Ethics and LLMs: Replacement and Deskilling

When people think about AI’s impact on the job market, it’s often very black and white: “I had a job, but then my company bought a computer that can do the job. I no longer have a job.” For those of us who have spent time with the LLMs, we know that they aren’t good enough yet to 1-for-1 replace many jobs.

But that doesn’t mean that there aren’t serious potential near-term job market impacts. For example:

“I work as a freelancer, and now a lot of people who would have hired me to design a book cover, write a blog post, or redo their logo will use an LLM instead of hiring me. The smaller projects for smaller clients are gone, now there are lots of freelancers fighting for the few remaining good projects. Competition is driving the price down, even for the more involved projects, and it’s more difficult to make a living.”

”I am one of ten people in my job role. We used to do tasks X, Y, and Z. Now, AI can do Z. Our time spent on Z is cut by 90%. Now, we need fewer people doing X and Y, because we all have more time for it. We are going to stop backfilling positions when people leave, and when we grow, we won’t hire as quickly. Across the company and the economy, there are less positions available than there otherwise would be for people like me.”

All of these scenarios ar examples of replacement, where a technology replaces a person for a job or task. The net impact of these more subtle types of replacement is still reducing jobs and pay for humans. These types of replacement reduce costs for organizations, but mission-driven organizations may want to consider the impact of replacement on their local community and the follow-on impacts on their reputation.

For example, if a mission-driven organization uses images or videos that are made by AI, some contingent of the local community may be upset that didn’t hire a human designer or artist for that task. Online, I have even seen people get upset about entertainers reposting AI-created fan art— a task that no one would have otherwise paid someone to do: some see normalizing AI art in any context as a threat.

As always, the reputational risk is higher for organizations where labor force and economic impacts are part of their mission; an organization working on workforce development or job training is likely to face bigger backlash around replacement and deskilling impacts than a youth sports organization.

Replacement isn’t the only job market threat posed by AI. Consider these scenarios:

“My job used to be difficult to train for. To get it, I needed schooling, certification, licensure, and/or lots of experience at a lower level job. Now, AI can do all of the difficult parts of my job. We still need people to monitor it, but now you don’t need all the background that I have. What used to be a high-skill job is now a low-skill job. I am paid less, I have lower job security, and my job is boring. I feel underemployed, but there are very few jobs out there doing what I used to do.”

“I used to be great at my job. It was rewarding and I could do all of the tasks well. Three years ago, at the behest of my manager, I started having AI help me considerable portions of my work. I and my organization got very accustomed to my being able do deliver consistent results very quickly. But now, we have a much more involved project to do and my skills are rusty and out of date. I let AI take over my job, and I don’t know how to do it well any more!”

This is deskilling, where AI reduces the need for humans to know how to do the things that it normally does. These two vignettes illustrate the deskilling of a job and the deskilling of an individual.

The International Labour Organization estimates that most jobs have tasks with exposure to AI. Their estimates indicate that women work in more exposed jobs: the risk is more than double for women than men.

From an organization’s perspective, individual deskilling could lead to an entire department that is unprepared to take on a opportunities outside of their AI-dependent workflow or being unable to pivot quickly to avoid risks. For a deeper look into deskilling and its consequences, I highly recommend Cautionary Tale’s episode “Flying too High: AI and Air France Flight 447.”

Depending its your mission and circumstances, mission-driven organizations may want to consider:

• When outsourcing tasks to AI may harm workers especially when you are making policy and high-scale decisions.

• When hiring a human may reduce risks to your organization’s reputation especially in communities where replacement is a big cultural touchstone. Note that humans may also offer reputational and branding benefits as well by leveraging human superpowers, like authenticity (future blog post coming on this).

• When deskilling can decrease your ability to take on new opportunities or avoid risks. If needed, create opportunities to practice and update critical skills.

• Where and when you might be breaking the law by using AI instead of people. The EU AI Act, U.S. Executive Order 14110, and similar bills in Canada, Brazil, and several U.S. states regulate AI with regard to to worker protection, transparency, and human‑oversight duties. I am not a lawyer; please consider asking one if you are in an impacted domain or location!

• Creating a policy around preserving human judgment and jobs. A policy will not only help guide consistent decision-making across the organization, but also communicate to your current staff how AI impacts their job security. People do better work when they are not afraid of losing their jobs, The AI Use Policy Template in your free Mission-First AI Starter Kit is a great place to start!

I think there’s a lot more to say on this subject: what kinds of tasks are at risk of deskilling or replacement? What jobs are likely to be affected? As an individual worker, how do I avoid being impacted?

Let me know what else you are interested in!

—

LLM disclosure

Note: when I went to try this query the first time, Claude was down! What would I do if my job were reliant on it?

I asked ChatGPT o3 instead.

”I'm writing a blog post about the ethics of deskilling and replacement for workers in mission-driven organizations. I'll paste in what I have so far. I am interested in your ideas about the implications of this change for mission-driven organizations. What can they do do reduce this harm while staying competitive? Are there any ways in which this is harder or easier for mission-driven organizations over profit-drive [sic] ones?

Then:
”Can you create a 1:1 square image to accompany this blog post on linkedin and my website? “ BUT now ChatGPT is down! So I just used the Mission-Driven AI Starter kit; my default when I don’t have an image.