The disparate impact of AI on jobs

AI job displacement is often in the news, but a more nuanced look at specific professions tells a complex story.

I was recently at a dinner party, chatting with three old friends. One is a speech pathologist, one a therapist, and one a sociologist. The topic of AI came up, as it often does when I’m in the room (I swear I’m not usually the one to bring it up!). In the ensuing discussion, something interesting occurred to me. None of my companions are terribly at risk of job displacement due to AI any time soon. And in fact, a story that is not talked about enough is that in at least one of these cases, AI will likely increase the demand for human workers. Although this is a small sample of professions, I think there is an interesting counter-narrative here to the AI-will-put-us-all-out-of work concerns I’ve heard way too much these days.

I’ll start with the profession that I am most familiar with: college professors, like my sociologist friend. I’ve written about this extensively before, but I’ll recap my beliefs here. The biggest issue that AI is creating for professors is what to do about student cheating, as AI can now do a reasonable job on most assignments in most college classes. As I’ve written about before, I don’t think there is one solution to this. Any solution is going to have to be highly tailored to class goals, content, level, format, etc. That creates a headache for every professor, as they are now forced to adapt all their classes. While this may be a major inconvenience, I don’t think it threatens the profession as a whole. One can worry that with intelligent AI tutors students no longer need to go to college to attain the same knowledge, but people have been making similar arguments since there have been publicly available textbooks. At least for the foreseeable future, college serves a valuable role in providing the necessary social motivation that many students need to learn, and the certification of that learning that is generally required post-college. In short, the job of the college professor is safe, although it will have to adapt to the new AI world.

The job of a speech pathologist is also quite safe. My friend works in a hospital, primarily with post-trauma patients who have lost some ability to speak (or in some cases, even swallow). Coming up with a treatment plan involves synthesizing a number of information sources, such as visual inspection of the patient’s throat, listening to whatever sounds they can make, x-rays of the injured area, brain MRIs, etc. While AI diagnostic tools are becoming increasingly accurate (and will be an important tool for speech pathologists), some of the tests require human touch. More importantly, once a proper diagnosis is made the patient moves into treatment, which is a very human interaction. There will always be a need for a human in these kinds of highly personal interactions with medical professionals.

Finally, I’ll discuss the job of the therapist, for whom I think the effects of AI may be the most interesting. Many have questioned the future of this profession, as chatbots can hold complex conversations. There are several LLMs now that have been specially trained to act as a therapist, and they are becoming widely available. Many people report it being easier to open up to an AI therapist, as there is complete anonymity. Furthermore, an AI therapist is available 24 hours a day, for as much time as someone might want to talk to them. And to top it off, talking to an AI therapist is MUCH cheaper than talking to a human professional. And yet, despite all this, I believe that not only is the job of the therapist safe from AI obliteration, I think the demand for human therapists will increase due to AI.

The reasons are simple. First, any therapist will tell you that good therapy is an experience that necessitates an intensely human connection. It is not uncommon for a patient to cry in therapy, for example, as they become overwhelmed by emotion. It’s harder to imagine someone having the same kind of experience while typing away to an LLM (although I’m sure that happens). On the other hand, some therapists report that asking patients to dialogue with an AI before their in-person meeting can be helpful, in a similar way that journaling can be helpful. It can focus patient’s thoughts and make therapy sessions more productive. For this reason, there is growing evidence that LLMs used in conjunction with traditional therapy may be more effective than therapy or AI dialogue alone.

So why do I believe AI will lead to more demand for human therapists? It’s because a lot of people just aren’t used to talking through the issues in their life that they are struggling with. However, because of the anonymity that AI affords, many of those people are opening up more, and its reasonable to expect that some fraction of those people will realize that to go further, they’ll need to talk to a professional. In brief, LLM chatbots offer a new inroad to traditional therapy.

What’s interesting looking at all three professions is that its clear (at least to me) that AI will have a real transformative effect, and yet will not lead to the job losses that many are worried about. Teachers may have to adapt their classes, but when used properly AI tutors can provide additional learning experiences that further teachers’ goals. LLMs can serve as valuable diagnostic tools for clinical medical professionals. For therapeutic treatment, LLMs may be a valuable augmentation.

Don’t get me wrong. I do expect that there will be job losses due to AI. However, job loss is only part of the story. Job transformation due to AI is another. And to top it off, new job categories that don’t exist now will be created by AI. As with all things AI, the true picture is complex enough that it’s hard to make universal statements.