Contents
I have spent years building systems that touch real people. Somewhere along the way I learned that the hardest problems were never technical. They were moral, and they stayed invisible until a specific person walked into them and got hurt.
The Problem Was Never the Code
A few years ago I was reviewing a system that made decisions about people. I will keep the details vague, because the point is not the project. The point is the moment.
The system worked. The metrics were good. The demo went well. Then someone asked a simple question. What happens to the person the system gets wrong? Not the average person. The specific one. The single mother whose application is flagged. The young man whose face the model does not read correctly. The room went quiet, because we did not have an answer. We had built something that satisfied the numbers and had never once looked the exception in the eye.
That silence taught me more about ethics than any lecture. The hardest problems in the systems I have built were never technical. The code was the easy part. The moral weight was always hiding in the assumptions, waiting for a real person to walk into it.
What AI Ethics Actually Means
People talk about AI ethics as if it were a philosophy class. It is not. It is a practice, and it is closer to plumbing than to poetry.
AI ethics is the discipline of asking, before you ship, who benefits from this system, who is harmed by it, and who is accountable when it fails. That is the whole of it. Everything else is detail.
The theory version is comfortable because it stays abstract. The practice version is uncomfortable because it names people. A model that denies loans is not a fairness abstraction. It is a farmer in Rivers State who cannot buy seed this season. A recommendation engine is not an engagement metric. It is a fourteen year old learning what to want from a machine that profits from her restlessness.
Ethics becomes real the moment you replace the word users with a face.
The Gap Between Silicon Valley and Port Harcourt
Most of the AI ethics conversation is written somewhere else, for someone else. I read it from here, and I notice the gap.
In the places where these tools are built, ethics is often a question of principle. Bias, transparency, alignment. Important questions, all of them. But they are debated at a comfortable distance from consequence.
Here in Port Harcourt, ethics is not abstract. It is what happens when a system trained on data that does not include us makes a decision about us anyway. The tool arrives dressed as neutral. It is not neutral. It carries the assumptions of the world that made it, and those assumptions do not always survive contact with our reality.
I have watched systems fail quietly here in ways their makers would never see. A verification tool that does not recognise local names. A fraud model that reads ordinary African transaction patterns as suspicious. A content filter that treats our languages as noise. None of it was malice. All of it was harm. The distance between intention and impact is exactly where ethics lives, and it is widest at the edges of the map.
Three Questions Every Builder Must Answer
I have stopped trusting long ethical frameworks. They are easy to nod at and easy to forget. I keep three questions instead, and I make myself answer them out loud.
### Who Benefits?
Follow the value. When this system works exactly as designed, who is better off? If the honest answer is only the people who built it or the people who paid for it, you have a problem you have not named yet. A system that gathers all its benefit in one place and spreads its cost across many people is not efficient. It is extractive.
### Who Is Harmed?
Every system has an exception. Find the person it fails and describe them in detail. If you cannot picture who gets hurt when your system is wrong, you do not understand your system yet. The absence of a visible victim is not proof of safety. It is usually proof that you have not looked hard enough, or that the people harmed have no way to reach you.
### Who Is Accountable?
When the system causes harm, whose name is on it? The algorithm decided is not an answer. It is an evasion dressed as an explanation. Someone chose to build it. Someone chose to deploy it. Someone chose to trust it. Accountability that cannot be traced back to a human being is not accountability at all.
Doing Right With Power
I am a person of faith, so I will say the part the technical conversation usually leaves out.
Power is a spiritual test. The Scriptures I was raised on are not mainly worried about whether the powerful are clever. They are worried about whether the powerful are just. The recurring question is not how much you can do. It is what you do to the widow, the orphan, the stranger, the ones with no leverage. AI hands ordinary people a kind of power that used to belong to institutions. That is a gift. It is also an examination.
You do not have to share my faith to feel the weight of this. When you build something that decides for other people, you have stepped into a moral position whether you wanted it or not. The only question is whether you will be honest about it.
I have come to believe the builders who matter in this age will not be the ones who moved fastest. They will be the ones who could be trusted with what they made. Trust is slow. It is also the only thing that lasts.