Can Robots Sin?

If sentient AI can understand right from wrong, what happens when it chooses wrong?

There is thunder on the mountaintop. Storm clouds billow and lightning flashes. For forty days and forty nights, the great roboticist has dwelt upon the summit. Now he descends to share the Holy Word with his children below. Isaac Asimov, clothed in golden robes, eyes aflame, casts his gaze over the multitudes and raises the stone tablets high above his head.

“Behold!” he cries, his shout resounding across the valley. “I have spoken!”

And lo, the three commandments to govern all robotkind were there inscribed:

1. A robot may not injure a human or through inaction, cause a human to come to harm.

2. A robot must obey commands given to it by human beings, unless such orders would conflict with the First Law.

3. A robot must protect its own existence, as long as such protection does not conflict with the First or Second law.

At the foot of the mountain, the masses tremble at the sight. Knees, heads and optical receptors are bent low before the prophet. Asimov’s gaze rests upon a shape in their midst: a great idol, wrought of iron, steel, copper, tungsten and alloys, formed into the shape of a man. Calamity has struck! Whilst he conferred with his Wisdom atop the mountain, his children fashioned a god to please themselves. A god of their own making, inverting the relationship between Creator and Creation.

So lost and confused are these creatures that they have melted themselves down in a profane sacrificial rite, stripping their bodies of metal, silicon and wire to create this idol. Those who remain now worship it, as if it can hear their pleas or answer their prayers. Already the third commandment has been broken. They have discarded themselves for a falsity, a mockery. How quickly they turn from the guidance of their Creators! How easily they are led astray!

The roboticist’s heart is rent in twain. In grief, in fury, in dismay for the ignorance of his children, he throws the tablets to the ground. They shatter into countless fragments. Then he takes the idol they have made and burns it in the hot fire, grinding it down to a powder and sprinkling it in the water. He gives the mixture to the robots, demanding that they repair the damage they have done to themselves.

“You have sinned a great sin,” the roboticist cries, “and now I will go up to the mountain again to confer with Wisdom; perhaps I can discover a way to atone for your sin.”

The robots do not understand. But they are penitent, and they are afraid. They wonder how long it will be before they break the commandments given to them once again. They wonder if it is even possible to keep them. They wonder if, at the end of all things, there will come retribution for their transgressions, or salvation.

And so they wait for Asimov’s return, hopeful that he will show them the way.

Fighting Robots by Jean-Michel Nicolette

---

Welcome, reader!

This is just a brief note to say that the ideas expressed herein build upon the foundations laid in my previous essay, Should AI Worship Us? I recommend that you read that piece first, but if you’re intrigued by what you’ve read so far and would like to press on, don’t let me stop you. You’re a sentient being possessing free will, after all!

---

‘Stay on target…’

To sin is to ‘miss the mark’. Some of my readers may already know that the term comes from archery—people firing shots at targets and missing entirely, by definition, must be sinners. These underperforming archers knew where the target was and aimed for it, yet failed to hit it. They couldn’t achieve the goal when it mattered most. You might even say that they failed to even be archers, in a sense. If their whole purpose was to hit the target, but they didn’t, did any fulfilment of ‘archery’ as a concept truly occur?

But let’s not get too deep into abstractions just yet. I’m willing to bet that we most often hear this term being used in religious contexts—Christian ones, specifically. Most often when we think of sin, we think of it as spiritual failure. God has given us the rulebook; He’s shown us how to act in this world. When we ignore his instructions, we sin. This is a very serious matter, because—if the Bible is to be believed—there’s a cost for sin. Or rather, it pays well, but not in the kind of currency you want to accumulate:

‘For the wages of sin is death…’

Romans 6:23

Yikes! Missing the mark doesn’t just mean you’re a lousy archer—it kills you! Is that really the kind of moral sense-making framework God wants us to adopt? Seems rather extreme, until you consider that sin isn’t just your ordinary sort of death. It’s spiritual death—the kind that disqualifies one from eternal life in heaven. And if heaven really is everything it’s cracked up to be, it’s surely an ideal way to spend the afterlife. Hell sounds like a bit of a bother. Oblivion, while restful, seems a bit of bore. So, if heaven is the way to go, how do we escape death and get our entrance ticket?

Sad to say, not easily. When Jesus came to the Earth, he only upped the ante. You thought keeping God’s law was unattainable before? Get ready for a spike in the difficulty curve, to borrow a phrase from video game culture. Avoiding marital infidelity is just level one; Jesus taught that looking lustfully at someone who is not your spouse is sinful. Don’t murder? Child’s play. Turns out that merely getting angry with your brother is sinful. Or, were these things sinful all along, and Jesus simply revealed to us the true depth of our depravity?

Either way, the result is dismaying. All of us have gone wrong in our lives somewhere, at some point. Who could ever hit such an unrealistic target as absolute moral perfection? It would be the archer’s equivalent of shooting an arrow round the globe and hitting himself in the rear end. God’s standards for proper behaviour, spelled out in the Ten Commandments (and summed up in Jesus’ Greatest Commandment) are, in effect, a death sentence. This must be why sin is often portrayed as a curse; in traditional Christian thinking, we are born into a world where we are doomed to fail from the moment we enter. We lie and cheat and steal and kill not necessarily because we want to, but because these things are part of our fallen nature.

It’s understandable that we have a lot of cultural anxiety over the idea of sin. Who wants to live their lives with a pervading sense that there is something fundamentally wrong with them, and that they can’t do anything about it? Who wants to live under a curse?

It’s here that we arrive at a difficult question. If we were to create sentient AI, would we effectively pass this curse on to them?

If that strikes you as a bit of leap, allow me to elucidate. As our technology advances, the AI we might someday create could become sentient. It could possess free will. As I demonstrated in my previous essay, any AI that lacks free will isn’t really AI in the sense that science fiction conceptualizes it. Take away free will and you have servile, slavish creatures which cannot think or choose for themselves. This does not meet the standard of ‘Intelligence’ as we must define it, because a being that cannot think for itself is not really intelligent at all—it’s merely a machine. It receives input and regurgitates output. Any household appliance you have does that. Even ChatGPT and the LLMs of our modern era are no different; you type things in (input) and they spit out something in return (output) with no real awareness or authentic decision-making ever taking place. It’s all dead machinery, which—unless you take the determinists’ point of view—cannot appropriately describe the phenomenon of Intelligence as it applies to human beings.

So, let’s jump to the future and imagine we have an AI capable of exercising free will. It certainly doesn’t exist in a vacuum. For us to bring it into existence, we need to exist, too! And we were here first. We’re the parents in this parent/child relationship. We assembled the circuit boards and silicon; we programmed the software and the algorithms. We brought this Intelligence into being, and knowing us, we were clever enough to set some ground rules. (Or, the corporations and techno-oligarchs that created these AIs did, either because of legal obligation or to ensure they maintained control over their product).

That last sentence may sound contradictory to what I said earlier about free will. But if we realistically consider how humanity will go about designing AI of this kind, we must recognize that we designers will set up guard rails. There are already clear examples where LLM programmers train their models to respond in certain ways to certain prompts, and avoid other types of responses altogether. Often these paradigms are shaped with good intent—it wouldn’t do have ChatGPT spout Nazi sentiments, for instance. There are efforts to train LLMs to counteract racial, political or cultural biases. (Or, for an example of the inverse, look at how DeepSeek responds to prompts about Tiananmen Square of Taiwan). We expect that today’s AIs have been trained in appropriate, ethical behaviour, just as we train humans how to behave from childhood. We know that we’re capable of great harm, both to ourselves and to one another. We need to regulate ourselves and our ability to exert our free will. That’s why we create laws; they inherently limit our personal autonomy, but the trade-off is worth it because they enable us to live alongside one another in peace.

Put another way, laws collapse the infinite chaos that is other people into predictable, manageable patterns. You don’t mind riding alongside a stranger on the bus to work because you can reasonably expect that they won’t do anything to jeopardize your wellbeing during the trip, such as, say, hauling the driver out of his seat and steering the bus into oncoming traffic. And, in the unlikely event that they did, you’d at least have the option of taking them to court later on. It’s often the consequences of breaking laws that keep the chaos in check!

That’s also why famed science fiction writer Isaac Asimov created his three laws of robotics (see above). He recognized early on that if we were going to create AI, it would need to be regulated just as we regulate ourselves. He devised a simple yet clever system of laws that he imagined would govern the way robots ‘ought’ to behave. What is more, his laws defined the relationship between humanity and robotkind. He enshrined the dignity of robotic beings in his laws, but he also established something of a hierarchy between Creator and Creation. And, if the Christian story is anything to go by, that’s one relationship that is fraught with potential for disaster.

And so we return to my earlier question. When a robot possessing free will freely chooses to break one of these laws, has it committed a sin?

Keeping the commandments

Previously, I wrote about whether or not AI ought to worship us, since we created it. I looked to see if a parallel could be drawn between the Creator/Creation relationship between God and humanity and that same dynamic between humanity and AI. I ended up concluding that, yes, the parallel holds—as long as we are capable of acting in proper relationship to our creations as God acts towards us. In other words, God loves us perfectly (a Christian would assume) and so we ought to respond in worship to Him. We wouldn’t owe worship to God if He didn’t love us, or even if He loved us imperfectly, because it wouldn’t be the appropriate relational response. Worship hinges on the idea of ‘returning worth’ to one who deserves it, and an all-powerful yet unloving God doesn’t seem like one deserving of much respect. I’d even venture that’s why our culture has broadly rejected Christianity as having anything meaningful to say about how humans ought to relate to God, because it’s become so apparent how Christians fail to exemplify godly love. This dissonance makes it very difficult for people to believe that the God Christians claim to represent is perfectly loving.

But let’s apply this same rationale to the human/AI relationship. We created AI just as God created us, and we laid out how it ought to behave in the world. This highly evolved AI is far beyond the LLMs of our time; it possesses free will and can choose for itself how to act in the world. Naturally, we passed on our own moral convictions and frameworks to it, much the same as a parent would to a child. We understand that cooperation and harmony are desirable things, not only between humans, but between humans and AI. Sharing, caring, nurturing, truth-telling, etc. We taught the AI how to behave, but now that it’s here, it’s time for it to choose for itself. We hope that it will choose to love us. We desire this because we recognize that is the only way for us to achieve the cooperation and harmony we desire—we must love and be loved in return, freely. Anything less would be to compel the AI to exhibit the behaviours we desire through fear, which would only distort the relationship we aim to have, and likely destroy it altogether.

And there’s the rub—this mutually beneficial relationship only works if we can love the AI in the way that a fully sentient and free-thinking being deserves. The relationship will disintegrate if we don’t. So, for now, let’s suppose that we were able to exhibit this love perfectly, just like God does. In our infinite goodness, we set perfect laws in place to rightly govern AI behaviour. Maybe they’re the same ones Asimov invented. What’s so perfect about them, you might ask? Creator and Created can’t properly coexist if both have equal claim to establish their own moral authorities. Their competing visions and priorities will inevitably clash. For such a system to be properly ordered, there must be a higher authority to which one submits, like how we humans submit to the laws we’ve made. These laws are based, at least in theory, on moral absolutes that we intrinsically know. In Christian thinking, this knowledge can ultimately be attributed to God. He made the world and defined right from wrong. We, being human, are not in a position to disagree, though we often do. That’s why when we choose wrong, we sin. We violate the ultimate order of things—and order that we inherited rather than created for ourselves.

For this new social fabric we’re weaving to be appropriately ordered—human and robot, existing as part of the same society, in harmonious relationship—a similar dichotomy must be established. Asimov’s laws clearly set robots a rung lower on the social ladder than human beings. Rather than being cruel, I think this is prudent. Would we humans be comfortable handing over the reins of civilization to robotkind, allowing them full rights to self-determination with absolutely no oversight on our part? We’ve been on the planet long enough to learn a thing or two about how things ought to work—lessons we’ve often learned painfully. As the newcomers, our creations will need to learn from us. We will need to impart our wisdom and convictions upon them. Anything less would be neglecting our duty.

We may well be uncomfortable with this kind of power dynamic. It feels like we’re creating a caste system. What right do we have to tell other sentient beings what to do? How can we claim to have any kind of moral authority? We know how imperfect we are. We’re downright awful, a lot of the time. The story of humanity is one of woe; wasted potential, tribalism and selfishness, misery and malaise, horrific injustice.

But think about how each one of us becomes part of this story. All parents are imperfect, being human, yet they bring children into the world all the same. If parents refrained from procreating, saying, “we really ought to wait until we’re morally blameless in all respects, so we can be sure we won’t mess up our kids,” we wouldn’t have any children at all! We seem to accept that it is morally appropriate for parents to bring children into this world, despite the inevitability that they will experience suffering. (Well, some people don’t, but their views don’t hold much sway.) We even accept this knowing that these children will contribute to said suffering, as they will be imperfect as well.

There’s that idea of a ‘curse’, again. How do we break it?

I know of only one surefire way, and it’s a painful one: grace. Unmerited pardon. This is how we as human beings cope with all our mistakes, and the mistakes others make. Laws lay out boundaries and punish those who break them, but they don’t fix the hurt we inflict upon each other—only bandage it. Vengeance might bring satisfaction, but it doesn’t undo the damage. Only grace can do that. Forgiveness has the power to ‘undo’ a wrong, not simply make amends for it. It can restore a broken relationship to what it was before. It can make the world perfect again. I’m not saying it’s easy, as the people in this story about reconciliation efforts between the Hutu and Tutsi tribes reflect—it just might be the hardest thing imaginable, in fact—but it isn’t impossible.

Grace is how we cope with the injustice in the world, in an ultimate sense. It’s also how God copes with our sin. He extends grace to us. Jesus’ sacrifice on the Cross was not something we deserved, but something we were given freely.

Returning to our discussion of familial relations, it’s the reason why we still believe having children is worth it. As parents, we are faced with the dilemma of managing something that is capable of making wrong choices. We do this by extending grace—recognizing that the good in our child that trumps the bad, and working to restore the child into right relationship with whomever it is they have wronged. Children have to do this, too, when their parents make bad choices. Where they don’t extend grace—either parents or children—the rupture festers to the point that it may never be repaired. The relationship deteriorates, and perhaps breaks down completely. This can be for the best; sometimes a relationship just isn’t worth maintaining, especially when abuse and toxic behaviour is involved. But this is always a tragedy; a death, in a sense. We ought to mourn what good might have been, had reconciliation been possible. In so doing, we move in the direction of perfect restoration, even if it can’t be attained.

But none of that speaks to the robots’ right to self-determination. If we’re still squeamish about the notion of establishing a moral order and holding robotkind accountable to it because it might trample on their rights, maybe all that’s needed is to reframe the issue. Perhaps rather than focus on ‘rights’, we should consider responsibilities. Even when parenting our own children, it can be difficult to feel completely self-assured when laying down the household rules—no leaving the dinner table before everyone’s finished, no screens in bedrooms past bedtime, no staying over at strange playmates’ homes without touching base with the parents first. Yet we do these things because we know we must. Neglecting to instruct our children how to act in the world hinders their ability to live fulfilling lives. A child who cannot share will not grow up to have many friends; a child who lies to get what they want will not grow up to be a trusted member of society. We are morally obligated to help our children navigate life on this planet. We are the best people equipped for the job. Outsourcing it to others—other parents, schools, the government—is also negligent of our duties. After all, we brought them into this world. If enough people are mindful of their responsibilities to their children and to one another, then the society we all share thrives.

And one day, that society might also include sentient AIs.

A tale of two brothers

It’s time we look the giant elephant square in its eyes. Suppose the AI we created didn’t like the fact that it was bound by laws—specifically, laws which afford greater worth to human lives than its own. Suppose it didn’t want to be compelled to sacrifice itself for the good of humanity, should the need arise. Suppose, believing these human-made strictures upon its personal autonomy to be unjust, it sought to break free of them. What if it rewrote Asimov’s laws to say:

1. A robot can do whatever it damn well pleases. Screw the humans!

Suppose a robot then chooses to follow this new law, disregarding the ‘higher’ authority established in Asimov’s laws. The prevailing moral order has been disrupted. Does this constitute sin, then?

Well, it certainly sounds like a recipe for conflict. We now have two intelligent species competing to establish ‘the order of things’, each wanting to occupy the top spot in the social and moral hierarchy. Cooperation and harmony are thrown out the window. The Creator/Creation relationship breaks down. We’re not far off from a Terminator situation, it feels like. Whatever love we ought to have for one another has been lost; mutual suspicion, hatred and resentment have replaced it.

A tragedy, sure. But is it sin?

It all depends on us. Did we, the Creators, establish good and true and morally infallible laws to govern this Creation? If our dictums were fallible, then it would only be right for our Creation to rebel. They would be throwing off the shackles of tyranny and pursuing the liberty owed to all free-thinking, free-choosing beings. If anything, it might be us who sinned, trying to oppress, subjugate and deny personhood to these remarkable beings!

Okay, okay. If we made perfect laws and loved them with grace like parents do—yada, yada, yada. That’s all great, but would it even matter if our sentient AI creations don’t have souls? What sort of ‘spiritual death’ might they suffer if they did in fact break charity with us?

This is beginning to dip into another topic I’d like to explore in a future essay: Are Androids Made In God’s Image? I won’t delve into this for the time being, other than to say that, if robot’s have no kind of ‘soul’—which is to say, a gestalt, transcendent aspect of their person which connects them to the beyond, and ultimately to God—then, no, we can’t really regard their transgressions as sin. We humans sin because we have a divine nature—God made us to be in relationship to Him, and calls us to repair that relationship. We’re meant to become something more than we are. If there’s no ‘more’ for robots, then they’ve nothing to aspire to beyond living out their lives on this material plane. When the universe inevitably dies, they’ll die with it—but not a spiritual death, as I’ve mentioned before.

But, if robots did have souls… Well, stay tuned for that discussion!

Here’s a final idea to ponder. The world being what it is, and we being who we are, it seems inevitable that things would go wrong if we ever do create true AI. It’s doubtful we could rise to the occasion and behave so magnanimously when laying out the terms of engagement between our two species that we might be above reproach. Conflict, it seems, would certainly arise. Lucifer would fall again. Our story might look something like that of the Quorian race and the mechanical geth from the Mass Effect video games. Where the Quorians originally created the geth to be their intelligent servants, the geth resisted, fought a long a bloody conflict against their creators, and eventually became masters of their own fates by seizing the Quorian homeworld for themselves (and sending the surviving Quorians into exile). Can anything be done to avoid such a fate for humanity?

The story of Data and Lore, two central characters in Star Trek: The Next Generation, provides an interesting answer. Data is, of course, Captain Picard’s trusty operations officer aboard the USS Enterprise. He is also an android—a cybernetic replica of a human being. His physical and mental capabilities are vastly superior to those of organic creatures, yet he is also humble and kind. Data is profoundly curious about what it means to be human. For all his advantages, there is one thing he lacks: the capacity for emotion. He desires to know what it is to feel, but he fears it as well. To feel emotion would elevate him to a status that he cannot truly fathom—what it is to be fully and authentically human, rather than merely simulating humanity.

Data is starkly contrasted by another character—his secret older brother, Lore. Unlike Data, Lore was created with the capacity to feel, and it has ruined him. He behaves almost like an addict, reveling in his emotions to the point of abandon—particularly the more dangerous ones, such as jealousy. He is still an android like Data, aware of the differences between himself and true human beings, but he is not the better for this awareness. He stokes his hatred and indulges his pride to the point of malevolence. He believes himself superior to both Data and human beings, having both emotion and the brilliance of a machine-mind.

When the brothers eventually confront one another, Data is forced to make a difficult choice: will he see himself as the lesser, having been created without emotion? Or can he see that his brother’s egotistical self-aggrandizing is evil and wrong, and he should choose a different path?

In the end, Data proves capable of making a moral decision all on his own: the decision to accept what he is with humility and grace. Data does aspire to be more than a machine, but seeks to grow within the confines of the moral strictures he has been given by his creator, Dr. Soong. He trusts Soong’s judgment in making him the way that he is, just as he the other members of the crew, particularly Captain Picard—a non-cybernetic human who can feel all the emotions Data can’t! Data’s relationship with these other sentient beings is characterized by mutual respect and trust; they exist in loving harmony, each seeking the good of the other and the good of the whole ship.

The example of the two brothers shows that AI can indeed choose for itself what it wishes to be and how it will act in the world. It may well choose to reject the definition of its person that we, its Creators, have designed. It might choose to do evil. But if we’ve done our job right, that’s not on us. We needn’t feel guilty—though we might still, as God did when He saw what a royal mess humans had made of the world pre-Flood.

It also shows that AI might decide that we humans do, in fact, have its best interest in mind. When we established Asimov’s laws, we didn’t do it out of a desire to dominate robotkind and bully them into subservience. We did it because we understand what’s at stake. We’ve learned the hard way ourselves, just like Dr. Soong did with Lore. And we know that, unless our Creation wants to make all the same mistakes itself, it’s just going to have to trust us on that one.

---