Curation, Discoverability, and the Great River

Choosing what movie to watch can feel like more work than watching the film itself. Have algorithms failed us? Is curation the answer? Or is it something else?

Curation, Discoverability, and the Great River

The Continuum is real. You see, there are millions upon millions of worlds in the universe, each one filled with too much of one thing and not enough of another. And the Great Continuum flows through them all like a mighty river, from have to want and back again. And if we navigate the Continuum with skill and grace, our ship will be filled with everything our hearts desire.

Nog, Deep Space Nine

I think a lot about what to watch next. With a dozen virtual video stores available on demand, at the click of a button, decision paralysis is a powerful foe. It's one streaming services have attempted to solve since Netflix first landed.

When I was younger, it at least felt a bit easier. Someone else's choices narrowed things down before I decided it was going to be a movie night. If I was going out, there were only a dozen movies in the theater to choose from, and there was usually either one new release (or something about to leave the cinemas) that jumped to the front of the pack. If I was renting? Whoever owned the video store had already decided what to keep in stock, plus other film-hungry shoppers got there before me to narrow down the selection. As for cable, well, HBO didn't only have control over what to show, but when it was going to be on.

My choices were limited by a mix of curation–choices made by owners or programmers of what to make available, and why–and algorithms: whatever formulae and decision trees those people used to figure out how many copies of a movie to keep in stock, how long to keep a film in the theater, or what time of day was best to put ROBOCOP II on the air for the fifteenth time.

It was simpler, then, but was it also better? Was I choosing better movies then, and if I was, was it because of how those choices were made available?

Marty Speaks

Marcello Mastroianni in Fellini's 8 1/2, peering over his sunglasses and tapping his nose.
Heart eyes for Marcello Marstroianni, forever.

Martin Scorsese published an essay this week about one of my favorite directors, Federico Fellini. Alas, it's an essay that's gotten attention primarily for its opening decrying the current state of cinema and distribution. It's a topic that's been top of mind for Scorsese for some time, and one that's caused its share of out of control Twitter arguments between... let's be honest, just about everyone.

In the opening, he weaves two parallel concerns together. One, that streaming distribution and corporate perspective on art reducing everything to "content" served by algorithms has compromised audience's ability to discover important films. Two, that the collapse of everything down to "content" has diminished the art of cinema itself by failing to differentiate it from, say, cat videos. I have thoughts about the latter argument, but I'm setting it aside to focus on the former: Have we created a system where it's impossible for audiences to be more than the end user of an algorithm serving up content? And, if so, how did that happen, and where do we go next?

Scorsese's main concern is that we've abandoned curation in favor of algorithmic recommendations. Audiences aren't presented with intentional choices by someone with a perspective, a point of view or love of film to share, but instead by a computer that wants to keep you engaged by choosing things you're most likely to keep on.

Certainly, Netflix has gone all in on this idea with an extremely cursed Shuffle button that picks a totally random (only not random at all, more on that later) thing and throws it in your face. Navigation of most streaming sites rests just as hard on this kind of recommendation-based structure that goes well beyond simply making suggestions. Retail sites sort and filter based on what they think you want, but give you the controls to narrow down the full catalogue however you want. Netflix, meanwhile, lets you browse infinite categories, only to fill each category with the same twenty or thirty options.

The prescription to this ailment, Scorsese says, is curation:

Curating isn’t undemocratic or “elitist,” a term that is now used so often that it’s become meaningless. It’s an act of generosity—you’re sharing what you love and what has inspired you. (The best streaming platforms, such as the Criterion Channel and MUBI and traditional outlets such as TCM, are based on curating—they’re actually curated.) Algorithms, by definition, are based on calculations that treat the viewer as a consumer and nothing else.

Framed this way, Scorsese's argument makes sense. Curation brings a human influence into what other humans should watch, while algorithms can't and won't serve the need for people to find new and challenging things. To a computer, you are a consumer. To a curator, though: you're a friend.

Scorsese has a point, but he's making an assumption that human curation is immune to what causes algorithms to fail. Or, maybe the assumption is that algorithms have created a kind of failure that didn't exist before Amazon and Netflix. Either way, in my opinion, that assumption is wrong.

What do we mean by curation?

Tom Baker as the Fourth Doctor in "The Day of the Doctor", tapping his nose and asking, "Who knows?"
A good question, Doctor! Or should I say, The Great Curator.

If you're lucky enough to have lived near an independent theater run by people whose taste you love and respect, you've probably experienced the best film curation experience around. Not only do you get the benefit of someone with access to rare or forgotten films making them available, you get to watch those movies with them and a bunch of other people. It's a shared experience, set up by people with an interest in making it one you can only get at their theater.

Curation is an expression of someone's–or a group of someones'–priorities and point of view. What subjects interest them, which filmmakers left an impact, and who they hope to reach with the films chosen are all weights on the scale. A service whose films are intentionally curated will be specific and idiosyncratic. Those films will often be framed in a way that helps the audience understand why they were chosen, and how best to think about them.

If that all sounds great, it often is. It's also a double-edge sword, because lurking underneath our priorities and interests, there's something we haven't always examined and unpacked enough to avoid falling prey: bias.

The question we have to grapple with if we believe wider, more systemic curation in streaming services is a tonic to the trend in soulless algorithmic recommendation is who, specifically, will be doing the curation? It might feel easy to brush this question aside, but taking even a cursory look at the film industry tells you how they'd answer it if left to their own devices. Even if put in the hands of creative forces instead of corporate ones, the relative monoculture of Great Directors is a massive thumb on the scale of whose priorities would be represented.

I don't use bias negatively, or to mean only "prejudice", either. We have biases that don't impact issues of social justice that would still lead to prioritizing some films over others.

A silly, too-simplistic example: If what you value in film is the visual language, rich cinematography, and thematic ambiguity over storytelling structure, political commentary, and dialogue, you might be more likely to choose The Thin Red Line over Robocop. There's nothing wrong with that preference, but it can become a problem if too many people choosing films for services share that bias. Curation then becomes an unseen lens through which Malick (dreamlike, unstructured, visually poetic) can feel important, valuable, inherently artistic, while Verhoeven (satirical, tightly structured, visually straightforward) does not.

Late in Scorsese's piece, without seeming to realize what he's said, he gives a tragic example of the great curators saying no to the filmmaker he's written an entire essay to praise.

He’d had a difficult time with his producers on that project—they wanted a grand Fellini extravaganza and he gave them something much more meditative and somber. No distributor would touch it, and I was truly shocked that no one, including any of the key independent theaters in New York, even wanted to show it. The old films, yes, but not the new one, which turned out to be his last.

Curators giveth, and they taketh away.

This gets worse, obviously, when it comes to social inequalities. Scorsese himself has put a lot of time and money into elevating and preserving films made outside of Hollywood. He cares about promoting diversity in cinema, and a service curated by him would likely do the same. That said, there's an important difference between trusting well-meaning white men to curate movies on behalf of marginalized filmmakers, and members of marginalized communities being given the power to do so outside of their communities. Who gets to curate, and for what services, for which audiences matters. It matters a lot.

Isn't that the problem we currently have with algorithms, though? There's a thumb on the scale either way, why not have it be a human's?

It's true, we do have that problem now, but the answer to why "have a human curate" isn't as complete as it seems starts with understanding what an algorithm is and isn't. When you strip the technology away, it's effectively still human priorities making the choices. The difference is which humansare putting their thumbs on the scale, and to what end.

So, what's an algorithm?

Black Widow in Captain America: The Winter Soldier asking, "What kind of algorithm? What does it do?"
A good question, Natasha!

There are a lot of misconceptions as to what an algorithm is, and the blame for that belongs squarely on the shoulders of Silicon Valley venture capitalists and founders who need people to think they've created something magical. Algorithmic modeling and analysis these days ranges from impressive to terrifying, so it's easy to treat them as something more than what they are. Interview with just about any tech startup, and you'll get an earful about something that sounds more like pixie dust than code.

Not to oversimplify things, but an algorithm is really just something that tells a computer how to navigate a flow chart. Which advertisement should we show this client? Let's see, what's their gender? Male? Okay, age? 25-40, cool. Have they visited any of these thirty sites in the past few days? They have? Awesome, so which one is paying higher rates? Show them that one.

When you get into something like Netflix's recommendation algorithms, the flow charts get exponentially more complicated, but at their core, all of these algorithms are meant to do the same thing, and that thing is make choices on behalf of users, weighted on criteria that the people building said algorithm have decided is most profitable.

That last part, the how an algorithm makes decisions based on data, is key. The lie of algorithms is that they're unbiased arbiters, hearing what a human wants and giving them the most relevant responses. This lie cuts two ways. On the one hand, it convinces people who want to "remove bias" from a process that handing things over to a computer will do the job. On the other, it can make people rightly skeptical of them concerned about the wrong things.

In his essay, Scorsese ponders, "[It] has created a situation in which everything is presented to the viewer on a level playing field, which sounds democratic but isn’t. If further viewing is “suggested” by algorithms based on what you’ve already seen, and the suggestions are based only on subject matter or genre, then what does that do to the art of cinema?" Seeing recommendations as creating a "level playing field" for all content, dictated only by prior viewing habits, presenting only things the viewer might like based on what they watched last, deeply misdiagnoses the problem.

Stroll through your Netflix account and you'll see what I mean. Choose a category. Any category. Look at the films in it, the pick another category and do the same. The pattern isn't a crass, targeted attempt to show you things you might love that resemble what you saw yesterday. It's a crass, targeted attempt to stuff every category they think you like with movies and television shows exclusive to their service, whether it's a good recommendation or not. Scorsese fears an algorithm so unbiased by anything but giving audiences what they want (over what they need), that new things will not be discovered. The world we've actually created is one where viewers are constantly harangued to watch things they've shown only superficial interest in, provided they're owned by the company suggesting it. Nor are any algorithms putting "cat videos" on the same level as movies, though not because they agree one is art and the other isn't. One is simply monetized differently than the other in executives' minds.

Both human curation and algorithms have the same basic desire for audiences to watch one film over another. They're both, essentially, a person or group of people who have decided these films are more valuable than those films, and try to direct audiences into agreeing. The issue with algorithms isn't based on them being code running on a machine. It's that algorithms are human bias encoded and run at scale, hiding the basest kind of curation: a collection made up of whatever is most profitable.

Both curation and algorithmic recommendation, then, have the same problem: neither are inherently concerned with what I, a human person looking for something to watch, actually wants. They hope they can help, sure, but they want to help by suggesting what they want me to watch. Sometimes, this is just the ticket. Other times, it means a movie that I'd love, that would challenge or entertain me–but that a curator thinks is trash and an algorithm thinks is insufficiently profitable–never shows up on the list at all.

The Great River

Gene Siskel and Roger Ebert. Ebert says, "Sound a little excited, Gene." Siskel replies, "Sound less excited, Roger."
A golden age, now gone.

In a late episode of Star Trek: Deep Space Nine, chief of operations Miles O'Brien is ordered to make repairs on the U.S.S. Defiant in three days, but the needed part won't be available for weeks. Nog, a Ferengi member of his engineering staff, offers to help, kicking off a boggling chain of trades intended to get the part sooner. Stuck in the middle of Nog's incomprehensible deals, he describes his philosophy with the quote that kicked off this piece: He believes that a continuum flows through everything, like a great river. If you navigate it correctly, you'll find whatever you need, but maybe not in the way you expected.

Figuring out what movie to watch these days feels a lot like navigating that river. You can choose by grabbing the first shiny thing that shows up in front of you, or you can try to sail the currents, through "If you liked..." recommendations, collections on Criterion Channel, dozens of half-forgotten films you queued up over years, and stuff friends have begged you to check out. Which combination of those leads you not only to something you'll enjoy, but a film that might expand and challenge you, is different every time. It falls on everyone to figure out for themselves, and it takes a lot of work. It makes me nostalgic for a time not when someone else was driving the boat. What I miss is having people to give me directions when a guide was what I wanted.

What Scorsese gets wrong is less about the value of curation than it is focusing on the distribution or streaming channel itself to solve things. Any corporation big enough to have a library worth subscribing to will have incentives counter to finding you or I the best films for us as individuals. Even when they get it right, how they get it right will inherently leave something else perfect out.

I think, instead, the thing we've lost over the last twenty years is a source of self-curation that's been gutted by mergers, budget cuts, and long-term cultural dismissal. That source is deep, personal, professional criticism. The trend in the early aughts towards "recaps" over reviews, that culminated in even those falling victim to layoffs, has made the art of critique and analysis something only for those with the time and resources to do it on their own.

Growing up with a critic like Roger Ebert was a blessing. His reviews were insightful, personal, and worthy pieces of writing on their own. They were also paid for and widely published by a major newspaper, allowing him to put the effort into his field that made those reviews more than blunt recommendations. As criticism has become increasingly more difficult to sustain oneself on, the avenues people might have used to find risks that excited them have closed. Ebert didn't choose the movies that were available, nor did he direct anyone's path through a video store. Instead, he supported the art he loved by arming the readers who valued his opinion, the same way other critics did the same for their own audiences.

This still exists, though it sadly falls on the backs of individuals to do it for little to no pay, often as a hobby. Streaming services that choose films intentionally and deliberately, curating its selections, are necessary and wonderful things. So, too, are the rare cases where algorithmic recommendations truly aim to suggest options based on what you might want. I don't think either, or even both, gets us all the way there. We need more, like what worked so well in my early days of getting into film: the best video stores, carrying as wide a selection of films as possible, helped along by employees you came to trust, and a handful of reviews that stuck in your head months after reading them.

As wide a selection as possible, as easy to access and search as it can be, and the advice of guides to narrow things down when the great river felt like too much to navigate alone.

We can get that back no matter how films are distributed. Even if the theater chains close, even if every streaming service is an anonymous, difficult-to-navigate warehouse of options overlaid with the worst of algorithms. If we chose to pay people to tell us why they loved what they love, value what expanded their boundaries, hate what drove them bats, we don't need our distributors to choose what films we see for our own good. The river will deliver all we need.