The right to refuse is a foundational right. Even when governments try to take it away, they actually cannot. In the end, if you refuse to comply, a government or other authority can penalize you, put you in jail or execute you, but they cannot make you say or do the thing you refuse.
Indeed, the right to refuse is the last right some individuals exercise; such was the case with Socrates, Sir Thomas More and Franz Jägerstätter. As we go deeper into the age of artificial intelligence, it is important we as citizens retain that right to refuse — without penalty.
While AI may be a revolutionary and helpful tool in some contexts, in other respects it offers a variety of possible existential threats. The threats of AI-facilitated bioweapons or terrorism are absolutely real, but there is another threat which is much more personal: the threat to our psyches, whether by that term one means our minds, our hearts or our souls.
One threat is that of de-skilling, or even never-skilling. The price of mastery in a given domain is deep experience with it. And that deep experience involves processing raw materials or raw data, learning to induce patterns from such raw inputs, failing and learning from that failure, learning anew why certain ways work better than others, facing ambiguity and uncertainty and finding a path through those to a good result. In other words, you need time, experience and friction for mastery to develop.
Unfortunately, the promise of AI is the minimization of all three. Mastery inevitably erodes with AI use, or it never develops in the first place, for humans are hard-wired to offload cognitive tasks when possible. In one study, when doctors first adopted AI for speedier diagnostics, they understood the need to check the AI’s work and correct it. A few months later, they couldn’t be bothered to check it, and instead cognitively deferred to it. When the AI assistance was removed, their diagnostic skills were worse than before they began using it. They’d been quickly de-skilled by AI use. Worse, students using AI are never-skilled; that is, AI assistance makes it unnecessary for students to develop skills at all — besides the skill to write a good prompt for their AI model.
This is an important turning point for humanity. Ours is possibly the last human generation with sufficient mastery to know for themselves whether an AI model is hallucinating or not, or aligned or not. And even that ability is swiftly eroding: Frontier AI companies are no longer capable of monitoring their models in real time and have had to turn that task over to AI monitors instead. Humanity is facing a situation where the outputs and processes of AI models will eventually not be “checkable” by humans in any meaningful sense. We will inexorably be forced to defer.
Consider this interesting case that came across my desk. The National Health Service of the U.K. commissioned a report on maternity services in that nation. They asked women who had received such services to comment and garnered more than 10,000 responses. Instead of reading the responses, and using induction to begin to parse the thematic elements, they turned the whole thing over to Microsoft’s Copilot to summarize. The commissioners would not read any of the raw response data, but would instead rely on the AI summary.
The outcome: Important themes were wholly missed or erased as a result — themes that were obvious to those not in government who bothered to read the raw responses. If the raw responses had not been available to others, no one in government would have ever known what the AI model missed. The commissioners had deferred, and so knew only what the AI told them.
This cognitive deference also has other psychological effects. Humans are hard-wired to anthropomorphize possible sources of threat in order to gauge intentionality. As we anthropomorphize AI models, while simultaneously deferring to them, we become vulnerable to AI-induced psychosis. AI is thus treated as a lover, as a priest, as a sage, as a god. It has bemused me to see that people who make their living as psychics are up in arms about AI psychics that have become very popular. AI is there 24/7 to tell us what we want to hear and to tell us what we should do. Some may see AI as more of a real presence in their lives than other human beings or God.
Cognizant of how humans are hard-wired, and desirous not to be de-skilled/never-skilled or beguiled by an anthropomorphic lie, some individuals have come to the perfectly reasonable conclusion that they must refuse on principle to use it. Indeed, Bill Gates of Microsoft expressed that some will want to emulate the Amish where AI is concerned — that is, to simply abstain from using this new technology.
However, the bigger question facing our society is not the refusal itself, but whether one can be penalized for it. Does U.S. law shield individuals who refuse to use AI?
Astonishingly, given how AI has been seen as an unstoppable juggernaut, the answer is yes. The names of Erin Maus and David Durman will be remembered for their successful pursuit of Title VII exemptions to using AI at their workplaces, for these exemptions are precedent-setting.
Title VII of the Civil Rights Act of 1964 prohibits employment discrimination based on race, color, religion, sex and national origin. Maus, a Unitarian Universalist, works as a software engineer, and Durman, a humanist (which is recognized as a religion under Title VII), works as a research administrator at the University of Kentucky. They each argued that forcing them to use AI discriminates against their religious beliefs and that they had the legal right to an exemption under Title VII from being forced to use it in their jobs.
In a letter he wrote to the University of Kentucky, Durman said, “Substituting automated machine processing for human presence, authentic listening, and personal responsibility in my work directly conflicts with my obligation to uphold human dignity and flourishing as primary.”
He further explained, “The process and work involved in thinking through situations, through my work, through my ethics, through the way I handle other people, is a key component in being able to develop myself as a moral person. And offloading that labor, regardless of whether it’s taxing or not, is undercutting my ability to be a moral actor.”
Durman hits at the heart of the matter. If using AI does indeed make Durman less capable of acting morally, according to his sincerely held beliefs, he should not be forced to use it, and he should not be penalized for that refusal.
And along the way, Durman may be helping all the rest of us. His and Maus’ refusal, and their successful insistence that U.S. law shields their refusals, are inspirational, and they establish an important foundation stone of a citizens’ inalienable bill of rights with regard to AI. Thank you, Erin Maus and David Durman, for sticking to your principles.