Ingenious New Research Abolishes Real Respondents

I just read a delightful and thought-provoking snippet in Proust’s Swann’s Way (published in 1913) that, unfortunately, reminded me of artificial intelligence in market research.

Here is what Proust originally wrote, which is delightful:

But all of the feelings we are made to experience by the joy or the misfortune of a real person are produced in us only through the intermediary of an image of that joy or that misfortune; the ingeniousness of the first novelist consisted in understanding that in the apparatus of our emotions, the image being the only essential element, the simplification that would consist in purely and simply abolishing real people would be a decisive improvement.

Here is a proposed revision to Proust’s paragraph, which I hope you will find disturbing:

All of the insights we get from real people are derived only through the intermediary of the data they furnish. The ingeniousness of today’s researchers consists in understanding that in the apparatus of gaining insight, these data are the only essential element. Therefore we can simplify the process by purely and simply abolishing real people, which may be a decisive (transformative, disruptive) improvement.

I am intrigued by the possibility of market research being conducted via large language models (LLMs) derived from the remnants of what real people have said. We would be like modern day archeologists, attempting to reconstruct human attitudes and behavior by excavating artifacts.

But market research can never abandon the real connection between data and people because it is real people that we want to understand. Abolishing real people from our research would push our work squarely into the world of fiction where novelists like Proust are far more eloquent than we are.

(P.S. It doesn’t work, either. See our companion post: Can You Really Use AI to Create “Synthetic” Survey Respondents? The answer is no.)

—Joe Hopper, Ph.D.