The advertising industry has spent decades perfecting the art of measuring things that may or may not have happened in ways that may or may not be accurate. Now, according to one prominent agency executive, artificial intelligence stands ready to make this situation either considerably better or considerably worse.

Ben Hovaness, executive vice president of specialized capabilities at Omnicom Media, offered this assessment Tuesday morning at the Advertising Research Foundation's CIMM summit in New York, in a keynote address that toggled between warning and exhortation. The Coalition for Innovative Media Measurement, which has been convening industry executives to discuss such matters since 2009, had evidently decided to begin the proceedings with a bracing dose of anxiety.

Mr. Hovaness's central concern involves what might be called the measurement hall of mirrors — the practice, now widespread, of blending empirically observed results with modeled ones and presenting the combination as a single number. Google's approach to conversion reporting, which Mr. Hovaness noted has been standard practice "in certain quarters of the industry" for five years now, offers little transparency about which conversions were actually counted and which were estimated by algorithms.

The dystopian scenario, as Mr. Hovaness described it, involves a self-perpetuating loop: a vendor reports favorable performance data, an automated buying system interprets this as a signal to spend more, which generates more data of similarly uncertain provenance, which the system interprets as further validation. (One might observe that advertising has always had a certain capacity for self-congratulation, but at least the humans doing it knew they were guessing.)

"If you think that this is an overly pessimistic view, I want to assure you that it isn't," Mr. Hovaness told the gathering, "because in some corners of the industry, this is how we've been operating for years."

The more hopeful alternative, he suggested, involves using artificial intelligence to accelerate the production of legitimate measurement studies — collapsing the time and labor requirements that have historically made rigorous analysis expensive and infrequent.

Whether the industry chooses the golden age or the hall of mirrors remains, as Mr. Hovaness put it, a matter of collective will. The machines, it seems, are agnostic.

Original story published in MediaPost: "A Tale Of Two AI Measurement Futures: Dystopian, More Utopian"