This is a meta-thread for bringing together and re-organizing a selection of The Wikipedia Review's many splendored but many-splintered discussions of Governance Issues.

Jon Awbrey

Dynamic Page —

That means you can read it again & again & again & …

Critizendium
Cyc-Borg Idiology
Democracy And Inquiry
Group Identity Myths
Purpose Of Criticism

Unintended Consequences Of The Community Metaphor. Despite the most elegant and engaging title, this thread went strait to perdition in near-record time. I'll list it here anyway as a reminder to salvage the title for another try at its intended subject matter.

Here is the original thread in the Citizendium Forum:

Unintended Consequences Of The Community Metaphor

Jon Awbrey

Random Bump

Okay, Not Really All That Random …

Jon

You're playing Διόνυσος Λικνίτης, obviously, not Διόνυσος Ενορχές.

I'm more a fan of Odysseus here …

Well, at least it's not a big font rebus. I suppose that's progress.

Oh, I know my Odyssey well and caught the allusion. There's a whole google search on the phrase "winnowing oar" to refer to the episode.

The winnowing fan (liknon) was, by the way, sometimes made of cloth. There is a story that the young Zeus was laid in a liknon (perhaps a winnowing basket) while being hidden from his daddy, rather like Jesus being laid in a manger. Dionysus Liknites (The Winnower) is one of the aspects of that god.

Here's a translation —


I really like the word for winnowing-fan here — ἀθηρηλοιγός — glossed as “chaff-destroyer”

Chaff Winnow-Maker

Jon

I assumed it was liknon, but no, you are right. Here is a whole article about the word αθηρηλοιγον (athereloigon)—and the object. It is indeed more basket-like. The wheat and chaff are put in it and tossed, and the wind blows the chaff out across the lip in a process much like using a pan to separate silt from heavier gold when the fluid is water rather than air.

The basket-like nature of it looks more like a craddle than an oar. But there were winnowing implements that looked more shovel-like and served the same purpose to toss grain-chaff-mixes up into the air, and probably this is what is meant.

http://vunex.blogspot.com/2006/11/unknown-...ct-part-ii.html

Incidently, the word "athere" here used as wheat chaff is related to "athero," the Greek for a gruel or paste made of wheat, and is the root of atheroma and atherosclerosis, which to 19th cen German pathologists (Virchow, etc) looked like a lesion with a pasty center.

Rudder Hell …

Jon

Indeed. Cybernetic (κυβερνητικης) hell at flank speed.

Wild fights! Wild fights!
On wikipee,
And MMO
Our RPG

Futile the edits
What have we wrought?
Done with the knowledge
Done with the thought

Rowing in circles,
Ah! The sea!
Put out an oar
On WP....


(Apologies to Emily D.)

Now, the whole thing about Error-Controlled Course-Engineering (ECCE) is that you must be able to admit the possibility of error into your thinking, and become aware of it when it happens — but that is the very thing that Wiki-Pontifical Wiki-Pilots are proscribed by their pretensions from doing.

Now, I know what you're thinking, but I'll let you go ahead and say it.

It will be good exercise for at least one of us.

Jon

Well, I've said some of it before. Cybernetic principles indeed simply required a closed quality control feedback loop and a mechanism to maximize the quality. That sounds simple, but the devil problem, as you point out, is how to define "quality." What's the metric? In biology it's survival and successful reproduction, which is not the redundancy that anti-Darwinists claim, because it's built on a deeper bedrock of some kind of adaption to capture the low-entropy energy flows in the environment, do it without being destroyed, and then use it to make more copies of yourself that do the same. And in the Deming-style process-control in industry the feedback metric is simple long term profit from repeat business from satisfied customers. So you cannot cheat your way into it-- not over the long term.

In engineering and the sciences, it's predictability that is the metric. Does the theory predict the future? Does the design perform the task (which is essentially the same thing)?

In artistic endeavors, these questions of improvement become woolier, because you're essentially working on audience appeal. Performance artists of all kinds hone their routines before small audiences before they open them or Broadway or try them on primetime. Politicians, too. The Gettysburg address reads wonderfully, but it's not that great a speech, because Lincoln never practiced it before an audience before giving it the first time. Martin Luther King's "I have a dream" speech reads like crap but sounds great in front of an audience, because he'd actually honed it by giving it numerous times for small audiences, before that big assembly on the Mall in Washington for the civil rights bill rally. All this is necessary.

So what IS Wikipedia? Right now, it's basically performance art, and not much more. There are none of the quality metrics that serve in science and industry. There are glimerings like article ratings, but no agreement as to how the ratings should work, and (more importantly) no proposals for how to improve the quality of articles of low quality at the governence level, except by screaching and complaining, until somebody who knows how to do it, does it. This is about as childish as complaining about the dust on the carpets, until finally mom gets out the vacuum. No business would run that way.

Were I the WMF CEO, the first marching orders I'd give my board would be to come up with some plan whereby the quality of WP could be rated by some method that its intended AUDIENCE would generally agree on. THEN, I'd want some way of MEASURING that objectively. You might have to have to solicit reader-user feedback . THEN, you'd like some plan whereby various experiments could be done on various sections of WP (say, all articles that start with "A") whereby you could implement some program of editing to see if it was better than some OTHER plan that you were trying with all articles that start with "B". And so on. This is how every successful science, technology, and business work. That is how they all improve, to the extend that they do improve.

Take this out, and all you get is fighting and schisms, on the order of the religious wars of the last two millennia. And to some extent, the international and inter-state wars, in the era when no states were democracies. Democracies rarely go to war against other democracies-- the various feedback loops in democracy generally prevent people from doing something not in their long term interests. And you can run a business without democracy, but not successfully unless you have some objective feedback from some other thing (profit, the market, etc). If you insist on no democracy AND no objective quality control, you're stuck with no rudder at all, like nations of the world in their non-trade activities though human history. That hasn't been pretty. And as we've remarked, WP seems doomed to repeat it all, so long as it follows no improvement-model that has been successful on large scales, in the past.

MR

Were you the WMF CEO, you would still be assuming that the purpose of Wikepedia Incorpulated is make Wikipedia into a high quality encyclopedia.

Which is precisely why you are not the WMF CEO.

For Moulton, with wub and squalour —


Jon

Defining the objective function is not the hard part. The hard part is constructing an accurate system model that can be inverted to compute the control maneuver that will gracefully drive the system to the desired goal state.

Only if you're concerned with time. Otherwise, any old random change (so long as it's small) has about a 50% chance of being a change in the right direction (thanks Ronald Fisher). For large random changes, it's less! And if you have a metric for improvement, you can use even random changes and still tell when you should keep something. and when you should ditch something. And you'll get a gift in the right direction, half the time. And know it.

Only in control theory, where you sweat about making most of the difference toward a local optimum all at once, in one fell swoop, do you have to start worrying about correct theories of how things work. But usually, systems are so complicated that you can't tell how they work. The nice thing about having a metric, is you don't NEED to know how they work. Just do what you're doing as long as things are improving, and you'll get there.

The problem with Wikipedia, is that it has no idea whether it's getting better or worse-- only that it's getting larger. And it has no idea what experiments it's doing at any one time, so it has no idea what to deliberately keep and what to deliberately toss. It's too much like politics in that regard (although other language wikis serve as natural "experiments" hardly anybody pays attention to them, as witness the de.wiki's experience with flagged revisions).

It's like our health-care system. You'd think we'd look at one of the better models out there (say the Swiss one, going now with universal insurance for 15 years) and say: "Hey, they said all the same things our conservatives did, at first, but now even their conservatives now like it." They get as good care as we do, for 80 cents on the dollar. And copy them some. But no. We have to reinvent the %*&%ing wheel. That's Wikipedia.

Building a system from scratch to a given objective function is not the hard thing. The hard thing is identifying the objective function of system already rolling.

Well, it's the hard thing for some people.

Jon

Not too sure if Jon posts with all his fac – with all his f-a-c-u-l-t-i-e-s intact.

The Curse of Dimensionality


That's true if you have one degree of freedom (like steering only left or right). But if you have more than one degree of freedom (also up or down, also faster or slower, also pitch and yaw, etc), then random perturbations become problematic, especially if some of them are already optimized. Once a parameter is close to its optimum, a perturbation in any direction is detrimental.

But the larger problem of the Curse of Dimensionality is that the search space becomes huge and most of the explorations are fruitless at best and counterproductive at worst. In medicine, such a treatment is said to be iatrogenic if all that exploratory to-ing and fro-ing is making the patient sicker.


Evolution will eventually get there, but not in your lifetime, oh mortal physician. Most random mutations are fatal. Evolution succeeds because the population size is considerably greater than one. It doesn't matter if 99.9% of a generation die trying, as long as one of them calls "Bingo" and goes on living and reproducing his success.


A more insidious problem is that some treatment regimes (e.g. blocking, banning, and reverting) only seem to increase in popularity, without any consideration that they might be toxic practices which only serve to speed up the growth of the cancer.


The supreme irony of the project whose objective is to compile the sum of all human knowledge is that it systematically fails to consult that knowledge base when treating its own illness.

In the Game of Evolution, obliviously failing to notice what killed the species next door is not a very wise practice.