Generative AI and Moral Injury

What if your job asked you to use amphetamines to be productive?

Source: Ewa Urban/Pixabay

Consider a hypothetical scenario: A company asks its employees to take amphetamines in order to increase their productivity. The company makes the case that the drugs can be taken in safe doses that won’t have adverse side effects. Plus, the drugs will make the employees work faster, help them concentrate better, and make them more creative and efficient.

Most of us would likely balk at the idea. The employees might reasonably doubt that the dosing is actually safe or that it won’t have long-term effects, despite the company’s claims. Even if the dosing was safe, employees could object that the company shouldn’t ask them to take mind-altering drugs in the name of efficiency. They might also object that using drugs is wrong or that it’s against their values or principles.

The arguments that the hypothetical company makes in favor of amphetamines are similar to arguments they might make in favor of using generative AI. Generative AI can increase productivity by decreasing the amount of time that employees spend on certain tasks. It can help them focus on other things that are more important. It can help jumpstart their creativity and make them more efficient. While we might think it’s fine to object to amphetamines, we might be less likely to object to generative AI. Why?

Some people might deny that drugs and generative AI have that much in common. Drugs alter your brain chemistry and have potentially harmful physical side effects; AI is just software. And yet there is research showing that generative AI also alters your brain: Several studies suggest that using LLMs weaken our cognitive abilities—what is sometimes called "cognitive debt” or “cognitive offloading.” There is also research showing that generative AI use worsens the negative effects of increased screen time, such as isolation, anxiety, and depression.

People might argue that generative AI is just another tool, similar to spreadsheets or email, but that’s not how it’s marketed or how it’s presented by employers. The rhetoric around generative AI is all about how it supercharges employees to be more efficient. But, as the hypothetical case of microdosing amphetamines shows, there can and should be limits to how much we demand that workers do what they can to be more efficient. What if employees have objections to using generative AI tools? Do they feel empowered to resist?

Psychologist Jonathan Shay introduced the concept of moral injury based on his work with military veterans. Moral injury occurs when members of the military are put in situations in which they have to betray their moral values or principles in order to fulfill their professional duties. It would be wrong to compare the moral injury that arises from combat trauma to the moral qualms that people might have using generative AI. The point, though, is that companies might be asking employees to betray their values if they mandate or even encourage generative AI.

People might object to AI tools on a variety of grounds—maybe they are concerned about environmental issues, threats to privacy, or incurring cognitive debt. While we can disagree about how serious these concerns are, they aren’t imagined. Even if microdosing amphetamines was truly safe, employees might reasonably object to taking them just to increase efficiency. They may have similar qualms about generative AI, but it’s not clear that employers are willing to listen.