Advertisements

The “data-driven” cliche

Two weeks ago, a website I write for recently decided to “pivot” (to steal another unbearable term) to being more “data-driven.” I asked the editor for the site what this shift would mean for the writers, and he explained that we should include more stats, mostly, and presumably fewer sentences of innumerate rambling. My first attempt in the new mode was a two-part series written in the voice of a Carl Diggler-esque character who likes to reference H.P. Lovecraft alongside stats from the U.S. Green Building Council.

When talking about being “data-driven,” whether in regard to marketing, writing, political forecasting, or running a startup, it is important to think about what proponents of such an approach are ostensibly pushing back against. There seems to be this widely shared notion that in fields as diverse as politics, sportswriting, art criticism, and business, there are vast hordes of people who can’t understand math and science and are finding their paths instead through intuition and personal experience (indeed, the Diggler character was willfully created as a parody of pundits who trust in “gut” and “personal experience”). It is a sentiment that dovetails nicely with the widespread myth that there is a “STEM shortage” (easily disproven) and that the best use of young people’s intellects is in helping organizations work on basically finished products that require little more than basic Excel skills and 8th grade algebra.

This attitude takes many forms. There is the lieutenant governor of Kentucky telling students not to major in history (the governor said something similar about French literature), and there’s the website FiveThirtyEight repeating slogans such as “I believe in Math” to push back against people skeptical of the site’s opaque (but “data-driven”!) statistical modeling that somehow failed to predict Donald Trump’s dominance or the breadth of Bernie Sanders’ support.

OK, so I’m one such skeptic, not just of FiveThirtyEight in particular (and its absurdly wrong predictions about the results of the Michigan and Indiana primaries in the Democratic Primary in 2016) but of the “data-driven” cult in general. To give them a fair hearing, though, let’s look at what the proposed benefits of this approach are. Here’s a sentence from a Forbes piece about Uber:

“In today’s data-driven sharing economy, pricing, tasking and hiring decisions are all based on the cold hard logic of computer models and data.”

What is “data,” though? Data is the result of what you have chosen to measure. For Uber, germane “data” might include the price of a discounted Uber ride versus a standard taxi fare, but not the amount of car insurance paid by its drivers or how much of its clout comes not from the free market “innovation” principles it cherishes but from the cold, hard logic of deep-pocketed venture capitalists (Google among them). Data is manmade.

The assumed infallibility and rationality of computers is a quaint idea, held by the same sorts of New Economy types that would be simultaneously dismissive of a sincere belief in God, despite such faith being almost indistinguishable from blind trust in Data, The Market, Algorithms, or any number of manmade institutions. Not only are computers of any type manmade (with particular baggage in the Cold War era) and capable of being stuck halfway between their binaries, but the way they work isn’t even an accurate approximation of human consciousness, i.e., the way that we experience anything. The brain is not a CPU running algorithms on data.

Yet you get the likes of John Gruber writing stuff like the following, about Facebook’s news curation efforts:

“The irony is that machines don’t have a point of view — they are ‘objective.’ Over the last half century or so, mainstream U.S. journalism has evolved in a way that has writers and editors acting like machines. They’ve made it easier for themselves to be replaced by algorithms. Most readers won’t even notice.

I do two things here at DF most days: find interesting things to link to, and comment on them. An algorithm may well beat me at finding interesting links. My job then, is to be a better writer — smarter, funnier, keener, more surprising — than an algorithm could be. When I can’t do that, it’ll be time to hang up the keyboard.”

“Machines” are whatever you make them; they follow their creators’ rulers. They don’t merit special authority for the particular way they solve problems, not any more than the the Old Testament did because it was written right to left in Hebrew instead of left to right in English. Being “better” than an algorithm is a meaningless statement; the human brain is not running algorithms itself, and the ones that it comes up with for other contexts depend entirely on intent. An algorithm has no sense of whether anything is “funnier, keener, more surprising” – it’s a fucking set of numbers! The impression it leaves is all in the interpretation.

The cult of data-drivenness has created both unbearable resignation (“time to hang up the keyboard,” gag) and equally off-putting arrogance. For the latter, look at the prima facie absurdness such as Golden State Warriors owner Joe Lacob saying that a team had never been built “around the 3-pointer ” before Silicon Valley VCs got involved in the NBA via the Warriors franchise (however, his statement did produce this amazing parody piece about the Cleveland Browns’ algorithm-driven approach to winning a future Super Bowl).

Like anything, obsession with “objective” “data” is not new. I recently re-read Aristophanes’ Frogs, a play about Dionysus descending into Hades to witness a debate on literature between Aeschylus and Euripides; it’s sometimes cited as the first major work of literary criticism. In Paul Roche’s translation, there is this passage about how Euripides analyzes drama:

“OLD SERVANT: We’re going to see something great:
            Poetry sold by measurement and weight.”

XANTHIAS: What, tragedy on the scales like pork chops?

OLD SERVANT: Yeah, with yardsticks and measuring tapes,
           Words’ll be fitted into little boxes an’ …

XANTHIAS: You mean, like making bricks?

OLD SERVANT: Sure thing, with rulers and setsquares,
          Cuz Euripides says ‘e’s going to analyze
          Poetic tragedy syllable by syllable.”

Data-driven analysis! No matter how much I rail against the arrogance of the data-first crowd, I could never come up with an image as striking as Euripides putting words into “little boxes” for analysis. Were those boxes determining what  was the “smarter, funnier, keener” #content placed before Dionysus? That’s for humans to decide.

 

 

 

 

Advertisements
%d bloggers like this: