A rough outline of networked, similarity-of-taste based rating in the context of a deviantART-like site

Preface

This is a very long essay (8190 words), starting out from such philosophical concepts of what is art and what is taste, detailing (and I mean detail) a completely revolutionary approach to restructuring the concept of favourites in an art community such as dA, and finishing in a lot of technical details concerning this implementation.

All this was written with deviantART in mind; I intend to post it there, but I might publish it elsewhere as well. deviantART can be found at www.deviantART.com and is possibly the biggest art community on the world wide web right now.

More specifically, I come from literature. I am a prose writer, which makes me a minority within a minority. While it might be helpful to keep my perspective in mind to understand some of my statements, I believe that all I'm saying here can be applied to dA, or any dA-like community, globally.

Guide to Reading

There's a lot you may want to skip. I've tried to use logical sections in all of this; every logical section begins with a subhead in bold print, surrounded by linefeeds. If you find a certain section irrelevant to you, do skip to the next.

I'm making a few definitions in all this, and in the second half I postulate a few things as they might be in a Utopian version of dA. These definitions and postulations are underlined and surrounded by empty lines as well.

I've used bold and underlining in the text body to put emphasis on certain things. I've used underlining to stress certain concepts that are not definitions or postulations and bold to put conversational stress on certain words.

Essentially, if you want to get the gist of this, scroll down to the first occurrence of “UtopiART” (if you're using Firefox, press ctrl+f for find and type the word in the resulting bar) and read all the underlined bits that follow.

Enjoy :)

Introduction

There is no more subjective area judged by the human mind than art. And this is for good reason! None of us would want to live in a world of perfectly identical tastes. Everyone has their own taste, and we shall assume for the rest of this article that all worded liking or disliking of art is based on a real difference of appreciation (as opposed to an artificial, constructed difference of appreciation caused by a set of factors like peer pressure, the desire to appear highly individual, pre-conditioned patterns of what is and what is not good art, and so on). That is to say: Person A can read story I with tears in his eyes, on the edge of his chair, while Person B will read the very same story indifferently and give a shrug at the end of it.

Having established the existence of highly subjective degrees of appreciation for the same piece of art, we must ask ourselves the first important question:

Why rate art at all?

Do you know the Mona Lisa? Have you ever heard of William Shakespeare? Ever listened to a Beethoven symphony? Among the overwhelming amount of artworks created by mankind, there are some that seem to be very well appreciated by a great majority of people exposed to them. Consequently, these pieces of art have been rated highly by critics, and deservedly, they are now well known. Here is our first, tentative (and surely faulty! We will soon see why) description of that elusive expression “good art”:

Good Art is art that will consistently be liked by a clear majority of people

I hear you mock me even now: “What, so Britney Spears is art?” Slow down! I am attempting to build this argument logically.

Who likes what?

Every piece of art has an intended or an unintended target audience. In the case of modern pop music, the target audiences are known down to their last statistical detail before even a single aspect of the music to be made has been discussed. Emily Dickinson wrote for herself and a few select friends, mostly in letters, and never intended for her poems to be published at all; still, we could come up with a well defined target audience for her poetry as well. Keep this thought in mind for a moment as I reveal another part of the puzzle.

What is the purpose of art?

To entertain, in the widest sense of the word. Art may inspire, stimulate feelings that will yield more art, it may plunge its consumers into the pits of depression, or it might merely make its onlookers smile widely. As we established in the introduction, everyone has a distinctive set of tastes. The following definition seems logical:

Good Art For Me is art that will entertain me sufficiently

Who likes what, reprise: Who likes whom?

Finding friends is rarely a conscious act. We interact with those who life puts in our path, be they schoolmates or prison inmates, colleagues from work or co-drinkers from the local bar. We stick to those whose likes and dislikes are similar enough to our own that interaction with them pleases and entertains us regularly enough. To some degree, then, we will find, at least statistically over a large enough period of time, that we will tend to appreciate the same kind of art our good friends appreciate. Thus we are approaching a usable definition:

Good Art For Me is likely to be art that entertains many of my friends sufficiently

The mechanisms at work and their faults

There's a new film in the theatres. It sounds really odd, seemingly discussing pudding-eating habits of cowboys in black and white. You are torn; do you want to invest the money to find out whether or not you will appreciate it? Worse, will you invest one and a half hours potentially wasted as you wince through scene after scene of “Howdy Partner! Fancy some pudding?”

Optimally, one of your friends will have seen the film. Even better, many of them will have seen it. You ask them; they are loving the film, and you go ahead and have a cinematic enlightenment. However, especially in the case of black and white pudding and cowboy films, your friends will not have seen the film in question. Who do you turn to?

Art Critics And Their Inapplicability to a dA-like system

An Art Critic is a person who is paid to rate works of art based on a highly faulty premise, namely that there is such a thing as an objectively good piece of art. I support the idea of the existence of an objectively bad piece of art, but that is a can of worms I do not intend to open in the space of this essay. Professional art criticism is a self-sustaining system: Critic A says that Film II is good; critic A is a critic of some public renown. The herds of the general populace view said film with the predisposition that they will appreciate it. While a such predisposition will not determine the reception of a work of art, I suggest that it can at least influence it. Thus, Film II is well received. Critic A is proven right. Critic A gains more public renown.

There are controlling factors to this system in the real world which keep it from turning into a veritable perpetuum mobile. These factors include for instance the possibility to influence public opinion of a film even after the public has seen it. If I provide a true barrage of gaping plot holes, internal inconsistencies, and artlessly shot scenes, for instance, I may get some people to change their opinion both about the film and the critic who originally praised it. The reason why such factors can work in the real world is that there is enough public interest in these works of art and that there is a sufficiently small set of films at any time.

A first case study: :devsuture:

A dA implementation of the art critic principle can be found in the :devsuture: group. Suture's mission is to pro-actively search deviantART for literary deviations worthy of highlighting. They provide no criticism as such (for good reason; the workload is too high as it is already), but by highlighting certain deviations, they influence costumer behaviour. Like a good review by an art critic might make you want to go and buy a certain CD, a reader looking for good material will probably look at the things highlighted by suture first. The system works in so far as it produces a decent number of hits, typically. Judging from the comments a deviation receives after being featured in suture we can suppose that at least some readers found their way to something they ended up enjoying through suture.

Wait. Did I not say a such system was doomed to failure in the context of dA? Yes, and while failure might be a harsh word to apply to suture, who are doing a decent job, I think it is, after all, applicable. If suture's mission was to highlight a relevant portion of literary deviation to people who will like them, I must speak of failure. The amount of pieces I have found and enjoyed through suture is negligible when compared to the amount of enjoyable pieces I found through other venues. Typically, every suture issue would have one or two pieces in it that I could truly enjoy.

Are suture not doing a good job? They are doing a wonderful job: Each one of their editors, contributors and what-not points out, absolutely correctly, a piece of art he personally enjoyed. The failure lies in the lack of similarity between suture's collective likes and dislikes and those of the wider dA readership. This could not be otherwise: As soon as a target audience exceeds a certain size, it becomes impossible to reliably predict their likes and dislikes. Let me put it quite clearly: This is an inherent statistical limitation and no fault of suture's.

So art critics have a sort of master-key taste that fits everyone? Far from it. The difference here is this: The ratio of art critics to piece of art of public interest is so that there is redundant criticism for each piece of art; there even is distinguishing criticism: This part the critic likes, this aspect he dislikes. I can read a Van Helsing criticism that rips the film to shreds and say “wow, rotating crucifix crossbows and quasi-naked faintly east-European women? Must see!”

Suture does not have the time to go through every deviation and say “this is good about it, this is bad”. This, combined with the self-sustaining critic principle (something which suture is constantly accused of: that they feature only themselves and their close friends. I cannot entirely agree with this, but the fact that this complaint is repeatedly voiced means there must be some core of truth to it. Ironically, the assumption that suture will constantly feature their collective friends can be directly derived from the first part of this essay and is a central thought in the second part of it) and the natural gap of taste which devalues a simple “highlight the good stuff” system make suture an interesting and applaudable undertaking, but one which will never even begin to satisfy the appetite for good literature found in dA's collective readership.

Wait, why is Britney Spears not good art?

Oh, but she is. If you're roughly 13-19, preferably living in the United States, and not part of a distinctive subculture such as Gothic, Heavy Metal, or Country-Western, Miss (or should I say Misses? I should really keep up with my tabloids) Spears may be very good art for you. You will dance to her music, find your life's turmoil reflected in her lyrics, you may enjoy your first arousal looking at her poster; you are enjoying her. According to our earlier definition, she is good art for you.

We will not go into what this says about you to me. That is a completely different thing.

However, if you are you forty-nine years old, a professor of music, and a passionate violin player, it is highly unlikely that you or anyone in your social environment appreciates that sort of music. Establishing what is good art for the whole of mankind, thus attempting to come up with a universal attribute of “good art” is an enterprise as hopeless as it is irrelevant. Who cares?

Why do we need to rate art, reprise: Filtering the noise

If you want to find information on the effects of the Thirty Years' War on the German city of Zwickau, you would ask a German, find out that the German term for that war is “Der dreißigjährige Krieg”, and you would enter that term and the name “Zwickau” into google. I'm pretty sure you'll come up with a wonderfully written essay on the effects of that war on Zwickau, including an interesting tidbit about a splinter from Jesus's cross, which was supposedly removed from Zwickau by a famous general during that war.

(In this and the following google examples I've chosen to neglect google's pagerank system which surely simplifies the signal-to-noise problem in many cases. Bear with me; this is theoretical)

If you want to find information about a man from Africa you met last night whose name is “Free Ecards” (you never know with those Africans), I wish you good luck, but I predict you will not find much.  If this example is too constructed for you, imagine you want to look up stuff about your old high-school pal John Smith.

This effect, the inability to locate existing information due to an overwhelming presence of similar, unwanted information is generally referred to as “information pollution” in information retrieval. We are talking about a signal-to-noise ratio, and if that ratio is low, then there is so much noise (sites offering you free electronic post cards) that you will not find your signal (Free Ecards is the son of an expatriated Rwandan rebel).

If you wanted to inform yourself about the inventor of the google-endorsed “Orkut” system, a straightforward google search will still yield you your result; even though there is a lot of noise, there is also a strong signal.

If you do our aforementioned historical research, there might only be one page in the whole google cache that you will want to see, but there will also not be any relevant amount of noise.

Now imagine this: In the course of one day, ten literary deviations are posted on deviantART. Three of them will entertain you sufficiently to prompt an avalanche of happy emoticons from your side. Will these three be hard to find? Probably not. Investing a minimum of time, you will locate them and be happy.

Or imagine this: During one day, two thousand three hundred poems and pieces of prose find their way into the land where peas are grey. Two thousand of them will be absolutely fantabulous in your opinion; the other three hundred disposable. You will not have the time to read all the two thousand fantastic pieces, but if you wanted to look for enough good reading material to fill a long winter evening, this would be a simple enough task.

Clearly this is not the way things are. Why we are so far from this Utopia of a writers' community (or is it a Dystopia after all?) I shall explore in the next section. Before we get there, though:

We can objectively find that the signal-to-noise ratio in dA literary submissions is horribly low. Among a hundred submissions randomly pulled from the recent deviations section, I can expect to find at most one that will please me. This necessitates a system of filtering; without a filter of some sort, we will be lost, forever doomed to shout wisdoms into a riot of noisy children singing Britney Spears songs.

But, why? Or: Whence cometh this turmoil and noise?

Post-modernism has sought to abolish the idea of an “artistic” stratosphere of society hovering untouchably over the grime of peasantry and the general art-less populace. Everyone, they said, can make art. There is profound and correct truth in this statement; still I think the world as such would be a better place had that statement never been made publicly, because for every truth to spring from that statement, a hundred misconceptions and false justifications for abominations mislabelled as art crawled from this Pandora's Box.

It is true; everyone can make art. Education influences but does not determine the chances a person has to create art which will be good for a decently sized group of people. Even a girl who has never been educated in any area of language can come up with powerful poetry.

Or she can write this:

U stabed a knive thru my heart,
When u said u loved me not
U onyl wanted sex
Is their know god?

(this girl was probably German since she did not realize that not and god do not really rhyme; us Germans are truly alien to the concept of final voiced consonants)

While neither birth into a certain social stratum nor the suffering of a certain educative process are obligatory prerequisites for an artist anymore, this does not free us, the creative people of the world, from the need to put sufficient thought into our work. In short:

Everyone can make art. Not everything anyone makes is art

This should be a simple enough statement. Just because Don Parker, 16, can write a prose piece about how he hates his father does not mean any relevant number of people outside his closest circle of friends will enjoy any part of it. Proof? Cinemas still show “The Incredibles” or “Collateral”, and not “How much Don Parker, 16, hates his father”.

Evil forces are at work in the world, however, and these evil forces counteract the expected acceptance of this simple truth. I want to get the one point out of the way first of all which will brand me as a racist: The typical American Affirmative Mentality. Everyone is special! You can do anything! Believe in your dreams! Every opinion is valid! If you think Shakespeare's sonnets deal with the apartheid in South Africa in a mind-bending prophetic gesture, why, that is your interpretation and thereby absolutely valid!

No, it is not. And you are not special. You can, however, do quite a lot of things, if you invest enough work in it. This is especially valid for art. Still, because a big aspect of art is the expression of feelings, be they your own or your characters, the mere expression of feelings in any form is often enough to qualify a piece as art in the eyes of the unreflecting. Similarly, being in touch with your chaotic side, the opposite of your ratio, is essential to the process of creativity; therefore, things that look sufficiently random and vague are silently assumed to be good art as well. Let me stress this one more time:

The distinction between good and bad art is made in the mind of the audience

Another factor facilitating the misconception of the simplicity in the process of artistic creation is an aspect of human psychology: peer groups. Again, a lack of reflection leads a group of people to label something as outstanding art (though they may not even begin to understand it); the next random passer-by will see this greatly appreciated work and assume that he should like it as well. These are the factors I mentioned in my introduction. I hear you contradicting me: What does it matter, you say, why these people like the art as long as they like it? Who are you, you ask, to tell them what may motivate them in their choice of what they like? Why, I am the alpha and the omega.

Seriously, though: My main arguments are these: Lack of comparison and sustainable enjoyment. I dare-say that if a member of an angst poetry jerk-circle was to be properly introduced to classic poets writing about similar subject matters with so much more style and grace, they would look back at the things they worshipped before and scratch their heads. At least I hope, for mankind's sake, that this would be the case. Furthermore, I will make the bold statement that in ten years I will still love Kurt Vonnegut and E. L. Doctorow and Neil Gaiman; I will still be able to read :devdarkcrescendo:'s poem “Winter” and enjoy the way the words and rhymes and rhythms roll off my tongue. Show gothicboi89 vampyrmystress's “My Blood” in ten years and see what will happen.

Do I say these people are wrong to like what they like and that they should be reconditioned, for their own good? In a perfect world, maybe. However, my problem is not what they spend their time with. It's the pollution they cause that worries me.

Amplifying the noise, or a second case study: DTFs

So :devsuture: did not work because they do not have five hundred people who are each ready to devote two hours a day to sifting through mostly mediocre writing. Fine. We have even more than five hundred people who are more than ready to sift through all the writing dA will fling at them: All the people reading dA. Why not give them a simple, automated means of “highlighting” certain things they particularly enjoyed?

Omg, +fav.

There are a number of conceptual flaws in the system of favourites. I will try to highlight only a few of them:

Ambiguity of function: So what are they, the little stars? Your own seal of excellence? Bookmarks? Tamagotchi-esque buttons on your friends you can press to keep them happy?

Probably all of the above. However, when viewed from the outside, only their first function is relevant. Overlaps from the other functions (“oh, let's fav this so I can find it later” and “damn, Diane sure posted hot pics of her breasts the other day. I should fav her poem”) (I hope there's no Diane amongst my watchers) distort any sorting attempted based on these favs.

Circle-jerks. This is the thing about everyone on dA has been accused of being a member of: A circle of people who give each other quasi-sexual pleasure through praise and favourites. I never quite understood how that concept could satisfy a thinking person for any amount of time, but hey. For the sake of the argument let us assume such circles exist. Clearly, if the amount of favourites you will receive for a certain deviation no longer depends on the deviation's quality but rather on the size of your jerking circle, the favourites are meaningless.

Mood swings. Rumours has it there are female deviants on this site, thus we should probably address the issue of irrational and unpredictable mood swings. All good-natured sexism aside, we are all subjects of our own unfathomable moods. Remember the definition of good art? Well, some days, nothing will pleasure you, no matter how good it is. Other days, mere trifles will tickle your fancy and make you fav a lengthy and obscure piece of Bringa-prose. On a binary system such as the dA favs, where there only exists “yes” or “no”, such mood swings will cause disproportional inconsistencies, especially when we are looking at a low number of favs. The absolute absence of an objective way of judging a piece of art makes a set of choices which translate into “GREAT!” or “nothing” very meaningless. Plus, what is your threshold of star-dom? Do you apply mouse-pressure to that celestial icon when a piece merely made you smile mildly, or does it take a full blown MacBeth to take you to the stars? Evidently everyone has their own threshold; this would be fine in an analogue system where you could factor in everyone's opinion with a certain magnitude (oh, the ominous flash-forwards!). In a YES or NO system, these different thresholds only serve to make the system even more useless.

Finally, the perpetuum mobile of fame. There is a certain ridiculous mechanism at work here: Once a piece of art has breached a barrier of favs, the sheer overwhelming number of favs in that box underneath the preview will cause a positive predisposition in the viewer; more favs happen; more positive prejudice. Factoring in the all-deciding pageviews, which also do profit from an enormously successful deviation, we can see how this snake eats its own tail.

Another issue has come to my mind; this one is very literature related. The vast majority of registered users of dA are there for the visual art. That is alright; we will not bite them. However, imagine this: A photographer with four hundred feverishly worshipping watchers writes a poem. See where I am going? Now four hundred people with practically no previous exposure to poetry will look at this poem (and the pwetty, pwetty pweview) and punch the poor star. Next, someone lists poems of all times by favourites. Warp city!

In summary, a universal system of favourites to help filter the signal out from the noise cannot work because:

1) With an audience as large as a million registered users, tastes will vary so much that a relevant overlap of likes and dislikes with those of one viewer becomes unimaginable

2) Favourites are not given with the motivation of marking this piece as excellent for the general populace; at least, this motivation will be partly overridden by others

3) Favourites are like the Tribbles: Once you have a certain amount of them, you'll drown in them

4) Circle-jerks, if they exist, and the fav avalanches lapping over from other genres serve to further distort favourites

The principle of DTFs has failed even worse than the approach of manually filtering literary deviations. It has come to the point where jokes about the horrible quality of an average DTF will not be met with disagreement but rather a defeatist nodding of heads and general acceptance of this evil as if it was human mortality itself. It is a commonly known fact that the only group of people who will enjoy a DTF is precisely that group of people who has made it, for whatever reason, a DTF. Not only does this not help us filter the signal out; it amplifies the noise, as a hapless deviant browser who will sort things by favs will be even less likely to find something that will please him.

As it stands right now, the abolishment of the DTF system would have only good effects.

But this is not what we want, is it? Surely there is a better way?

Concerning the best state of a dA-like community and the new website of UtopiART
A Truly Golden Handbook

(note: This is a direct parody of Thomas More's full title to Utopia. I do not propose to found a new website. Not until I have my army of flying robot-monkeys ready)

Actually, if I wanted to mock Thomas More consistently, this should be book two. We have laid out so far what is wrong with the literary aspects of dA, specifically that we cannot find good pieces, and we have tried to explain where the problems lie (there is no reliable filtering system). How do we fix things?

What are friends for?

Recall what I said about friends: They are likely to share your likes and dislikes. Let us approach UtopiART one change at a time:

There is an option to sort deviations taking into account only your friends' favourites

If I went to the deviant pages of :devinebriate:, :devtearstone:, and :devsaintartaud:, and if I would see a certain deviation in all their recent favourites, I would most certainly want to see that deviation. Now imagine I could look at the prose submissions of the last months and sort them by favourites they received from my friends only. Such a sorting mechanism would almost certainly provide a much clearer idea of whether or not I will like something.

This is enough. You have to realize this: We could stop here. If this change was implemented, it would bring with it a true landslide of adjustments in mentality: Seeing the favourite system finally producing good results, people would use it a lot more consistently and actively. This would, of course, again produce better results. We would find more people who do write good things, make them our friends, and with every friend who shares our likes, our picture of dA's literary landscape would come into focus a little more. From my limited experience of programming, I would suppose this to be minimal coding effort. If you want to correct me in this, please do go ahead.

But we are dreamers, right? A company producing one of the dullest of products, CAD systems specifically customized for layouting complex printed circuit boards, has the following motto: “How Big Can You Dream?” (don't ask how I know these things; I just do). Well, we're poets and madmen, surely we can dream Really Big, no?

Who likes what how much, or: rating your friends!

Say my friend A is truly a Butterfly reader in the H. G. Wells meaning of the word: He likes pieces of 200 words or less and favs carelessly. Friend B is my closest friend, whom I have cloned from a piece of my nasal tissue. He likes exactly what I like (because I will destroy and remake him if he doesn't). Now Friend C is indiscriminate and a hippie: He likes everything and everyone. Even trees. Especially trees. From a list of thirty deviations, C faved all of them; A faved those I will be least likely to enjoy, and B finally, loyal henchman that he is, faved exactly those I will love. The group of A's favs and B's favs just so happen to be distinctive; no deviation is in both sets, but the sum of fav(A) and fav(B) is exactly fav(C), that is all of them. What do I get? A list of 30 deviations with 2 friend-favs each. Great.

Have you never felt the urge to rank and rate your friends? I know I have. I used to dedicate entire diary entries to friend top 10s and my reasoning behind the ranking. I was also thirteen at the time. Imagine this:

For every friend you can set a trust-factor on a scale from zero to ten

Zero means, right, this guy is whacko; I want him on my friend list for some reason, but I want to disregard his opinion concerning other people's art entirely. Ten means: I have cut this friend out of my nose. His opinion is worth ten times as much as the opinion of a single normal friend of mine.

Ten is a rather arbitrary number. I like it. Our numeral system happens to be based on it. A simple zero to three scale might work as well, but I would prefer to be given slightly more control here; besides, I cannot seem to shake off the idea of friend top ten lists.

WOW does not equal wow

I realize previous versions of dA had a more discriminate system of rating deviations. It can be assumed that it was somewhat like the current comment-rating system, from 1 - Bad to 5 - Very Good. Why this system was abolished is quite evident to me: Arbitrariness. Deviant A may give the rating five to one deviation every three months while deviant B rates everything he even remotely likes five. This is essentially the same problem we faced earlier with different faving thresholds and mood-swings: What may be a three one day might as well be a five the next day. It was that problem multiplied by five.

Now that I admit the existence of an inherent design flaw in this system, I demand its return. I do this merely because I think crass contradictions are cool.

No, not quite. Two things have changed:

1) We filter WHOSE arbitrariness we want to factor in

and

2) We have a corrective measure at our hands; if a certain deviant proves to be too random, we tune him down in our trusty trust-o-meter. His opinion will no longer have a big influence (or, if we so desire, no influence at all) on the listings we see.

The re-instatement of this system would also at least palliate another of our previous worries: Mood-swings. As I explained, overly moody (read: female) (okay, who else do I need to provoke and make my enemy? ;P) deviants can be muted out. However, we are all moody sometimes. But now a difference in one or two steps on the scale weighs a lot less than the difference between fav and no fav would have.

For reasons of consistence and Arabic counting systems I would once more advocate the usage of a scale from zero to ten. Zero should be the default for every deviation. One should mean “this was slightly enjoyable; it had at least one semi-interesting aspect”, and so on. The rationale behind the omission of negative ratings is simple: No one will bother rating down bad deviations because they are the rule much rather than the exception. We can see this principle in action when it comes to comments. The only person in the whole, wide deviantART world who consistently rates all comments is :devinziladun:, and if you look at the typographical intricacies he weaves into his poetry you can make an educated guess as to how much free time he has. Occasionally people will rate really damn good comments a five; I imagine this happening in a sort of drowsy state of after-shock.

I would like to add one more thing before I formulate this step as a state of affairs in UtopiART: The artist should only be notified of a general fav; not of the exact rating they received. That rating should remain a secret between the rater and dA's sorting algorithm. The reason should be obvious: Favs will become commonplace for at least decent deviations (and they should be that!), and since we are faving everything anyway, thoughts of circular masturbation will be prevented from creeping into the rating process. Why should I rate Diane's poem a ten if she will never know what rating I gave it?

Plus, this adds an element of mystery and should provide us with a great many conspiracy theories. And those are consistently the best fiction I read from dA's general populace.

Thus:

Instead of simply marking a deviation as a favourite, every deviant is given the option to rank it from 1 to 10. There are no negative rankings. The artist is informed of the fact that his deviation has been ranked only, not of the ranking. The ranking is used in sorting. The trust-factor is applied to the ranking each friend has given a deviation.

Are we there yet?

Almost. One last step and we are walking on UtopiARTan soil. Let us pause and reflect on what we have so far:

We have a discriminate rating system yielding in a friend-filtered ranking of every deviation your friends have rated, with the option of adjusting every friend's opinion by means of a trust-factor. Assuming that a good number of your friends will have read and ranked a certain piece, you can now be very certain that you will like it too. If you go to, say, browse prose -> general fiction, of all time, sort by favourites, you can be pretty sure that the top pieces in there will be some of the pieces you yourself liked best.

As an aside, and this is purely aesthetic, it might prove useful to add another option to not list pieces you yourself have ranked in a “sort by favourites” view. This must be an option only, of course; I imagine it could be useful when you are looking for new art.

This is wonderful, you say! Almost as good as pre-cooked boneless hens flying into your lazily opened mouth only to digest themselves in your stomach. Almost.

There are a thousand worlds out there

What of that wonderful writer who merely uploads his pieces and hardly ever interacts with the rest of dA? You would love his writing! If only you knew it existed.

But maybe there is a connection. Let's see. Your friend A has a friend, C, who does not happen to be your friend (you simply don't really know him). C finally is a friend of Z, that mysterious antisocial writer. C is very fond of Z's writing; he merely never mentioned it to neither you nor A, because it never came up. However, C has consistently given Z's pieces high ratings. That's all good and great for A, but assume for a moment A is an ignorant toward Z's unique genius and never rates any of his pieces. What do we do?

Spread the love like herpes

The gift that keeps giving. This is the most advanced feature of my proposed perfect state of art, and surely the one that lent this ugly essay its even uglier name. This is where the word “networked” finally comes in. I'll just get the final underlined feature out of the way:

There exists a system by which the opinions of the friends of friends factor in; indirect opinions weigh less. This system works recursively down to a hard or a soft limit

That sure looks ugly! Let's go through it.

In order to deal with ACZ-like situations, we should be given the option to consider the opinion of those our friends call friends. After all, my friend's friend is my enemy, no? Logically, not all of our friend's friends will be equally trustworthy. However, we don't know them; they would be our friends if we knew them. So we should ask someone who does know them about how trustworthy they are. It just so happens our friend, A in the above situation, has already ranked all his friends in his trust-o-meter. Why not use that? Now for the privacy zealots amongst you, we need not know, personally, the ranking A has given to C. The only party who needs to know this, and who knows it anyway, is the dA backend.

Can you see how this set of Russian figurines works out? Z is also C's friend, so we will want to consider Z's opinion on things as well! After all, he's a good writer, he will know good writing when he sees it. How much should we value Z's opinion? Well, let's ask C. He will know.

All the computer scientists amongst you just threw their hands up in panic. With, say, an average 100 friends per person, we would face an insane exponential growth of things. Not only does this sound like an incredible amount of computing effort, but doesn't this also break the filtering function of our system? Surely, soon enough you're connected to such an amount of people that the discrepancy in taste will void this new system once more. I'm certain it could be proven that any deviant could be connected to any other deviant (islands notwithstanding; as “about a boy” taught us, some men are islands) in a maximum of, say, six steps. Any Orkut user will know what I am talking about.

First of all, I will quickly address the computing load concern: surely there will be simple ways to avoid undue load. It would be perfectly sufficient, for instance, to generate a list of influencing deviants and their influential factor for every user once a week. Optimally this would be spread out over the week; it should be evident that doing all these calculations at the same time would be a Bad Thing [tm].

[Upon proof-reading, I decided to insert a paragraph on downward compatibility. People hate change, I know that for a fact. Convincing the high-ups in dA to implement this might not be the hardest part; that might be convincing the herds of deviants out there what is in their best interest. However, change may come gradually. The basic features of the current system are all there: You still have people giving or not giving favs to things. During a “beta phase”, the new system could be an optional setting, where every rating translates into a good old fav for those outside the system. This might allow for a creeping change.]

Next, we will definitely need a “degree of separation” factor. I will not go deeply into the mathematical aspects underlying this concept, for two reasons: I do not pretend to have the mathematical skills to think this through absolutely thoroughly and secondly I do not think that a system of this complexity which depends on so many unforeseeable factors can be planned accurately; this is a clear candidate for a beta phase with a lot of trial and error. However, here's a sketch of how it could work:

Say, for instance, that every friend's friend factors in at 0.09 * (your trust-factor for direct friend) * (your direct friend's trust-factor for the indirect friend).

Thus, if you trust your friend A with a factor of 5 and A trusts C with a factor of 9, C's ratings would factor in for you at a rate of 4.05 times what he rated the deviation.

This would mean that no indirect friend could ever influence a rating more than his connecting direct friend could; it also means a certain dampening of influence that comes with distance. (This factor is obviously in grave need of tweaking; 0.09 was just the first thing that came into my mind)

The ambiguity of paths should be resolved so that whatever path is most favourable is chosen. For instance, if A trusts C with a factor of 9 and you trust A with a factor of 5, but there's also B, whom you trust with a factor of 10, and B trusts C with a factor of 10 as well, C's effective trust-factor should be determined by following the path of greatest trust at any given node.

Margaret Thatcher naked on a cold winter morning. Have I lost you all now? Good, just what I had in mind.

I am mentioning and arguing the technical details of my proposed implementation as a proof of concept of sorts: I want to show that this is logically feasible. If you do not care for the technical details, you can merrily skip ahead to the next bold subhead. You may miss a few more naked politicians, but hey, you gotta take the bitter with the sweet.

The final point of interest is terminability. Obviously, measures against circular recursion should be taken. There is a quite simple solution to this: whenever we encounter a deviant on a certain level of recursion who was present in a higher level (for instance, when we encounter a direct friend as an indirect friend of the fifth level, or when we meet ourselves in a Final Fantasy Seven compatible scene of revelation), we discard the current thread of recursion. Boris Jelzin, passed out drunk in his underwear.

Still, the soft limit imposed on the actual recursion through a degree of separation factor lower than 0.1  (0.1 would mean that a line of 10s in the trust factor could result in a theoretically infinite recursion; theoretically because this would necessitate an infinite amount of deviant. A rate of 0.0999... and lower means that in an infinite recursion our effective degree of influence would tend toward zero) (do I need to prove this through complete induction? No? Thank-you) might not be sufficient. For instance, if the degree of separation factor is chosen too liberally (that is, too close to 0.1), the sheer amount of people whose opinion will finally influence a rating might prove to be too much pollution once more. One option would of course be to choose a radical DOSF (figure it out! ;P), such as 0.05, which would reduce a 10 friend's 10 friend's (say that one ten times really quickly) weight to a 5 already, and a consecutive 10 friend's opinion to a nearly-negligible 0.25. However, this would limit the usefulness of this system gravely, since only one or two levels of recursion would ever matter. Also, it might not take away enough of the potential load, since our super-exact computers would merrily keep calculating effective influences down to 0.000000000000001 or so. Thus, I propose a hard limit and Nelson Mandela in a bikini on a rocky British beach.

A hard limit would mean that after an arbitrary number of friends, our algorithm would simply cut off. A sane hard limit would be five, because I really like that number.

If you wanted to get really fancy, you could combine the soft and hard limit approach. (NB: even in a pure hard limit approach a DOSF will be needed to keep indirect friends from drowning out the voices of your direct friends. 0.1 will always be the absolute maximum possible for a DOSF) One could tie the DOSF to the current DOS. A linear rise in the DOS would yield an exponential decay of effective influences. George W. Bush, full front nude with nipple piercings.

I made a picture of it. Really. I mean, of the resulting curves. Not of good old W. Took me all of ten minutes. I like ten. I think I will use the picture as preview, to scare off even more people. Go look at the preview now and understand it!

With these simple proposals in mind, and employing either the hard limit or the combined model, we should get an astonishingly exact prediction of the relevance of any given deviation to the currently browsing deviant.

Arrival, at long last

That should be it. We're on UtopiARTan soil now. After a month of getting the general populace to use this system, results should show.

Next step: Twenty hours working week and true communism.

In one fell sweep, we have done away with the unfair system of an established, untouchable group of artists who will get recognition no matter what they do and created a system through which even completely obscure artists will soon find those people who will appreciate their art most looking at it. All they need to do is browse dA, find a few art-works to their liking, befriend these artists, and then post pictures of naked breasts.

No, I'm actually absolutely serious this time (not about the breasts though). If we implement this system, it will work. It will be nothing less than a revolution.

I have written this essay with the literary community in mind; I could delve into the reasons why the general problems of dA can be found more extremely in the smaller subparts of it, but I think I have bored you all with enough theory for today (or this month, depending on how long it took you to read this). I think it applies easily to all of dA, and the benefits might even be felt more strongly in the bigger communities. I imagine it's very hard to get recognition as a starting photographer in a field so amazingly equipped with really good artists. I mean this without irony; I have very little clue when it comes to photography, but from where I'm standing, I see a lot of incredibly good photographers in this community. I imagine this will make it very hard for a starting photographer to become known.

To be honest, I dare not hope for the full range of this to ever be implemented; too great would the change be and too irrelevant for dA is my voice. However, if I could gain enough support to make maybe even only step one come true, we would all be treated to a largely different dA experience.

I dare not hope, but I dare to dream. If you share this dream, it may become a vision.

Thank you for your interest and your patience,

and good night to you all.