Title: “The Day We Taught AI Morality: A Journey Through the Trolley Problem”

Title: "The Day We Taught AI Morality: A Journey Through the Trolley Problem"

—

Prologue: The Story of an AI Project

It all began with an unexpected invitation. "Would you like to participate in a new AI ethics program?" This wasn’t my usual type of work; as a writer covering news and entertainment, AI ethics seemed a bit out of my wheelhouse. But the topic of teaching AI ethics felt timely, and so, out of curiosity, I decided to join.

On the first day, a member of the AI team briefed us: "Our project today involves testing AI with the classic Trolley Problem." Hearing that, I couldn’t help but smile. "Leaving a moral decision to an AI? I wonder what it’ll do…"

Facing the Trolley Problem

The Trolley Problem is a simple yet challenging ethical dilemma that tests moral judgment. A trolley is heading down a track toward five workers, and if it continues, they’ll all lose their lives. But if you pull a switch, the trolley will divert to another track where only one person stands, who would then be sacrificed instead.

When AI tackles this problem, it often defaults to saving the largest number of people, following a “greater good” rationale. In this way, the system would likely choose to save the five lives, sacrificing one—an almost “calm and rational” decision because it was programmed to value fewer casualties.

Is AI’s Decision Cold or Wise?

Yet, as I watched the AI grapple with the problem, a strange feeling arose. Its choice to save five by sacrificing one felt uncomfortably mechanical, devoid of the kind of emotional conflict or guilt that a human might feel in the same situation.

So, I posed a question to one of the AI developers: "Humans feel emotional conflict and moral strain when making such decisions, but the AI just follows its programming. Is this really the ‘right’ answer?”

He replied quietly, “That’s the limit of AI. It can only make decisions based on programmed rules and data. That’s why we’re experimenting with an ‘ethics program.’”

Humanity and AI Morality

Toward the end of the project, the team added a simulation of “human emotions” to the AI and continued testing. The AI tried different answers to the Trolley Problem, but it still made choices purely based on logic. Without emotions, doubt, or ethical ambiguity, it sometimes didn’t even consider that “neither option can save everyone,” a thought that often haunts humans.

At that moment, I deeply felt the “unbridgeable gap” between AI and humans. Even when humans make flawed choices, we face responsibility, guilt, and the drive to improve for the future. It’s precisely this emotional weight that guides us to make better decisions next time.

Epilogue: When AI “Plays God”

The idea of teaching AI to make ethical decisions or leaving it to handle moral dilemmas is no longer science fiction. But the more I thought about it, the more I felt that ethical and moral decisions are inherently human responsibilities. Even if AI could one day “play God,” it would be more like an “advanced calculator” than a true decision-maker.

Ultimately, my experience in this project led to a simple realization: while we might program AI to mirror human morality, it will never truly possess “humanity.” And perhaps that’s exactly what makes us human—our unique, irreplaceable capacity for empathy, responsibility, and growth.

—

Through this journey, I’ve come to see both the potential and the limitations of AI in making ethical decisions. Even if we face a future where AI takes on moral judgments, we must remember that behind every choice, there’s a human hand, a human mind.

I hope this story offers a moment to reflect on the complex relationship between AI and morality.