How We Trained an Algorithm to Predict What Makes a Beautiful Photo
EyeEm
33820

The debate about whether computers can do what human minds do sidesteps the more important humanist question of what a photo can accomplish rhetorically. Sure, aesthetic standards of “good” form may be roughly translatable into data—controlling, of course, for cultural and historical variation—and that translation seems like compelling work.

But to know whether or not a photo is “good” is to know something of astounding insignificance. That a computer can make that judgment makes the judgment of no greater significance. We are still left asking, So what?

A rhetorical perspective on photography asks a different, and much more engaging set of questions: Whether “good” or “bad,” what can a photograph do? What social work can a photograph accomplish? What power structures can it reify or disrupt? What cultural narratives can it put on display? These are questions that matter in a humanist sense. They focus our attention not on the quality of form, but its social and political function.

Get a computer to crunch that kind of “data” and you’ll really be onto something.