MVP Discussion: Reputation

Let’s talk about “reputation”, or in a more general form, “points”.

  • Do you think we should have a point system?

  • How would these points be awarded?

  • What is the incentive to earn these points?

  • What rewards/privileges do points give you?

  • How do we prevent the system from becoming overly elitist?

Feel free to cover any of these points or any others you want to discuss.

1 Like

The reputation points are very useful in motivating users and they help to separate good and bad questions/answers.

  • However the point system is not always working so well in order to separate good and bad questions. There are many false positives and false negatives. (e.g. older answers tend to have some advantage and this may already work for an answer that got posted a few hours earlier)

  • The motivation (that the reputation-points give) may have negative side effects and may become a perverse incentive to score points.

I do not believe that the reputation point system is a necessity and it might be a nice change in the current narcissistic driven internet to have again a website without likes. Wikipedia seems to be doing very well without it.

In addition without reputation point system you have less technical issues associated with it.

Also it becomes easier to copy content across different websites. Say one wishes to use codidact with a copy of thousands of questions and answers from SE/SO, but the votes can not be copied, then one has a problem with a large flux of un-voted questions and answers (which will not be a problem if the community of the site is organized to not care about votes).

2 Likes

The flip side is that having votes mean that:

  • “Better” answers, whether old or new, float to the top. (Yes, there could be some adjustment to the simple model of this as used by SE to allow for “newer” = “better” on a scaled basis and/or moderator override, etc. But in principle this does work.)
  • Tying votes to a reward (i.e., reputation) gives people the incentive to put more work into producing better answers. Without the reputation bump, voting means something to the reader but not to the writer. If it means nothing to the writer, no reason (except pure altruism) to write a better answer, which means that many people just won’t bother, which means less/poorer content.

I am open to alternatives, but I see none. The specifics - how many points, what rewards, etc. - can certainly be different. But the basic premise is fundamental to the site unless someone can come up with a truly different idea.

5 Likes

I added some points on a possible point system here What are we trying to build?

Just got an email from SE that they are changing Question votes from 5 to 10. Despite https://meta.stackexchange.com/questions/337843/should-the-weight-of-question-upvotes-be-increased-network-wide which clearly showed a consensus in the community against it.

We must not Deliberately override a clear consensus EVER (except, of course if there is some true emergency “fix the system” problem, but I mean policies, reputation values, etc.)

And for reputation specifically, the reasons as listed in the link (generally) hold for me too. Basically there are a LOT of reasons to value Answer votes higher than Question votes.

4 Likes

There’s obviously exceptions where there’s nothing else we can do - I’m talking emergency situations - but yes, I agree. This is a community site and needs to stay that way.

Sort of relevant: I made a post on MSE and I’d appreciate if you all read it.

1 Like

Although I’m not deeply involved in the details that are already planned, some thoughts:

Do you think we should have a point system?

There will likely be some voting mechanism for answers. The votes will reflect some sort of “quality” (or maybe “usefulness”). And when the answer is connected to a specific user (and not posted as something like a “Community Wiki”), then there will be a point system - even though it may be implicit: There will be users who have 10 answers that each have been upvoted by 100 people, and there will be users who have 5 answers that each have been upvoted by 2 people…

How would these points be awarded?

That’s what it boils down to: Should upvoted answers be translated into an otherwise meaningless number that could then be called “reputation”? I think it has advantages: It allows e.g. to incur a small cost for downvoting answers, or implement the concept of “bounties” (which I think can be pretty useful).

What is the incentive to earn these points?

This is in part answered by your next question (“What rewards/privileges do points give you?”). The approach of outsourcing moderation work to users who have proven (via valuable contributions) that they want to make the site better has worked astonishingly well on SE for 10 years. Sure, there are corner cases and issues. But the general idea of considering reputation as a rough measure of trustworthiness isn’t that bad.

The second incentive is related to things like competition and visibility. Many people are (or have been) proud of their contributions. Of course, considering the low value of these internet points, contributors still have to be altruistic. But maybe that’s the point where “reputation” turns into a “measure of altruism”…

How do we prevent the system from becoming overly elitist?

This question is not so clear (for me). What exactly does “elitist” mean? And more importantly: Is it bad?

5 Likes

This is a difficult question. If you think there’s an obvious answer, you don’t understand how human beings work — specifically, you underestimate how different people work differently.

Reputation (or karma or EXP or whatever you call it) has positive and negative aspects. It’s a motivator for some people, because they like having bigger numbers next to their name. But that’s a motivation to produce content that gets upvotes, with an incentive to produce popular content in preference to useful content. There’s a reason why the term “rep whore” is (or was) common on Stack Exchange meta sites: people who seek reputation for reputation’s sake often do not make the site better.

Reputation, the way it’s done on Stack Exchange, is a fairly good measure of engagement. I’m sure most of you can name a user or three on your community(ies) whose engagement clearly favors quantity over quality, while others are unsung heros who have helped more people but didn’t get a big number from it. And even as a measure of engagement, it’s only mildly accurate if it’s “fair”, which requires preventing plagiarism and sockpuppetry.

[citation needed]. Reputation points are useful in motivating some users, I’ll grant you that. But.are these the users we want? And how does reputation (as opposed to voting) help distinguish good posts?

No. It gives some people the incentive to produce more upvoted answers. There’s a huge gap between this observation and concluding that it leads to more useful content overall.

Sure. I’m proud of my contributions, too. But that comes from the contributions themselves, not from the number next to my name. I do realize some people work differently. But are they the people we should primarily cater to?


If we build a point system, it’ll be difficult to go back on that, and to make it evolve. Stack Overflow has had trouble with that: there’s always an outcry when the reputation mechanisms change. There’s always an outcry when reputation goes down — SE ended up implementing a compromise system where deleted posts sometimes give you reputation, because although deleted posts don’t help anyone, there was too much whining about losing points.

Given that it’s easier to add a prominent number than to remove it, I think we should start without any kind of reputation-like number. Then we can experiment with some unofficial numbers — maybe calculated through a separate engine, and shown on a third-party site. If at some point we think that one or more of these numbers is useful, we can add it to the official system.

7 Likes

Gilles, I guess that I completely agree with you, only you were able to describe it much more eloquently.

My post was supposed to be much more nuanced than ‘reputation points = a good thing’ and I actually even end up with the (opposite) conclusion that I find reputation points unnecessary.

Reputation points has a good side as it motivates (some) people and allows to differentiate (some) posts (yes this should have been more nuanced by stating ‘some’), but… it is a bit simplistic and very one-sided (it is based on popularity only).

My viewpoint is that this popularity voting thing starts to become a bit old-fashioned and is a thing of the last 10’s in this century but not something for the next decade

(The voting has been mainly used on commercial websites - copied, taken over, on other smaller sites - and is used to get as many people hooked up, addicted, as possible - creating more income. But now we live in a world where everyone is living their lives on internet. We are moving away from those commercialised spaces. We want safe spaces where we converse normally as we do in real live. In real life I may give compliments to other people, but I do not press like buttons continuously or search for people/conversations in lists with high scores.).

I personally want to go back to experts that I trust rather than some massive community generated popularity number that has been created with several sources of underlying randomness and several sources of bias (like how long the post has been on the main page, and how many times it got bumped, is a large influence).


There are several communities that operate without ‘points’ and without a ‘score-board’. Some have become very large and important, like Wikipedia, and some may have some alternative system of endowment like ‘Linked-in’. We should be careful in introducing the voting-system, just because SE/SO does it. Any voting system will fixate how the ‘community’ values the contributions and contributors. But this is unnecessary. Many communities that are without an explicit voting system will still have some sort of way in validating the contributions and contributors. You feel this whenever you step into an online community, you notice directly that there are some players that are widely regarded as the alpha monkeys and act as the high rep contributors. It happens naturally (but when you let it happen naturally it is much more what the community develops and less based on what ways of validation the system/software has created).


A possible way of endowment could be the number of ‘followers’ for posts/tags/contributors (more like twitter). This still will relate to ‘popularity’ but in a milder form (not every contribution is being scored).

1 Like

I am not sure about a reputation system in general, however I think we need a way to grant privileges over time.

I think Discourse’s trust level system is quite nice for that. What is your opinion?

2 Likes

Raising an eyebrow when reading about “safe spaces”, but…

LinkedIn has a different goal, and consequently, different incentives: The goal is, roughly speaking, to be a network of professionals, and thus, the “reputation” is measured by the number of contacts. Wikipedia may have more similarities at the first glance. But its nature as a “collaborative encyclopedia”, where one page may have dozens of editors, makes it far more difficult to attribute contributions to individual persons. On a Q/A site, there is a (nearly) 1:1-correspondence between a user and the question, and one user per answer. Rewording a paragraph of a Wikipedia article may take a few minutes. Writing a good answer can take hours. People want to say: “That’s my answer”. (Free for everybody to use, but it’s still mine).

I agree that it may appear to be “shallow” to care for internet-points. And it’s tempting to say “I’m here only to (altruistically) help others”. But there is a reason of why these persons still do not turn each and every answer into a “Community Wiki” answer, and not publishing each and every code snippet with CC0/PublicDomain license: Attribution and confirmation are important.

Imagine something like Instagram, where people are posting crappy holiday photos. Friends are “liking” these photos. Fine. These people could also send out the photos via mail, and the firends could respond with “Thanks, I liked your photos”. Why doesn’t this work? I think there’s one aspect that is crucial for a contribution/confirmation-driven community to work: The confirmation has to be visible for others!. Receiving a “Like!” is worthless, unless everybody can see how many “Likes” one received.

(I’m not an expert at the details here, and one could go very far with reading and critically evaluating dozens of medical studies. So correct me if (you not only think, but know that) I’m wrong, but as far as I know, the connection between “Likes” (aka upvotes), dopamine release, and the resulting symptoms of addiction is pretty clear. Yeah, addiction is bad, okay, but on a Q/A site, it at least fosters valuable and useful contributions (i.e. something beyond crappy holiday photos))

However, I’m also aware of some of the problems that directly connecting “Upvotes” to “Reputation” may cause. It’s frustrating to see certain Q/As being upvoted to the moon, just because they appeared in the HNQ. The self-reinforcement that is caused by the Q/As to appear in the “Questions with extreme votes”-section of the mod tools also contributes to that.

So I agree with luap42: Compared to “rep from upvotes”, the discourse trust level system is much more sensible as a measure for trustworthiness. It has many parameters and degrees of freedom, but things like the “time spent reading”, “number of days visited”, or “number of posts read” certainly are clearer indicators of whether someone cares for the site.

3 Likes

I certainly agree about the positive effects of the voting and reputation system. I do not want to argue with that. (I argued that one may find side effects unwelcome and the positive effects unnecessary, but I did not want to argue that the positive effects don’t exist)


Why do you believe there is (should be) one single user per answer on a Q&A site?

1 Like

Some very quick thoughts/replies:

  • A quick look at the Discourse level system (from @luap42 ) - it looks to me much like SE-style reputation except that you only get the badges (= privileges) and not the reputation points. So the result is sort of the same, except that you don’t get the constant (for better or worse) feedback as you climb up.
  • I suspect that a lot of high rep SE users started out by getting hooked on gaining rep, learning the system in the process, until they graduated beyond that and continued contributing anyway (and/or contribute far more than their daily 200 rep cap).
  • I really do think that rep in general - though not perfect - and voting - though not perfect - is a good way to keep people coming back to the site. Which is key to getting people to make it one of their main sites rather than a place they just go to if Google sends them there. Without rep or some substitute, my hunch is that instead of 3 groups (one-time visitors, occasional/moderate visitors (say rep. between 200 and 2,000 and gradually climbing over time) and heavy users (high rep, on the site every day, moderating (whether as actual moderators or casting flags & VTC as high rep users, etc.), we lose that middle group - i.e., except for a few people with a passion for the site topic (which is another issue - as mentioned before, main SO is just too big), you won’t get people to climb up, and without those people you don’t get enough users to make the site work. Not an absolute, but that’s my general concept.
4 Likes

I thinkt it depends. From a human point of view: Sure, why not.

However from a technical point of view, there are some concerns IMO. This is just what I thought of right now:

  • How should ownership be stored in the system with performance concerns in mind? This looks easy at first, however as soon as we want to distinguish “major authors” from “minor authors” it gets complex.
  • How should reputation (or whatever) be stored? How should it be split among the contributors?
  • Who should be able to vote on the answer? Nobody who edited it?

I am not sure how exactly you mean this.

In the end any reputation/privilege-like system will be similiar too SE as it always is about granting more moderation privileges to the active, trustworthy users. The distinction is how the trust level is calculated.

Stack Exchange uses the model that maps specific actions from other users to points. Actions indicating that your post is good result in a gain of points, actions indicating that your post is bad result in a loss of points. In the end there are certain boundaries that need to be crossed in order to earn a privilege.

Discourse, on the other hand, uses a looks at your overall commitment. Users that spend some minimum time interacting with the site are granted more functions. Higher levels can only be achieved by maintaining a positive interaction rate.

Both systems have advantages and disadvantages, of course. I’d favor Stack Exchange’s model, however if we decided not to implement a reputation-score-like system, I’d think the Discourse model were a good choice.

1 Like

I guess that I have a (relatively static) system in mind that looks more like wikipedia or other wiki’s (in which case those concerns are much less of a problem). In my view the StackExchange system of community building by making many questions/answers (ever growing in number) and gathering votes is dying. (the source with the nice questions is drying out; almost like a pyramid scheme)

What sort of life is still going on at StackExchange?

  • Do we see a cute little energetic bambi jumping around
  • Or do we see vultures picking the old rotting meat from bambi’s carcass? Soon it will become a skeleton that is good to be framed in a museum but is not working well to keep the reputation-minded bloodthirsty community alive.

I couldn’t disagree more. Wikipedia is a fantastic system. I use it frequently. I contribute occasionally. But it is a reference work. It is alive unlike a printed encyclopedia. But the concept is the same. The people who contribute a lot tend to do so based on their particular interest area, and do so by starting and/or updating articles. Which have no relation (except for current news events) to what people are actually looking for.

On the other hand, a Q&A site (SE/SO, Quora, ExpertsExchange, etc. - i.e., no matter the corporate structure or other details) is about I have a question + I can answer the question. It is immediate. Both in function (“I’m traveling to xyz country and not sure if my documents will work, help!” which needs an answer “right away”) and feedback (“votes” = “points” = “new privileges”, all very quickly if you put in the effort).

Can the system be manipulated? Yes. A good structure (adjusted over time with community agreement, unlike some others) and good moderation (algorithmic like Charcoal + real people who care) can make it work.

3 Likes

I think:

  1. this depends on the Stack Exchange site.
  2. it has been consensus from the beginning that we should not build a site vastly different from Stack Exchange (except for management) right now and that the friction of moving from SE to Codidact should be minimal.
  3. that our main consideration points are (a) how to make Stack’s management issues very unlikely and (b) what is Stack Exchange failing at.
1 Like

I believe there usually is. I didn’t say that there should be. There are Community Wiki answers, or answers that are heavily edited (e.g. due to now software versions, with the caveat that the original author may not even be active on the site any more). And that’s fine.

(However, now I say it: I’d personally prefer to generally have a clear “ownership” of answers, but that’s just one opinion…)

Some of the issues raised by luap42 in view of a more “collaborative” structure of answers perfectly map to the “Documentation” approaches that SO started a while ago. Dozens of people made (not always perfectly consistent) edits to things like the documentation of “How to create arrays in Java”, and one could immediately observe the difficulties of distributing reputation (i.e. quantifying the contribution) - and the “Too many cooks spoil the broth”-effect…

3 Likes

A concept I came up with: user experience is made up of three aspects. Kudos, privilege and badges.

Users can give and receive Kudos on an answer, question, comment or useful comment thread. Kudos is a signal of how active and beneficial that member is to the community - kudos could be used similarly to upvotes where a new user can be given small amounts of kudos by going through a tutorial and learning the different aspects of the community but is distinct to privilege. Kudos mostly matters to newer members. It provides a way to gate basic interaction until a user has completed the tutorial or basic easy to complete interaction in the community. Gamification is an important part of kudos and at certain levels could unlock perks (this is optional). A high kudos does not denote a high privilege.

Privilege is a separate aspect that is given to a member of the community by the community via votes. Unlike SE, different levels of moderator rights are not earned through Kudos but are voted on by the community. Votes would be held regularly and would be at varying levels. Privilege can be seen as moderator privilege. Privilege is earned and given to users who will not abuse their moderator rights and is given at varying levels. For example a low-level privilege could be to have the power to open up or accept comment threads on definitive questions. A higher level privilege could be to close questions or move questions into different areas etc.

Badges are given when a user receives a respectable amount of kudos on a given topic. Badges could also be represented in levels of accomplishment. Badges allow a user to be seen as a domain expert on a given topic. Badges allow a user a certain amount of privilege-like interactions but only on questions/answers within their domain knowledge. Most of these badge privileges will be specific to understanding the subject. Badges provide an aspect of gamification but with actual value.
[/quote]

1 Like