Data Science?

Prometheus’s Fire or a Monkey’s Paw?

I went to a presentation for the opening of UVA’s new data science building. The presentation started off with a DALL-E-generated image of Prometheus with his fire overlooking a horde of humans holding torches of fire looking up. The idea was to display data science in a positive light, a technology that positively empowers humans.

This portrayal of data science is a bit disingenuous, considering its potential and ongoing challenges. The Promethean story of bringing fire to humanity is different from data science. Data science is patently devised by us, humanity. In either case, though, fire can cause wildfires, and with “AI” who knows what.

Lessons from Psychology & Data

In psychology, they often talk about different forms of “biases”—mainly “heuristic and representative bias.” These concepts can be mapped into the current discourse on why models fail:

  1. Relying on variables that don’t measure the phenomena.
  2. The dataset does not accurately portray the phenomena.

The above reason is why “Fundamental Attribution Error” exists in the first place. In the same light, the point of the “correlation vs. causation concept” highlights an aspect of “Fundamental Attribution Error,” which is that two variables can be correlated but not necessarily cause each other.

And suppose the “problem of other minds” has anything to say. In that case, there is already an existing epistemic gap between any observed behavior and the reasoning that went behind it (regarding using AI to predict human behavior).1


Citations

Eronen, M. I., & Bringmann, L. F. (2021). The Theory Crisis in Psychology: How to move forward. Perspectives on Psychological Science, 16(4), 779–788. https://doi.org/10.1177/1745691620970586

McCarthy, J., & Hayes, P. (1981). Some Philosophical Problems from the Standpoint of Artificial Intelligence. In Elsevier eBooks (pp. 431–450). https://doi.org/10.1016/b978-0-934613-03-3.50033-7


“The first task is to define even a naive, common-sense view of the world precisely enough to program a computer to act accordingly. This is a very difficult task in itself.”

McCarthy

  1. This epistemic gap has already been detailed by McCarthy (the founder of the AI field), and I feel like modern practitioners have forgotten about his work. ↩