By dmitrizzleHeader image credit: Lara Berch
I don't think that having something curated is a big deal. Having it done well is, naturally, different.
For the past three years, I've been working with algorithms that curate content. We're way beyond Web 2.0 now. Interactive websites that react to our input can source our preferences and transcend them across devices. They're becoming smarter. It still feels like there is something missing.
Recommendation systems are incredibly complex. For the most part, they work by scoring a range of variables and assigning that compound value to each item. Afterward, the items are sorted according to the score. For example, if the computer knows that you like black and white photography, it'll assign a higher score to monochrome images - thus placing them at the top. That doesn't solve the immense technical challenges, which are associated with the complicated nature of a human being and our society.
The Harry Potter problem
This got nothing to do with the Wizards or the story of Half-Blood Prince. Sorry. The problem is in the popularity of this book and what it can do to recommendation systems.
Let's take Amazon to illustrate this example. Their algorithm recommends related products as "users also bought X." It works by calculating a relationship between multiple products that others have purchased. The problem occurs with overly popular sale items. The system "thinks" that particular goods are related to almost everything else in the store - since they often end up as one of the items in the shopping card. So the Harry Potter books are then absurdly promoted as "customers also bought" with screwdrivers.
Nagender Parimi talks about this fascinating and complicated issue in more detail on his blog. It's worth adding that there has since been a solution proposed and implemented. It's called the Netflix Challenge solution, and it is working by normalizing the undesirable effects in a way that I can only comprehend as magic.
But that's not enough. Our attitude towards discovery is a complex construct that doesn't (and perhaps shouldn't) have consistency. What if today I felt adventurous and wanted to find something that's the opposite of popular, something that very few seen? But tomorrow I'd like to learn all the new trends.
The biggest concern that I have, shared by Eli Pariser in his TED talk "Beware online 'filter bubbles'" is what I'd like to call a Bad DJ Problem.
The Bad DJ Problem
Loosely referenced by Amit Sharma in his excellent answer on Quora under Impact of Recommendations and Eli (see above). In short, "who are you to tell me what I want, robot?"
A practical example. Top forty stations are great. They are seemingly some of the few surviving styles of music broadcasting which can still make money on FM radio waves. I loved them when I was a kid, I still think they're OK, some of the time. Nowadays they seem to get old very quickly, though.
The "filter bubble" of the top forty formula amplifies the issue of monotony by ensuring that we only see content that's popular and liked by our peers - while increasingly disregarding anything, that isn't an "instant hit." I believe that this is the cause of our digression into the progressively simpler, easier to digest forms of entertainment. Which ultimately breeds boredom and a desire to consume more to compensate.
A good DJ would never let that happen. She'll dig through old classics and fresh new tracks. She'd know you well: your style and preferences and appreciate the fact that sometimes you'd want to hear jazz even though you are into hip-hop. She'll grow with you as a performer and a human being who desires more complexity and spice with the passing years. But give you a simple, catchy tune on Wednesday afternoon when the work gets rough.
Can a machine replicate the efforts of a "good DJ"? Perhaps. The algorithms being capable of predicting a desirable outcome for humans are getting better every year. But nothing close to a talented human being as of yet; there is even some debate that they could never actually catch up. Here's a link to a 4,000-word article by Tim Urban on AI if you'd like to explore it further.
On ArtSocket, there's a recommendation system already in place. You can see it in action depending on a number factors. A prompt that comes up with a suggestion for the most read article. I could make it base its judgment on many factors. For example, a number of print sales the page made or a measure of whether people read the article or just skimmed through. Even the types of visitors that came across the page. But I still can't let it control the kind of art I'm showing on the front page, or which articles get to stay in the magazine section.
I'm trying to be a "good DJ" when curating my art and essay collections on ArtSocket.
This job means working with beautiful, flawed pieces of art which someone poured their soul into. Not just data. And although they often look great on their own, together they must make another whole. An exhibition should demonstrate purpose, story and curator's personality. Reflected in what gets to be a part of it and how it's arranged.
My heart breaks when I get some submissions. Sometimes they're just bad. That's not a problem. But often there is a brilliant piece out there that would look fantastic on someone's wall, yet I can't accept it. It doesn't work with the collection or the style of the gallery. And that's the biggest, and perhaps the most significant challenge of human curation. To be able to use personal preferences, the vision, and expectations to benefit, not sabotage, the project.
Those objectives are to be broken. Sometimes.
There's a set of technical requirements, which dictates images to be of a certain size and of a particular quality. I'd like to think that I'm capable of overlooking them, in favor of meaningful, beautiful creative pieces. There are submissions that are blurry or have a ridiculous amount of grain. And that's OK. They make me feel something. They are beautiful despite the flaws. In fact, those imperfections make them what they are. And so they belong.
Everything on this website and leading up to its creation is just a series of events. It is a story, a narrative. We all construct life narratives - for ourselves, our pet projects or anything worth remembering. Narratives that we augment, edit and rewrite every time we choose to. Our entire life is a creative process.
The very first image posted is the beginning of a curated collection. It is the first sentence of a story that the following art pieces will extend and elaborate on. No list of images, events, songs or products is a collection unless it is a coherent, meaningful story. Best ones of the kind would be an artwork in their own right, speaking from the heart of the curator and producing emotions and understanding from within the viewer. A relationship that gives art its name.
The (online) medium
It took me a year to find the right kind of ink and print medium that would compliment the digital imaging on a screen. And even then each image has to go through additional color processing before being submitted to the shop. It obviously doesn't have to be this complicated, just owning a high-res file would generally be enough. But since a lot of energy was put into selecting and arranging all the work, might as well go an extra mile and ensure that it looks it's absolute best on the screen and on the paper.
Art curation and care are beyond what automated suggestion algorithms can do today. It is a meticulous process of selection, rejection and creative analysis that can only be done by someone who knows and appreciates creativity. Furthermore, that same person or team needs to have the technical expertise to understand what it would take to have digital images look great both in print and in pixels.This article is an edited version of what was originally posted on December 5, 2013.