related traffic, I'm sure other sites have other types of recommendation. I don't understand the reason for the arbitrary scope.

How are the existing criticisms about recommendation questions mitigated here? The "SE Format" has oft been used as an excuse for closing recommendations on other sites (infinite lists, opinions, rot, etc). I see a lot of fanfare about quality and scoping but these are not the things in the minds of the masses who will be asking the questions after beta. They simply want answers and all the quality in the world doesn't help with some of the issues:

  • How do you combat older answers outranking newer, better answers?
  • How do people vote objectively for a subjective topic? What wins between a well furnished recommendation for crappy software versus an awful one-liner for the best choice? Should somebody try every solution before they vote?
  • How do you handle near-duplicates? There seems to be a focus on making questions having a pin-point scope (with an aim to combat infinte-list-of-X syndrome) but doesn't that just mean there'll be an amorphous blob of nigh-on-identical questions? How do you deal with that? Just keep handling them?
  • How do you handle exact duplicates with "... but none of those work for me" addendums? Bounties only go so far.
  • And what about exact duplicates where there has been significant rot, either in the question or its answers?
  • An entire site of this seems like the most ripe target for astroturfing and illicit product placement. Can you do anything to protect against that?
  • With such exacting rules for questions, how do you handle people who don't meet them? What about when people want multi-purpose or deliberately open-ended software?

And finally, drawing on both ends of this little investigation, if we can mitigate these issues here, why can't we just do that on Ask Ubuntu, Super User, Server Fault and Programmers.SE (to name a few)?


While I can see that asking these questions here is a bit like wrapping myself in bacon and running into a lions' den, I'm honestly not trying to crash your party.

I want to know how SR is going to fix the very real problem that affects most SE sites. At the moment, this just seems to be more of a shift than a solution. The same old problems as using the software-recommendation meta-tags apply, but now they're inexplicably in one place, caked in extra rules that people aren't going to read before they ask.

Given the enthusiasm for it, it seems likely I'm missing something. Could somebody be a dear and poke me in the right direction?

  • 3
    To cherry pick a little, we're wouldn't create a general "request for anything" site because we don't scope sites so wide that would would be "shouting out a calculus question in a football stadium" — stackexchange.com/about – Robert Cartaino Feb 11 '14 at 16:56
  • And on the main site, this question would probably be closed as "too broad", with comments to "stick to one question per post". You rise a lot of questions here, many of which already have been dealt with on this Meta. To get good answers, I'd say this one has to be split into at least 4 (if not more) parts as well. Hard to cover them all together in one "thread" – hard to follow that if we tried. – Izzy Feb 11 '14 at 17:18
  • For one part of your question, please see: How to deal with similar but not duplicate questions. For another, I can just second Tim's comment (and would even go so far as to leave games out completely – but that's my personal opinion, and I can "ignore" them if decided differently). – Izzy Feb 11 '14 at 17:22
  • @Izzy While long and winding, I'm really only after an answer to one question (that the whole post feeds into from different angles): "Why is SR the right solution for the SE recommendation problem?" – Oli Feb 11 '14 at 17:23
  • Is it? Where does it say so? Why should it be? Was it created for that reason? Do we pretend to be the "saviors of the universe"? I don't think this was the intention. And I personally would answer all the questions I just made in this comment with a clear "No". But I see no problem with that: this site has a good scope, I enjoy to participate here, and I feel it very help- and useful: this is not the "solution to the SE rec problem", but the "place to get software recommendations". – Izzy Feb 11 '14 at 17:29
9

At a recent Stack Exchange company summit, we discussed how we might be able to host a broader range of content, including things like "recommendation questions." We listed many of the problems, as well as possible solutions for overcoming the inherent challenges these questions present. But even if we could eliminated all the technical issues of ferreting out bad questions/bad answers, we were left with a social problem that isn't as easily fixed through software.

These questions break our reputation system.

Commonly known as the bike-shed example, on any other site, there would be a disproportionate level of participation in these trivial issues where basically anyone can chime in with an opinion. We simply do not want to overrun an expert community where the top-ranking users are those who were quickest to type in the name of a well known application.

But this site (Software Recommendations) is a scope unto its own. There is nothing else here to "break the reputation system" or "keep people from answering real questions," or any other activities which might otherwise detract from the rest of a site.

If you haven't read this far, you might conclude that {software/hardware/equipment/shopping} recommendations can never work on our sites. Not so. The way we mitigated the bike-shed problem on this site is by requiring that problem statements be specific enough that only a few people out of many could answer them. We don't make the internet better by simply copying the top results of any Google search and calling it Q&A.

But avoiding such questions (and vetting the answers) takes a lot of self-discipline and community self-moderation which isn't easily integrated into a mainstream site. It will be interesting to see how this site fares when it enters public beta. The main concern is that we'll have to beat unwary users over the head with a lot of obscure meta discussions with unfathomable rules just to ask a simple question. Few people are willing to take those knocks just to ask "what do you recommend?" At the time time, it's either keep up the quality or end this experiment. It a tough balancing act.

But who knows; we may just be able do do this better than anyone else on the Internet.

  • I certainly don't think they'll never work on any site. I'm for them in general because they are real problems. The old fashioned way of dealing with the "bike shed" problem on AU was just to enforce CW on subjective questions like recommendations. If reputation is the only aspect of that problem couldn't there just be a way of hooking up *-recommendation tags to automatically be CW? Layer on the other rules ("be specific" et al) and I'm back to wondering why we can't just handle recommendations on the right sites. – Oli Feb 11 '14 at 22:34
4

This answer draws from my experience mainly on

In a nutshell, recommendations can work, but they need strong quality control.

Many sites (SO, SU, SF&F) have decided to ban recommendations altogether rather than apply quality control. Others (U&L, Literature) apply quality control (see Literature's book recommendation policy).

On a site like SO, SU, etc., for most questions, an answer solves a problem, in a way that can be tested empirically. Either the solution works or it doesn't. This gives a direct, concrete test for answers: does it work? Yes → upvote, no → downvote. A solution may be preferable to another because it is simpler, cheaper, or more widely applicable; this mostly affects upvoting vs not voting. Another type of questions is the understanding question: explaining a concept in context; a good answer explains the concept well. A better explanation yields more upvotes. On a site like SF&F or Literature, explanations make up the bulk of the question.

Recommendations are a bit peculiar because you can have zero-effort answers with no concrete test. What usually happens is that people upvote the answer if they've heard of the recommended item. This is at best a popularity contest, and in fact in practice primarily a race as to who posts first. For recommendation questions to produce quality answers, we must insist on quality answers that explain why the recommended product is good.

Recommendations are also very easy to ask about. You don't really need prior understanding or effort, unlike “debug this code” (which requires that you go through the effort of writing that code) or “explain this” (which often is asked by someone who has tried, and failed, to understand).

So both questions and answers to recommendation questions suffer from the fact that it's particularly easy to ask a bad one. This is why the average quality is bad.

The premise of this site is to apply quality control to posts that goes beyond what is usually done on sites such as the Trilogy. Note that this here is my opinion; it is controversial: some of the participants here want a free-for-all.

So, should recommendations live on their respective sites (SO, SU, AU, etc.)? Ideally, yes. The problem is that in practice, most of these sites have failed to apply quality control to recommendation questions. I see the dedicated SR site as a way to enforce quality control in a more uniform setting, and to refine our definition of quality and the methods we can use to foster and enforce quality.

For me, the best way for SR to succeed is for it to become redundant, by coming up with quality control in a form that other sites can apply. I'm not very confident in the ability of SO and SU to apply better quality control; I'm only slightly more confident in AU, which is a bit overwhelmed under requests of all kinds from first-time users.


Now addressing some of your specific concerns.

as an Ubuntu user, why is it more logical or better to come here over Ask Ubuntu to get recommendations?

That is an issue with a separate site. In practice: you ask for a recommendation on AU, your question gets closed and you're refered to SR. In this process, SR's question quality guidelines should be shoved in your face. (There won't be a test, but really, there should…)

How do you combat older answers outranking newer, better answers?

Mainly, by requiring requiring good answers. One-liner answers tend to be rated by age, but on answers with in-depth explanations, the first-timer bonus isn't nearly as big and newer answers have a chance.

How do people vote objectively for a subjective topic?

Recommendations are not that subjective. A good answer should include detailed explanations of why it is a good answer, which makes it easy to rate: does the explanation convince you that the application is a good fit? Hopefully, if the facts in the answers are wrong, someone will comment. This is no different from, say, voting on Stack Overflow where most voters won't actually try the code provided in the answer.

What wins between a well furnished recommendation for crappy software versus an awful one-liner for the best choice?

If we're doing our job correctly, a third answer that is well-written and recommends the best choice.

How do you handle near-duplicates?

See How to deal with similar but not duplicate questions and Dr. Strangedupe: Or, How I Learned to Stop Worrying And Love Duplication

How do you handle exact duplicates with "... but none of those work for me" addendums?

Why don't they work for you? It has to be because you have additional requirements. Edit your question to add those requirements and your question will no longer be a duplicate.

And what about exact duplicates where there has been significant rot, either in the question or its answers?

Most software doesn't change that quickly. If a question is obsolete, edit it to indicate what kind of obsolete state of the art it's referring to (e.g. requirement: works out of the box on Ubuntu warty), and don't treat it as a duplicate target for questions that do not have this obsolescent requirement.

An entire site of this seems like the most ripe target for astroturfing and illicit product placement. Can you do anything to protect against that?

Flag as spam. As announced in the ground rules, we will be particularly strict on astroturfing. Answers that consist only of ad copy (whether coming from the vendor or not) should be summarily deleted anyway (see above).

With such exacting rules for questions, how do you handle people who don't meet them?

Close. Close. Close. And each time, comment to explain in what way the question is deficient and point to the meta thread (I expect we'll have a dedicated close reason or two, we need a little more time to work them out).

What about when people want multi-purpose or deliberately open-ended software?

A good recommendation question describes both a task and requirements. A task can be both broad and precise.

if we can mitigate these issues here, why can't we just do that on Ask Ubuntu, Super User, Server Fault and Programmers.SE (to name a few)?

See above — SU and SO, especially, aren't even trying to foster and promote good recommendation questions and instead ban them outright. Which is comprehensible, but not ideal.

Note that U&L doesn't really have a problem with recommendation questions. We do ask that people provide precise purpose and requirements, and a lot of distribution choice questions are pure bikeshed, but it's a minor issue on U&L, probably because of our elitist appearance¹.

Even though U&L is not, in fact, elitist — we do try to provide didactic answers to elementary questions — it looks this way, in part due to its self-sustaining domination of command-line-based content.

  • I really appreciate you being thorough. With the assumption that everything goes to this plan, I can see how this would work. I also agree that the bigger the site, the less likely they are going to be able to handle a strict format of question like is being suggested here... However, my inner moderator has seen things. I know how bad things can get and if you survive, you're going to be the close-OT dumping ground for many sites. Scale will quickly be as big an issue here as anywhere else. Has anybody discussed altering the /questions/ask form to make it more for-purpose? – Oli Feb 11 '14 at 22:50
  • @Oli My inner moderator has also seen things, even if the sites I've moderated have nowhere near the volume of AU let alone SO. I don't remember if /questions/ask has come up yet, but it's on the logical path. We're spending another week in private beta to hash out this kind of things. I don't know if we'll have enough experience to write a /questions/ask from the private beta alone. We will ask not to be mentioned in close-OT reasons for the first 6–8 weeks, as we first see what kind of traffic gravitates to us. – Gilles 'SO- stop being evil' Feb 11 '14 at 23:14

You must log in to answer this question.

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