People can’t help seeing signals, even in noise.

A recent NY Times article included a line that has stuck with me over the last few days, “So great is people’s confidence in their ability to glean valuable information from a face to face conversation that they feel they can do so even if they know they are not being dealt with squarely.”

The article is about job interviews, and the frequency and degree to which people will overestimate their ability to rate a candidate better than a set of data. Even when given nonsensical data, people had no trouble crafting a narrative about the candidate and using that to justify a hiring decision.

In how many other scenarios, do we create stories from unrelated data, and use it to justify a decision?

This is perhaps why AI has become so appealing in healthcare. (The New Yorker also examined the human vs machine decision making this month.)

As someone who works in AI and algorithms to support health decision making, this struck me as another place where it isn’t either/or, but both. The challenge is that finding the role for human and algorithm/machine will vary depending on the health condition and the data available. Behavior change counseling isn’t diagnostic radiology. But in both cases, we should be interested in whether people are making up compelling stories from the data because they can.