Employee Monitoring and Your New AI Supervisor

Meta (the parent company of Facebook), began to extensively monitor employee computer activities in April as part of training AI agents. After ongoing pushback from employees, they have recently implemented some constraints such as 30-minute surveillance breaks. The situation speaks to the complicated way artificial intelligence and digital surveillance can mix in the workplace.

Employee monitoring software tracks employee actions such as keystrokes and mouse movements to ensure the employee is working when and how their employer wants them to. The level of invasiveness varies, but almost 80% of US firms use some kind of employee monitoring software and 55% even monitor call/email tone and content. Some of this surveillance is comparatively mild, such as time and attendance logging. However, artificial intelligence enables monitoring at a scale and level of granularity previously impossible. Software makers often recommend employee monitoring for improvement, not punishment, but even if employers follow that advice, should employees be so extensively surveilled in the workplace? What might justify it?

One possibility would be a general allowance for employers to practice business as they see fit. We might think, all things being equal, that the government should not interfere in how employers run their businesses. After all, part of the aim of market-based economics is to unlock the benefits of differences in employer approach and management style. But all things are not equal. Granting employers the freedom to surveil comes at the expense of substantial employee freedom, respect, and privacy. And there are very real legal limits to employer power – employers are not free to beat their employees even if it is their preferred management practice.

Alternatively, it could be argued that overweening workplace surveillance is self-limiting. People either find such monitoring unacceptable and leave the workplace, or they tacitly consent to it. However, even if the problem generally takes care of itself, this does not justify allowing it. Almost any practice can be allowed under the argument that employees are always free to find a new job. It may still make sense to have legal standards of acceptable behavior. Moreover, the ability of employees to “vote with their feet” is often highly limited. A slow labor market, limited opportunities in an area, or widespread adoption of workplace monitoring in an industry can all prevent an employee from exercising their exit option. (I’ve previously discussed some of the challenges of consenting to working conditions.)

Another potential argument is that employee monitoring often occurs via employer property, such as employer-provided computers and employer-owned premises. First, it is notable how slippery consent is here. There is often no way for employees to not work from such materials, so they cannot meaningfully consent to these practices unless they do not want to work at all. Second, again, while one often can do what they want with their property, limits can be imposed to protect the rights and well-being of others. For example, one is not allowed to booby trap their house.

What about the ostensible benefits of employee monitoring? Presumably the primary benefit of such surveillance, and why employers engage in it, is to increase employee productivity. Should this factor into the ethical debate? This depends on our ethical approach.

Under ethical approaches that emphasize respect for people as such, as opposed to treating people instrumentally for their economic contributions, even the idea of sacrificing dignified treatment for productivity is ethically dubious. Other ethical approaches instead look at the over-all good that would result. From these perspectives, it is at least theoretically possible that so much good results from employee monitoring that it outweighs the harm to employees.

But, in fact, studies indicate the relationship between employee monitoring and employee productivity is complicated. Careful monitoring can increase employee output in some cases, but it can also damage trust, autonomy, and employee morale, ultimately decreasing productivity.

That said, the technology is still under development. If employee monitoring software reliably increased employee productivity, would it then be ethically justifiable?

Plausibly, with an increase in employee productivity, a company can produce its goods more cheaply, leading to some general public benefit. But this is highly idealized. In practice, negative externalities (such as pollution), harmful or addictive products (such as tobacco or, arguably, social media), and other factors, complicate the path from company efficiency to public benefit. All a successful product needs to do is satisfy the short-term desires of a particular customer base; it doesn’t need to lead to an overall increase in human well-being. In short, improving employee productivity is a compelling argument at a board of directors meeting, but of less clear ethical significance.

Moreover, it’s important to recall what’s on the other side of this discussion about economic efficiency — employee freedom and well-being. As a matter of practice, we do allow companies to behave somewhat autocratically, such as deciding how their employees dress and spend their time at work. But just like government infringements on freedoms, we should demand compelling justification for such restrictions. The more onerous the imposition from the employer, the stronger justification we should demand – workplace monitoring can be far more serious than having to wear a uniform. Depending on the invasiveness of the specific employee monitoring software, it can represent a serious infringement on employee privacy, autonomy, and dignity, resulting in deep concerns. If we value employee well-being, then workplace monitoring should be highly restricted unless carefully designed to preserve employee privacy and dignity. Employee productivity is an insufficiently compelling argument to overturn the prioritization of human dignity.