Friday, October 19, 2018

ARS (beta)

Eric Kim seems to be the driving force behind this new thing that he wants to see change the way we give and receive critique on photos, which I guess is named ARS but is in the "beta" stage of development? Anyways, it's called ARS Beta and it has "moonshot" dreams to "disrupt the photo industry and social media" well, me too, bud. You can go look at it here.

Essentially, you log in to the thing with a google account and upload some photos. You have to participate (rate photos) in order to upload photos. Your photos are in turn shown to other users for rating ("keep" or "ditch" are the choices) and, optionally, a line or two of written critique.

Key points for rating photos: you get no indication of the photo's author, you do not see other ratings or critiques, and the photo shown is remarkably small.

The anonymity does a lot of good things. There's no pile-on effects of any sort. You're on your own, judging this thing on its own merits, as it were. It also seems to completely eliminate the site's dreams of being social in any way shape or form, so I am not sure where they're going with that. But let's get back to the photos.

What the photo's owner sees is what percentage of ratings are "keep" and any written critiques (unattributed).

The algorithm is heavily biased toward showing you recently uploaded work, which makes sense. Over time, you get more and more pictures in the system, but not proportionately more and more users, and you've got to get the new stuff rated. So your new picture will get examined for a day or two, and after that you're pretty much done. I have been getting anywhere from 0 to 5 written critiques, and I think something on the order of 50 ratings. You can tell for a while how many you have: 50% keep probably means "two ratings" if you see it in the first hour. 33% or 66% shows up next, generally, and so on. So, most of the time people just click one of the two buttons.

Based on some glancing at the "top users" ad "top photos" (which is one of the few ways you can begin to connect content with people, but it's very very tenuous) and the ratings their photos have received, people mostly click "ditch" -- it appears that you are doing pretty well to get a photo in the area of 50% "keep" ratings. The results in the "top" links are, maybe?, from the previous month, but only the top 40-odd pictures were over half "keep" ratings.

This makes sense, of course. People on a site like this are likely to consider themselves to have good taste, and especially given the anonymous nature of things (no incentive to be kind) are likely to click "ditch" constantly. Even I do, although I write critiques much more often than average.

You would think that with all these anonymous elements, that the top photos would be an eclectic mix.

They are not.

What ARS Beta users like is high contrast black and white photos, invariably with a single human figure in the middle or far distance, with strongly graphical shadow play. There are a few exceptions in the top photos, but not many.

The fact that the pictures are shown small tends, I think, to produce a strong preference for simple graphical designs. Anything not very simple tends to collapse into incomprehensible busy-ness. The fact that the users are weighted toward effete Eric Kim fans (wannabee street photographers) means that pictures should look as much like their lazy ideas of what an Henri Cartier-Bresson looks like as possible. The top photos include two direct copies of Henri's bicyclist shot, both much weaker than the original (they're sharp, for one thing).

They do not go for just any random garbage. I tested it, turning some random junk into a high contrast black and white photo, and achieved a record low 3% keep rating:

I tried some high contrast black and white geometry, and got up to a 23% keep percentage:

At this point I started throwing in people. This one, which the attentive reader will recognize, achieved 76% keepers, my best rating:

I tried a couple pictures of just people, portrait-y things. This is not a brilliant portrait, but at 17% it's clearly not just people in b&w that's floating the boat here:

Oddly enough they liked this one enough to give it a 62% rating, which is quite favorable. It's a very pretty picture, highly graphical, and it's b&w?

To their credit, they do not just like girls. This was just straight-up bait, and garnered only 27% keepers:

But onwards and upwards.




And on and on. I could hammer these things out endlessly, and some of them would do pretty well. Interestingly, the only actually good photograph I put up is this one (again, familiar to the long time reader). It's the only one with any "story" on its own, it's actually interesting, although perhaps it's been ruined by the ubiquity of the distracted boyfriend meme. It got a 45% keeper rating:

And to give you an idea of the level of written critique:

[Ditch] A man's attention is taken by a passing woman. Could be cropped a little tighter. Should have been taken a second later.

[Keep] Nice, he is looking at other woman, that's cool.

Neither of these commenters actually looked at the picture in any depth. The first is just stupid. 10 milliseconds later the shot was gone. The second guy is missing about half of the byplay.


ARS Beta is a bold effort, but I think it illustrates perfectly that virtually any community will have some sort of weird inbred tastes, and that people will learn to cater to them, and then you've got a hideous feedback loop of terrible.

But I think more importantly it shows us just how bankrupt the idea of the single iconic image is. The idea that you can meaningfully critique a single picture, without any context, without any notion of what it's for, is simply stupid. People will reach for whatever crutch they can, in this case "does it resemble one of the 2 or 3 photos by Henri Cartier-Bresson that I can remember?" with a dash of modernity thrown in -- they like sharp, highly contrasty, and a little of that "HDR look" lightly salted in will do no harm.

Ultimately the whole thing is a deeply stupid exercise in attempted ego-stroking.

I don't see how Eric Kim is going to monetize this thing, but I suppose if he can get enough users some idiot in Silicon Valley will give him a billion dollars.


  1. This seems like Tinder for photos, but I'm waaay to cynical. How would it be if the photos could post themselves? Now that sounds like a short story.

  2. How about a model where in order to be able to rate & critique, the critic has to demonstrate a certain level of knowledge of photography, and basic critical theory ? And a feedback loop whereby the photographer can rate the received critique ? Then _maybe:_ it starts getting useful and relevant - for those who crave the crutch of approval...

  3. "if he can get enough users"

    Bot farm...

  4. I concluded exactly the same as you. Though I did get them to fall for funny animal pictures which rated much higher than some classic Raghubir Singh street shots I pitted them against. The use of sexuality to prop up photo was a no no too I found. The critique was blown highlight iso too high abysmal too.