A recent survey of AI use in the workplace found that “AI chatbots have had no significant impact on earnings or recorded hours in any occupation.”
Phew. I was worried that the whole point of putting AI to work was to put human beings out of it.
The working paper, published by the University of Chicago’s Becker Friedman Institute for Economics, provides volumes of stats collected from 115,000 study participants to “challenge narratives of imminent labor market transformation due to Generative AI.”
And, as with any study, the devil is in the data:
First, the project was a multiple choice opinion survey, so any of the numbers reported as objective statistics started as subjective choices categorized by the options presented to the participants.
While there’s complicated math involved that’s supposed to correct for formatting biases, the survey doesn’t measure what’s happening as much as what people think and feel about what may or may not be happening.
Opinion polls have given us the wrong predictions for elections and misled corporate marketers into thinking their brands mean something. A study of actual behaviors correlated with reported measures of productivity or profitability would have been far more conclusive.
This study wasn’t that one.
Second, the study was conducted in Denmark. Now, Denmark is a lovely country but it’s the size of the Upper Peninsula of Michigan and there are loads of externalities that make living and working there different from, say, where I live and work (starting with publicly-funded healthcare and education).
Employees there also can get up to two years’ worth of unemployment benefits if they get fired, so it’s fair guess that they think differently about work. It’s also quite possible that employers are less eager (or incentivized) to replace them?
Third, the respondents admitted that they’re using AI all over the place; three pages of aggregated replies to an “open-ended question” list the ways AI is transforming different occupations. Generative AIs are doing research, drafting content, and developing new ideas, and employees are seeing themselves spending time overseeing these activities (instead of accomplishing them).
This begs a question: Why? If they’re not seeing any impacts, why are they doing it so aggressively?
It could be that the warm embrace of Danish social services makes it safer to putter around with new tools? Maybe employees have been forced to do it?
Or maybe the point isn’t to yield any additive productivity or profitability benefits but rather train their replacements?
The study has nothing to say on this point, as its conclusion is narrowly constructed to gauge users’ opinions about AI’s utility…which is that they’re using it all over the place without seeing any return for their efforts.
Maybe they’re not looking at the right thing?
[This essay appeared originally at Spiritual Telegraph]