Previous Entry Share Next Entry
How pervasive should conversation be in Querki?
device
jducoeur wrote in querki_project
I'm still spending much of my time in the guts of the syntax, adding new functionality left and right. I just updated querki.net to the 0.3 release yesterday, and I'm starting to work up the wedding invitations themselves. But let's pause and go to a tangent that will become relevant in the next couple of months by asking a very important question: how often are folks going to *want* discussions inside Querki?

I haven't talked about this feature area much so far, but it's terribly important. One of the dozen ways in which Querki is rather radical is that it is intended to be a "social database". While it's going to be useful for little personal Spaces that nobody else cares about but you (eg, your shopping list), it is specifically intended to have a sweet spot for *collaborative* Spaces, where multiple people are working together to build and maintain the data.

So I've been pretty casually assuming that Querki would have comments -- well, pretty much everywhere. The idea is that any Thing can be the root of a comment thread. That's powerful and useful, and I know from my experiences with the ProWiki prototype that I am frequently going to want to use this.

But this recent post from siderea brought me up short. I strongly recommend that folks read it -- in particular, if you want to comment in this discussion, please read it before doing so, because it is crucial context.

I don't want folks to over-interpret her post -- it is *not* saying anything as simplistic as "Comments are Bad". What it *is* saying is that comments have subtle social effects -- and moreover, that the *default settings* for comments can strongly influence the cultural shape of an online service. That's a subtle and important observation, and I believe it's dead-on correct.

This has led to me pondering what the settings for comments in Querki should look like. The following is a strawman, handwave design. I'm specifically looking for discussion of it.


First of all, there is a top-level setting for the Space, between the three broad options:
  • Allow comments on all Things

  • Allow comments only on specified Things

  • No comments
The last two are similar in day-to-day practice -- the second is just like the third if you don't turn comments explicitly on anywhere -- but the last allows the Space's Owner to prevent any other contributors from turning on comments, which I suspect may sometimes be useful. (And the last should also suppress the UI option to allow comments.)

My gut instinct is that the second one should be the default -- that it should be easy to turn comments on case-by-case, but to make them opt-in. IMO, the most important observation here is that, if the default is "things have comments", it turns into a Statement (often unfortunately so) when you *don't* have comments. I don't want folks to have that sort of social pressure.

Note, BTW, that you will be able to turn comments on in a Model, and all of its Instances (yes, after all that argument I went with "Instances") will inherit that setting. And yes, individual Instances can override that setting to turn them back off again. Yay for a nice clean inheritance paradigm.

Closely related is the question of *who* can comment, and *how*. My current inclination is that, if comments are turned on, the default should be that "members" of the Space -- folks with edit privileges -- should be allowed to comment freely, and that anyone else who can read the Space gets screened. The other option is that everyone else can't comment at all. There is no setting that simply allows every Querki user to comment. (And there is absolutely, positively, no capability for anonymous comments, period.) Yes, I'm making a strong statement here: commenting in a Space is a privilege, *not* a right, and random people wandering into a Space *always* get screened.

Why? Partly it's simply to deal with trolls -- in general, allowing random people to comment kind of begs for trolls and spammers, and I want to very strongly discourage that. So my general theory is that, if you've been given edit/add privileges, that's a statement that you are part of the community. Otherwise, you have to be added explicitly -- either comment-by-comment, or with an explicit action by a moderator to whitelist you.

This is likely to be controversial, but I think it's correct. Querki is *not* intended as a light-touch, everybody-to-everybody melee as ordinary social networks are; indeed, it isn't intended to be a social network. Conversation inside Querki should, by and large, be *purposeful*, focused on the contents of the Space. To that end, I think we want to encourage a social standard that accepting someone into the community of a Space should be a deliberate act, and one that you need to earn.

(BTW, I am deliberately side-stepping the question of how comments themselves work, since that's going to evolve. Initially, I expect simple comment threads attached to the end of a page, as is typical. But in the long run, I expect to inject somewhat more Wave-like discussions, adding the ability to place comments *inside* objects, to help them focus better on the content.)


Opinions? While my gut says that the above is correct, I am *strongly* interested in folks' take on it, pro and/or con...

  • 1
I suspect that "Comments" are not the magic thing alone.

I believe that identity is bound up with the quality of comment use. Anonymous, pseudonymous and sock puppet identities have a lower cost associated with bad behavior.

Partly true, but I think you're making a common error.

I'll state my biases here: I am strongly in favor of pseudonyms, *provided* they are persistent and have real weight. On the flip side, I believe that anonymity is a complete blight on the Net, and should not, by and large, be encouraged.

I don't come to those conclusions lightly or quickly, and draw an extremely strong distinction between anonymous vs. 'nyms, largely based on persistence of *identity*. A name isn't important -- an identity is. An identity carries consequence. An identity matters as much as, and only as much as, the social capital that has been invested in it.

I regard Google as *staggeringly* stupid in the way they have treated this matter, because they put all the weight on the name, thus entirely missing the point. What matters isn't a real-world name -- what matters is the reputation associated with a persistent identity online.

There is *vast* evidence that displaying a real-world name does *not* guarantee good online behaviour, especially when interacting outside one's real-world crowd -- precisely because it does *not* have a high cost for bad behaviour. But real name or 'nym, if you've spent effort establishing an online identity within a community, then there *is* a high cost to bad behaviour.

Hence my approach (and on this I am standing firm): I want to encourage people to use consistent identities, precisely so that they can build up established online reputations. This implies that I want a non-zero cost to enter a given community, and social norms that nobody is automatically trusted -- you have to earn trust.

(Not to mention that trollishness is in the eye of the beholder. Different communities have different norms. That matters.)

In other words, I agree with the core of your point -- that the only way to reduce bad behaviour is to make it expensive, by making sure that the behaviour is associated with "you". But I disagree that wallet names are anything more than slightly relevant to the definition of "you" online. The only real cost to bad behaviour online is the way that it adheres to the identity that you are presenting to that community. Therefore, we want to encourage communities that consist primarily of well-established identities, and should build systems that encourage such structures.

(And yes, it is possible to fake this approach out with a sock puppet. But I see no practical way to truly prevent that, so I'll satisfy myself with focusing on making that more effort than it's usually worth...)

We are Silverwinging in violent agreement, by and large.

Consider LJ, where neither Goldsquare nor Jducoeur are our names. But there is a consistency of identity, and a social consequence that comes with that consistency of identity. I have several LJ "friends" who I have come to know over the years online, but I could not tell you what city they live in or what their real names are.

But the pseudonymous situation, as you point out, leads one to quickly develop sock puppets and/or special purpose names with no reputation or use other than perhaps abusive feedback.

I wonder if there is merit to actually implementing a "social currency" metric, so that new names have a zero score, established and largely positive names have a positive score, and some names have negative scores - without making the manipulation of scores a "gamification" of interaction.

On the Internet, no one has to know you are a dog. :-) But one should consistently be the same dog, within any individual context.

But the pseudonymous situation, as you point out, leads one to quickly develop sock puppets and/or special purpose names with no reputation or use other than perhaps abusive feedback.

Correct, and I think that's the heart of it -- that's driving some of my thinking. That's why I'm leaning towards what I might harshly call a "guilty until decided innocent" attitude towards commenting: I think it's appropriate to require the community to *consciously* admit you to full membership. (Where commenting is treated as a typical privilege of membership.)

I wonder if there is merit to actually implementing a "social currency" metric, so that new names have a zero score, established and largely positive names have a positive score, and some names have negative scores - without making the manipulation of scores a "gamification" of interaction.

Quite plausible, and I've been pondering the value of having something like that. But it's a lot of work to do well, and seems to usually turn pretty heavyweight and complex (see for example the systems on Slashdot and SO -- the latter, in particular, is almost brain-breakingly ornate). So I'm inclined to see what we can do with something lighter-weight and community-specific to begin with, and add something more Querki-wide later, if it seems useful.

(Part of why I'm inclined to put it off is that I have thought about this problem *way* too much, and concluded long ago that, to make it really hum, it doesn't want to be a simple number, but wants to be a metric calculated relative to the social network in question. That's an interesting graph-theory problem, but outside my scope right now.)

On the Internet, no one has to know you are a dog. :-) But one should consistently be the same dog, within any individual context.

Heh. Yes, exactly...

I wonder if there is merit to actually implementing a "social currency" metric ... without making the manipulation of scores a "gamification" of interaction

I suspect that this simply isn't 100% possible, no more than 100% site reliability is possible. People will attempt to game nearly any system with visible inputs or outputs: the question is to what degree, with what level of success, and what side-effects their attempts to do so engender.

(In this case, visible outputs are needed because the point would be to make a visible reputation score. Visible inputs are not strictly required, though they can serve well in actively incentivizing easy-to-measure desired behaviors... not all side-effects are bad. :)

A Comment thread could itself be a Thing that people can add to a Querki (page? app? Thing.) itself. So the thing creator can add commentability as a property to individual objects or to global scopes. If someone chooses to make a tumblr-esque Thing, they could leave comments off. But if they want Instagram or Flickr or any other Blog system, they can turn them on. Or they can turn them on for one part of a page (submit your captions for this picture!) but not for everything.

P.S., Twitter also doesn't have comments. It does have replies, retweets, and direct messages though. How do those models play in Querki? Do they?

And of course, a basic discussion forum could be just a comment box, but I suspect there are better ways to do that sort of function.

So, yes, Option 2 as the default puts the authorship of that feature back with the author, and makes them decide on every object, unless they decide a global scope of on or off, which is also a decision of authorship then, and not a default choice made by a third party. Tyranny of the default and all that.

I'd say more and more services are moving away from the wild anyone-can-comment model. As pointed out, Tumblr and Twitter do this by pushing everything back into the account of the person making the two way conversation. major blog platforms like Gawker Media and Ars Technica require you to register or link to a facebook/google/twitter account. and so are a lot of the smaller services too. But some places, like Livejournal, allow the creator to turn on anonymous commenting. Perhaps Querki could still keep the restrictions of someone needing to be a logged in Querki user (so it isn't truly wide open to spam bots), but allow the creator/owner to permit commenters to have their identities screened out?

Perhaps Querki could still keep the restrictions of someone needing to be a logged in Querki user (so it isn't truly wide open to spam bots), but allow the creator/owner to permit commenters to have their identities screened out?

Hmm -- are we using the same definition of "screened"? I mean it in the LJ sense: that your posts have to be moderated and approved before being publicly visible. (Heaven knows there are a lot of terms for that idea.) The point that I'm focused on today isn't about public visibility of identity (although that's also interesting food for thought), it's about moderation...

A Comment thread could itself be a Thing that people can add to a Querki

I ponder that point frequently. I confess, I'm actually disinclined to go quite that far down the rabbit hole, for a variety of reasons, ranging from the technical (I want to build some *very* specific and powerful UI for comments) to the economic (I don't think I want comments to count towards your Thing quota).

Otherwise, though -- yes, I think it's important that this be very configurable, so that you can adapt to the needs of the Space. The question that hadn't occurred to me until reading siderea's article was how relevant the *defaults* are.

P.S., Twitter also doesn't have comments. It does have replies, retweets, and direct messages though. How do those models play in Querki? Do they?

My gut is to say that it's a non-sequiteur -- that those concepts are peculiar to the Twitter *application*, and I'm not writing an application, I'm writing a platform. You might eventually be able to build something Twitter-like *on* Querki, and this suggests that you might want to turn comments off on that.

But truth to tell, I don't think it was nearly so deliberate in Twitter's case -- most of their functionality evolved from their semi-self-imposed limits, often as emergent effects within the community that then got reified by the service.

And of course, a basic discussion forum could be just a comment box, but I suspect there are better ways to do that sort of function.

Yep. But in fact, one of the first Apps is going to be blogging (moving this journal into Querki itself), precisely to act as a testbed for this and a few other features like RSS...

Seems like (as is usual in any discussion I get into) I'm thinking of this in far less theoretical terms than everyone else, in that my principal reaction to this post was "if you default to off, make it *really* obvious where the toggle is".

I do think it says something, though I'm not sure what, that my reaction is that strong only to the default being off; I don't have anything like that same intensity about the obviousness of the toggle if comments default to on.

But you're actually looking for theoreticals, judging by the tone of the rest of the comments, so... if you're looking at siderea's contexts, I imagine one could take the "what is this meant to be" and work backwards: you're talking a lot about communities and admitting members in this post, so I'm going to take it for granted that you want to encourage communities. To have a community, you need the comments so you can have a conversation. It's worth noting that she mentions *possible* to put comment areas on Tumblr. But in the plethora of links I get from people to essays on Tumblr, I don't think I've *ever* seen one with comments enabled. Creating the social expectation of comments off seems to lead to comments *staying* off and no detailed conversations happening at all.

Which is fine if that's what you're going for, but I think you do run the risk, by freeing people from the social expectation of having their comments on, that you create a social expectation that comments will be *off* and no one will join the conversation even in situations when comments are turned on, or people will feel social pressure against turning the comments on.

my principal reaction to this post was "if you default to off, make it *really* obvious where the toggle is".

Yes, that's a good point. My current handwave is that it'll be front-and-center on the Create a Space screen, right below things like Public/Friends-Only/Private and the name of the Space. But we'll have to pay attention and make sure that folks notice it, and that it's clear how to change your mind later.

To have a community, you need the comments so you can have a conversation.

Actually, that's not obvious. Remember that the point here is collaborating on data, so the data *itself* is really the focus of the community.

Mind, I suspect you're correct that comments will be relevant to most communities. But it isn't clear to me that that'll always be the case.

Creating the social expectation of comments off seems to lead to comments *staying* off and no detailed conversations happening at all.

Interesting point. I think we'll have to pay attention to that -- if you're correct, it may argue for a more nuanced position. (Eg, that comments are on by default, but only for members of the community.)

I don't know how easy or hard it is to turn comments on in Tumblr, which is relevant to the question -- I do think it's essential that it be very easy to turn them on for the major Models, so you don't have to think about it over and over again.

But yes -- we'll have to consider this an experiment, regardless of which initial position we choose, and see whether the results seem to match our intent.

Also, I need to keep chewing on the relationship between comments on/off and the access-control questions. It is possible that, eg, the right answer will often be to have comments on throughout the Space for members of the Space, but not even *visible* for people who are simply viewing the Space.

If the collaboration is in the data, and the comments are meta-data about the data and its state, why would it be wholesome to allow people to access the data without the meta-data?

You may want to take a fast look at how Google Docs is handling comments, and the options it offers when jointly editing text.

But if you just have data without any discussion of that data, how does that work for collaboration? For anything more interconnected than "I need this data from you, give me it in the survey form below," you need some way to at least say "this data is from XYZ and I collected it using Q method" and for someone else to respond "but Q method isn't useful to us, you should have used Z method." Or if you think the back-and-forth is irrelevant, some way for Person B in this example to *say*, when Person A posts data, "where did you get this data from and how did you collect it?" if Person A doesn't specify. (Yes, the fact that this was the first thing to leap to mind proves I'm an academic, but I'm having trouble conceptualizing a situation where shared data without any commenting abilities won't just lead to edit wars.)

"Creating the social expectation of comments off seems to lead to comments *staying* off and no detailed conversations happening at all."

This conversation has been focused around "default on" or "default off". There's another path: have comments neither on nor off by default. Force the Space creator to choose an option explicitly, just like any other required field.

  • 1
?

Log in

No account? Create an account