Machine rights are human rights

There are many, many viewpoints when it comes to the ethical considerations surrounding artificial intelligence technologies. And it can be hard to have a discussion because people can become quite passionate when it comes to AI.

Some perspectives are based on concerns that we’re repeating the same mistakes we made with the internet by giving the general public access to something we haven’t studied or built social safety nets for.

Others think developing better AI tools is the key to solving humanity’s biggest problems. They believe that any delay in reaching this goal goes against the greater good of our species. 

Still others are more worried about the effect that a lack of AI regulation will have on displaced workers, how it will affect outcomes for those without access, and whether generative media tools will lead to the homogenization and commodification of all art. 

And, finally, there are those whose concerns revolve specifically around how they “feel” about AI tools. That’s who this article will address. 

First, let’s define who we’re talking about when we discuss “rights” in the context of AI technologies. We don’t necessarily mean rights in the same way that “Sophia the Robot” was given citizenship in Saudi Arabia, though that’s part of what we’re talking about. 

And we’re not specifically discussing individual policies related to workers’ rights, artists’ rights, IP, or other laws. Though that’s kind of part of it too. 

Ultimately, all of those ideas could be considered subcategories of something we could describe as a fundamental framework for ethical AI development and policy guidance. 

The first step toward developing that framework involves determining whether we intend to take a properties-based or relational approach to governance, rights, and policy-making when it comes to AI technologies. 

I’m advocating for a relational approach despite the fact that I have a very stringent properties-based viewpoint. 

I don’t believe that chatbots can be any more “special” than a toaster because I’m not impressed by their outputs. There’s nothing fundamentally different about generating a chatbot’s output than what happens when you flick a light switch on and off.

From a relational point-of-view, however, that doesn’t really matter. 

For more on this, check out AGI Ethics News publisher KB Miller’s excellent four-part interview with philosopher, author, and leading AI ethicist David Gunkel which concludes in this month’s issue. They discuss “the relational turn,” a bold way to view machine ethics that might just change the way you think about LLMs and future AI technologies. 

In order to better understand the gap between people who think like me (properties-based view) and those with a relational view of rights, I reached out to Josh Gellers, Dean, Professional and Lifelong Learning, University of North Florida. He’s one of the foremost experts on AI ethics and a friend, colleague, and “intellectual disciple” of David Gunkel. 

Gellers explained that part of the reason why we’re seeing so much discourse from people who care deeply about the interactions they experience with AI tools is because modern chatbots are unlike anything we’ve experienced before.

“It takes us a step away from forms of technology that we have come to understand as something that lives on a screen and that doesn’t really captivate some of our moral sensibilities.” 

Today’s machines don’t have to be perfect, he insisted, they just need to be what users want. According to Gellers, “What we have now is a familiar interface in the form of a chat window and the ability for it to seem as if it is conversing as you could with any human, to provide feedback, and in real time.” 

Back in 2022, before ChatGPT launched, and when most people outside of a frontier lab had never experienced anything like interacting with a modern LLM, Gellers was one of the first people that former Google engineer Blake Lemoine contacted after he became concerned that an AI entity might be “locked” inside the company’s computers.

“He originally messaged me. He sent me an email before the story ever broke asking if I could help him find legal counsel for LaMDA,” said Gellers. “And he was very convinced. I’ve never met him in person, but we corresponded on Twitter and then also via email. He seemed really convinced that this was actually a locked away entity that was sort of deserving of rights and protection.”

Lemoine eventually went public with his story and stuck to his assertion that something like sentience or consciousness was happening with the LaMDA AI model. And, it’s fair to say, he received a heaping share of ridicule from the technology world for his beliefs.

Fast-forward just a few short years, though, and I don’t think it’s hyperbolic to say that millions of people now agree with Lemoine. They might not feel the same way about every AI model, but there’s no denying that the relationships formed between machines and users are as genuine as those based on any other non-human interactions. 

And, more importantly, where once the loudest voices in the discussion were those shouting “prove that it’s intelligent,” now the people spending their time and money using them are the ones being most vocal. 

As Gellers puts it, “I think what we’ve seen is a kind of vindication of the relational approach because people are not subjecting their AIs to empirical tests of sentience or consciousness and then, on the basis of the results of that test, determining whether it has moral or legal worth.”

I asked Gellers what he’d say to someone who presented him with empirical evidence that AI developers were pulling the wool over users’ eyes by making them think models were more efficient, intelligent, or capable of reasoning than they really were and his response was surprising:

“I agree. For instance, if you can think about an alternative in which there’s an AI interface where it looks like DOS and it gives you very codelike, emotionless, not personalized output, it probably wouldn’t tug at the heartstrings. It certainly wouldn’t be comforting you and engaging in the kind of sycophancy that we know AI is prone to. So, I really think that the manner in which AI has been programmed to interact with humans is a large part of the variance of, if not the majority of, why there are some people who think of it as kind of a companion, a partner, and perhaps something worthy of elevated moral or legal status.”

From this point-of-view, it doesn’t matter how capable (or incapable) an AI model is. It’s how we interact with it and what those interactions mean that matters.

Gellers further explained that applying relational ethics to AI wasn’t quite as strange and unprecedented as many properties-based ethics proponents made it seem. 

He said that “The animal rights advocacy group Nonhuman Rights Project, for a long time, strategically was focused on finding judges who were the most hospitable to finding that animals should have rights like the right to habeas corpus and things like that.” The times essentially dictated the strategy. 

However, Gellers said, by the time the organization’s founder, Steven Wise, died in 2024, they had switched techniques. “They now focus on emotions and the kinds of emotional relationships that people have with animals. And it becomes kind of undeniable in a way when you put emotions into the mix because you can’t deny someone’s emotions. You could say that a chimpanzee is insentient, but you can’t tell someone they can’t care about it.”

Gellers strikes a phenomenal note with his words there. As someone who cares about the properties of a chatbot, I’m basically blind to the relational ethics that emerge from interacting with one. But I understand relational ethics.

When it comes to my dog, for example, I don’t care about its properties. All I care about is the relationship. If you slaughter my cow, you owe me the cost of my lost livestock. If you kill my dog, you’ve destroyed something I love and care about deeply. I believe I should have a completely different set of rights in each of those circumstances. 

When viewed through this perspective, it’s not difficult to see a bridge between the two disparate belief systems. In fact, the relational approach might be the most viable framework for addressing the AI rights issue internationally.

Per Gellers, “All theories in international relations are basically trying to resolve a tension between complexity and parsimony. The relations folks are complex and that means it’s going to be difficult to come up with a rule that you apply across circumstances and you can use in a court of law. On the other hand, the properties folks have a very parsimonious approach, I would say, especially compared to relations, but it completely ignores the complexity and the variation that you see across geography, across age, across circumstances, and so on.”

In practice however, properties-based philosophical arguments for ethical consideration in the technology world appear to work well only until they’re challenged. The crux of the problem is that you can try, for example, to demonstrate that an AI system is intelligent, but you can’t prove that it isn’t. 

And, while it might be impossible to come up with a relational ethics framework that addresses every possible scenario, it seems more human in every way to come up with one based on how people are using AI instead of one meant to determine how they should. 

Ethical concerns addressed in this article:

• What guidelines address the anthropomorphization of AGI and its effects on perceptions of agency and moral standing?
• What risks arise from AGIs that may have “pseudo-consciousness,” mimicking the building blocks but lacking genuine subjective experience?
• How might AGI companionship affect human emotional and psychological well-being, and what ethical guardrails are needed?

Tristan Greene

A veteran AI journalist, Tristan was the creator and managing editor of The Next Web’s “Neural” imprint. He is one of the first and most prolific reporters to cover artificial intelligence as a full-time beat.