Data Driven Creativity

My school started much earlier than my kids' school this year, so I spent a couple weeks at home while the rest of the family visited relatives across the country. I am not too proud to admit that I bingewatched an obscene amount of TV during the two weeks they were gone while I was completing some writing projects. It's really the first time I have done so; while I have shows that I like, I rarely get to watch them all at once, or to pick the next one on the list in rapid succession.

So, it was with a new interest that I enjoyed The Second Digital Disruption: Data, Algorithms & Authorship in the 21st Century by Kal Raustiala (UCLA) and Chris Sprigman (NYU). A draft of the article is on SSRN, and they blogged about it in a series of posts at Volokh Conspiracy. Here is the abstract:

This article explores the intellectual property ramifications that flow from the explosive growth of mass streaming technologies. Two decades ago rising internet usage led to what we call the first digital disruption: Napster, file-sharing, and the transformation of numerous content industries, from music to news. The second digital disruption is about the age of streaming and, specifically, how streaming enables firms to harvest massive amounts of data about consumer preferences and consumption patterns. Coupled to powerful computing, this data—what Mark Cuban has called “the new gold”—allows firms such as Netflix, Amazon, and Apple to know in incredible detail what content consumers like and how they consume it. The leading edge of this phenomenon—and the primary vehicle for our examination—is the adult entertainment industry. We show how Mindgeek, the little-known parent company of Pornhub and the dominant player in pornography today, has leveraged data about viewing patterns to not only organize and suggest content but even to dictate creative decisions. We first show how the adult industry adapted to the internet and the attendant explosion of free content. That story aligns with many similar accounts of how creative industries adapt to a loss of control over IP by restructuring and recasting revenue streams. We then show how content streaming firms have used data to make decisions about content aggregation, dissemination, and investment. Finally, we consider what these trends suggest for IP theory and doctrine. A key feature is that by making creative production less risky, what we call “data-driven authorship” drives down the need for strong IP rights.

I thought the discussion about how data drives what to create to be fascinating, and the article is well worth a read. I think the perfect example of what the authors are describing is the Netflix movie Bright, in which Will Smith plays a cop who teams up with an Orc on the LA Police. The movie was critically panned. Rotten Tomatoes: 26%. But viewers seem to like it a lot: Rotten Tomatoes Audience Score: 84%. Netflix is surely on to something here.

I could certainly see it playing out as I watched. I watched "The Five," a show by one of my favorite authors, Harlan Coben. So then Netflix gave me nothing but mysteries and suspense to watch, plus another show by Coben, Safe (both were great, by the way). But I'm not really a mystery show person - I like sci-fi. So, I watched one show, and then the suggestions got weird: do I like mystery? sci-fi? sci-fi mysteries? I wound up having to dig a bit for the next show.

But here's the interesting thing: the quality of the shows varied wildly, even among the genres that I liked. The writing, acting, editing, and direction mattered. I don't know about the Mindgeek and porn clips, but I will note a couple distinguishing factors. First, there is likely a...er...utilitarian factor associated with those works; people are not watching for the articles, as it were. Second, the works are much shorter; it is much easier to have a highly focused 15-25 minute clip than a 10 episode series. Even with these differences, I suspect viewers have their preferences about what they see in the different clips with the same data driven attributes.

My broader point, then, is that how we consider the effect of data driven works will depend a lot on how we view creativity. The data certainly reduces the creativity in certain major plot points, as well as the quantity of different types of works. But to some extent studios have always done this, only with rules of thumb and intuition rather than actual knowledge. In that sense, data will democratize creativity - if viewers want more women in better roles, there will be more women in better roles; no need to rely on a male studio executive's views on the matter.

Beyond selection, though, I suspect there is still room for surprise, storytelling, differentiation, and other forms of creativity. Consider Bright: write what you want, but it just has to star Will Smith, include the police, and feature orcs and elves. At the limit, too much data may constrain creativity, of course - the more you add, the less you can create.

To be clear, Raustiala and Sprigman don't say anything that contradicts my intuitions here. They make clear that creativity is on a continuum, and that data merely slides to one side. But they do question how viewers will perceive works, and it is there that I disagree with them. I suppose that we could hit that limit where everything is automated, but my gut says that despite having preferences for particular story aspects, viewers will always be able to separate the wheat from the chaff (though not the way I would - as just about every American Idol vote shows) and thus will always look for something new and different within their preferences. At least, I sure hope so.