AYESHA RASCOE, HOST:
Is AI coming for your job? Well, if you're a woman or live in a city or voted for Kamala Harris in 2024, AI is more likely coming to your job, if not for it. That's according to researcher Jonathan Hersh. He teaches economics and management science at Chapman University's business school, and he joins us now. Welcome to the program.
JONATHAN HERSH: Thank you very much.
RASCOE: You're not claiming that artificial intelligence is targeting working women or urban workers. Give us a thumbnail of your findings concerning AI and demographic groups in the workforce.
HERSH: So AI is changing what we do at work, not whether we have work. But we see that the people for whom AI is changing their jobs first are typically white-collar, urban and often vote Democratic, and that really flips the script on previous tech disruptions, which hit manufacturing and rural communities first.
RASCOE: I've been thinking about this a lot 'cause I've been moving. AI cannot move your furniture, right?
HERSH: Yeah, so that's the starting place of a lot of this work, is that what we do is we look at the job level and we decompose the job into all the tasks that the job performs. And for some jobs, AI is very good at a number of those tasks. You know, think computer programmer or a data analyst. AI can assist with a lot of those tasks. And, you know, this could be good or bad depending on the job. You know, if you're looking to break into computer programming and you're working in a framework or a programming language you've never worked in before, AI could be really good 'cause it teaches you how to work in that. You can produce more code, and that could be really beneficial. But if you're in a job in which AI is moving too fast for you, that could actually, you know, have downward wage pressure on you.
RASCOE: Why, on balance, does your research say women will be more exposed to AI in their jobs?
HERSH: We're aggregating over thousands of different jobs, and so it just turns out that on net, women happen to be slightly more exposed to very high-exposure AI jobs. And we see men tend to be in very low-AI-exposure jobs. Men tend to work in the trades that have more physically demanding tasks that have less relative AI exposure.
RASCOE: On a tour recently, Bill Gates used education in health care as prime examples of AI disruption because basically, AI in 10 years could be giving out great medical advice and great tutoring, and so it would really disrupt those industries. What do you think about those sorts of predictions?
HERSH: Yeah, if we look at the vast majority of health care workers, a lot of the tasks they perform are not going to be aided by assistance from large language models or generative AI. These are very physically demanding tasks - you know, moving patients, examining patients. Where we see artificial intelligence particularly helpful is in the diagnostics realm, in the analysis realm, in developing treatments.
You know, the other thing to consider with health care in particular is that regulation plays quite a big role here. So even though AI can assist within that, there's a level of regulation that might not permit AI to work within that task. In addition, there's a degree of humanness that people expect from their health care workers, and they wouldn't necessarily want AI to perform at that role. So you have to have a balanced approach with this. Just because AI can do this role doesn't necessarily mean that it is going to have an impact on it.
RASCOE: We mentioned Harris voters in the introduction, and you've done research for, or at least research used by, Democratic pollsters. Why would AI exposure skew toward Democrats?
HERSH: I think of it more in terms of the reverse. It just so happens that people who are knowledge workers have tended to support Democrats. You know, so it's less about the impact of AI as moving in one direction or another. It's more about, you know, the characteristics of those workers who tend to support Democratic and the fact that they're in these roles that can be augmented by artificial intelligence.
RASCOE: You say, you know, you're in the early stages and that policymakers need to be poised to help people navigate these changes that are driven by AI. What kinds of things do you think lawmakers should be looking at?
HERSH: So there are some no-brainer policies that we could pursue which specifically are around job retraining or job training. Take a marketer who's been working in that marketing job for 10 or 20 years. Now AI is part of that marketing job. There should be some type of policy that allows that worker to understand that new way of being a marketer or transitioning that worker to a job that has relatively less AI exposure.
You know, the second thing is that right now in America, we couple health care with your job. We should think about sensible policies that allow people to keep their health care that's not tied to their employment. I think, in general, we really need to have a rational and sensible policy around AI that keeps dignity of workers, that maximizes its benefit, while downplaying the risks that it might pose.
RASCOE: That's Jonathan Hersh of Chapman University. Professor Hersh, thank you so much for speaking with us.
HERSH: Thank you very much. Transcript provided by NPR, Copyright NPR.
NPR transcripts are created on a rush deadline by an NPR contractor. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.
