Lambda the Ultimate

I think Raph Levien's great work on attack resistant trust metrics (PDF)—and associated code—is woefully underutilized. It might also be nice to implement some kind of slope-1 collaborative filtering system as LtU grows.

Paul Snively: slope-1 collaborative filtering system

I might need a collaborative filtering system some day, so I'd add to any discussion if we start here or elsewhere. But I'm biased toward offering first principle analyses over evaluating old published schemes. (I find folks can too quickly simplify during problem definition so solutions ignore whole classes of pertinent issues; not that I mean to criticize.)

In the last year in my day job I refined replacement algorithms grading relevance of parts of large data sets in a manner attempting to be fair to new data even when old data has advantage of tenure. New material must not be held up to as high a standard as old, or it can't break through an "old-boy" glass ceiling. That's one important thing to mind in algorithms. New data needs continuous aging behavior from loose to more stringent metrics.

Also, one dimension rankings oversimplify how people actually rank things. (Well, there's always a high-grid population subset seeking aristocratic total ordering of people & things in 1 dimension to simplify pecking order management and exposure control. But one needn't encourage elitism except by choice.)

Anyway, I think multi-dimensional ranking would be more useful and generate less nonsense (to non-mainstream thinkers) due to homogenization. Inter-dimension ranking effects could be calculated in some manner like to slope-1. I guess I'm thinking of some kind of Bayesian "consider-the-source" ranking of voters' attitudes towards topics to scale effect of their votes. For example (for some topic foo) a positive vote ranking a foo-centric post from someone who dislikes foo might count more than positive votes from foo fans.

But that approach might incite a semantic web fan flash mob swamping discussion with attempts to categorize in typologies, and we'd be reduced to librarians mediating vocabulary fights of newbies. (I almost said peasants. :-)

(Sorry for replying to myself.) I thought some folks might not have heard the term "grid" before. It comes from Mary Douglas in anthropology. I found a decent intro on http://en.wikipedia.org/wiki/Cultural_Theory_of_risk

Americans in general tend to be high grid, low group: 1) (high grid) believe everything is sensibly according to plan and everyone lives in social positions befitting what they deserve; 2) (low group) the influence of others, even in one's own family, is considered minor and often completely inconsequential.

Technologists tend to experience amplification of high grid perspective by self selection when society rewards such a career path (in many ways) and when creation of tools to order the world encourages a world view that's regimented and ordered. Resulting ethnocentrism makes one dimensional ranking look more reasonable.

[Edit: Sorry I'll try to keep my generalizations in check. In fact, I should try to keep my yap shut for a while, for practice. Thanks for all the good work, which is seldom praised but much appreciated.]

Americans in general

Are you trying to prove that we need a moderation system? Seriously, straying a bit off-topic is one thing, but let's keep the generalizations under control.

I'm not convinced this is way off-topic when it's still relevant to spam-removal and post-rating tools. Nor do I think it's inappropriate to talk about cultural issues affecting tool design, which will require talking about tendencies. None of this says anything about individual Americans in the specific.

I don't really see the connection of "Americans" to this issue. The LtU audience is both wider and much narrower than that, and I see many more possible downsides than upsides to bringing that kind of generalization into the discussion.

I think the point about "technologists" was closer to the mark. I'd agree that it's something like the "amplification of high grid perspective" that leads to people wanting to develop automated systems to moderate discussion. I have some concerns about such systems, though. (Perhaps because I'm not American ;-)

One concern is that an automated system may just become a proxy for discussions we'd have to have anyway. Two of the entries on the advogato.org home page at the moment are about the trust metric system. Is it better to be discussing a trust metric system than just the posts that are considered out of line? The latter discussions have a benefit in that they're more directly about what needs to be discussed in order to communicate and perpetuate the community's values. Adding indirection here could be actively harmful.

There's also the point that automated systems are easier, and thus more attractive, to game. In "Gaming the system: How moderation tools can backfire", which is an interesting piece, Derek M. Powazek wrote "if your goal is thoughtful, positive conversations, beware of adopting the qualities of a game."

With a trust metric system, one specific game is simply the gaining of a credential, such as "Master" on advogato. This could lead to people signing up and posting purely for the purpose of achieving the credential, increasing the volume of discussion but probably not helping quality. (Perhaps LtU, being learning-oriented, could address this by adopting "Ignoramus" as the highest rank, reflecting the fact that the more you learn, the more you realize you don't know.)

Moderation and/or trust metric systems do seem useful in some of the large communities that rely on them, so perhaps Paul is right that something like that will become necessary "as LtU grows". But despite the relatively large size of LtU's silent readership, I doubt that the active membership will soon grow to the levels where such a system becomes essential, because the site is so subject-specific.

On the subject of spam, I see that as a fairly orthogonal concern, which happens to overlap the above in some places. E.g. spammers create admin issues even when their comments are being blocked, just due to sheer volume of traffic -- we had a multi-hour outage a couple of Sundays ago related to this. We may do things such as adding a captcha-like filter to the registration page, but I don't think we're in danger of needing a moderation system just to deal with spammers.

Or perhaps if we just got a good spamm filter, the moderators might get enough time to do more higlevel moderating, oblivionating the need for a automated moderating system.

Technical efforts against spam are ongoing. As I mentioned, it's not just about filtering the content -- just the traffic volume alone, which has more than doubled in the past year, creates issues.

But I think it's worth repeating what Ehud wrote in the topic: "the conclusion we reached - that is that the community should "police" itself - is the right one. If you find the content or style objectionable, post about it (in a separate thread, if needed)."

In other words, everyone's a moderator, so in theory, we should have no shortage of moderation resources.

Perhaps if we just solve the halting problem, other software engineering problems will follow!

It's an arms race and it always requires effort to deal with.

I am an American (and actually from the brief descriptions of "high grid, low group" it also seems to not be that bad a fit; I am a capitalist,) but I'm also one of the more vocal advocates of implicit community moderation (again, I am a capitalist.) The "Gaming the system" link is quite good. I guess, however, that I'm not a particularly good representative, as I would not use a filtering system anyway. Actually, to be honest, if reading LtU required or even highly "benefited" from using a filtering system, I would cease reading it.

Along the lines of your "Everyone's a moderator" post, you do seem correct in that established community members (including myself) are not doing much to emphasize community standards. One case, not so much a problem but a trend, is the growing amount of Forum Topics that are not based on anything, just random thoughts, wonderings, questions, ideas. Part of this is, I believe, a sense of not wanting to be "unwelcoming" to new members. I guess I should state here that I seem to have a bit more extreme attitude toward new users/being welcoming than most people here, so take what I say as you will in that light. Essentially, I care about quality more than quantity. Even in programming language communities, where just having more users has benefits, I am still more interested in quality. In the case of LtU, it is far from clear that more is better. As such, I have no compunction about having the bar to entry be higher, e.g. things like waiting periods and also just a less accommodating attitude. I don't come to LtU to read unstructured grab-bag conversations whose goal is to communicate well understood ideas to a single person. While LtU is a good resource for learning via old conversations (and the wiki when it comes), I don't think new conversations should be a vehicle for explaining well established ideas.

Anyway, as I suggested above, it's probably a good thing that the majority of the LtU community (and the Haskell community) are much friendlier than I am.

For an alleged rant, that was remarkably useful. Thanks. I agree with many of your points, and I know there are quite a few others who feel similarly.

However, there are some good arguments in favor of giving some leeway to newer members - the main one, I think, is to help keep the site and its community healthy over time, i.e. too high a barrier to entry will just keep new members away, presumably leading to stagnation. This is closely related to the point McCusker raised about the old-boy glass ceiling. (It's the same sorts of reasons that lead universities to put up with students...)

However, this requires striking a balance, which is made all the more difficult by the fact that it's not done in a very systematic way. I do think we need to work on this, although it's more of a medium- to long-term goal from my perspective right now.

...not so much a problem but a trend, is the growing amount of Forum Topics that are not based on anything, just random thoughts, wonderings, questions, ideas.

I've been guilty of this is in the past; but the nature of my offense wasn't clear to me until I read this post. Good guidance on a thread gone bad is a rare thing -- too often, the thread is tolerated until it's really rotten.

However, it's definitely unattractive to see a thread wander from technical to sociological to completely meta, disciplinary stuff. A mechanism for petitioning a user to change their behaviour could help alot -- it would keep discipline out of the forums while giving it adequate space.

If a top level Forum Topic post does not have a link in it, reject it. I think this heuristic would almost perfectly discriminate between "good" and "bad" topics.

Did the early posts about Cat have links?

Most of them seem to. Certainly policy 4.a seems relevant.

Rys McCusker: Americans in general tend to be high grid, low group: 1) (high grid) believe everything is sensibly according to plan and everyone lives in social positions befitting what they deserve; 2) (low group) the influence of others, even in one's own family, is considered minor and often completely inconsequential.

You say that like it's a bad thing.

:-)

McCusker wrote:

[Edit: Sorry I'll try to keep my generalizations in check. In fact, I should try to keep my yap shut for a while, for practice.

The latter definitely wasn't what I was aiming for. Your comments on LtU are always interesting and thought-provoking. I was perhaps being overly preemptive in attempting to head off a discussion about national identity.

Thanks for all the good work, which is seldom praised but much appreciated.]

Thanks very much. LtU is one project where I don't think I've ever questioned whether the effort is worthwhile.

Don't take the harping personally. The culture of non-anonymity dates back to the early days of LtU. It's understood that not everyone will conform to that, for various reasons, but nevertheless new members are encouraged to use their real names. Exhortations about that aren't really directed at established members.

I'll second the real name issue. I think it's important to use a name that is identifying/unique at least. For instance, I have used 'naasking' since I first started posting online, and a google search on that string yields my real name on the first page of results, so I'm not trying to hide my identity; more of my online history is under my pseudonym than my real name, so isn't it more useful? :-)

It could be argued that in a significant sense 'naasking' is your real name online, and it's been said before that such cases're fine.

We may very well try some version of this after our next Drupal upgrade, in the form of the Flag Content module.

We had a similar experience on schemecookbook.org. We'll probably be doing something like that here. LtU does get many more human spammers, however, presumably because it has much better googlejuice.

LtU is more challenging to think up a relevant but usable question for, though. It's not about "programming"! Perhaps something like this: "Please reduce the following lambda calculus expression: ((λx. x) ltu)"

"Please reduce the following lambda calculus expression: ((λx. x) ltu)"

Lovely. But soon everyone will visit this site for nothing but solving Captchas ;)

That seems to accomplish three things at the same time:

1) It would eliminate any auto-generated spam and most other spam. At worst you'd have people pushing their products who actually are in the community

2) It would establish a low but reasonable barrier for entry.

3) It would work towards establishing what topics are considered absolutely vital for entering the conversation.

Since the other major topic is types, if I could suggest another one:
(toRational (3::Int)) + (toRational (4::FLOAT)) =
(answer 7%1)

You are right. We have been talking about it, and some changes will hopefully be implemented soon that will help with the situation. If the technical measures will not be enough, I will certainly raise the option of changing our policies.

Let me use this opportunity to note that several members have offered to help with spam control and their help greatly reduces the amount of spam you see (imagine how much worse things could have been!) They should be thanked by all. I have also been more proactive recently, and many spam accounts are deleted or put on probation. As you say, all this has not been enough.

Well, when I remove spam, by marking an account as a spammer and going through the known spam page, stories go back to their original order in the recent posts page.

But at least one other admin seems to be using a different code path, which seems to have slightly different effects along the way.

One thing that happens is that if a user replies to a spam message, that non-spam reply will affect thread order even after the spam has been deleted. Deleting the non-spam message manually doesn't fix this directly, but the next time the spam deletion script runs, it fixes the thread order.

Improving some of this is high priority right now, and there'll be changes in the next few days.