My Grandfather Was a Schoolteacher in Maharashtra. His Values Built My AI Ethics Framework.

My Grandfather Was a Schoolteacher in Maharashtra. His Values Built My AI Ethics Framework.

My grandfather walked to the same school for forty years.

Not the kind of teacher who gets written about in newspapers. Not the kind who gets invited to education summits or quoted in policy documents. The kind who wakes up at the same hour every morning, walks the same road, teaches the same subjects with the same patience to a new batch of children who will one day grow up and tell their own children about him. The kind who, without a single published paper or public platform, manages to shape six children who all carry the same four ideas in their bones.

He was called Pujari Guruji by the people who knew him. Not because he was a priest. Because he carried the kind of moral seriousness that small town Maharashtra once placed on its teachers. The schoolteacher was not just an educator. He was a reference point. When there was a dispute in the neighbourhood, people came to him. Not because he had authority, but because they trusted his judgment. And they trusted his judgment because it was consistent. It did not bend for convenience. It did not adjust based on who was asking.

I did not fully understand the weight of that until I started building AI systems.

Let me tell you about the four ideas.

My grandfather lived by them. My mother and her siblings grew up retelling them the way other families retell legends. Not as commandments carved in stone, but as the quiet standard by which you measured whether you had lived the day properly.

The first was truth. Not truth in the philosophical, abstract sense. Truth in the sense of: say what you know. Do not pretend you know what you do not. Do not present guesswork as certainty. When you are wrong, say so. When you are uncertain, say that too. Truth was not a lofty ideal in our household. It was a daily practice.

The second was fairness. Treat people equitably. Not identically, because my grandfather understood, long before any textbook taught me the word, that equality and equity are not the same thing. He treated his students differently based on what each one needed. But the standard, the respect, the investment of attention, was the same for every child. No matter whose son or daughter they were. No matter what surname they carried.

The third was humility, which in our family expressed itself as something very specific: do not think you are above anyone. My grandmother was cut from the same cloth. She served people unconditionally, no calculation of what she would get back, no mental ledger of favours given and owed. The ethic was not transactional. It was dispositional. You served because that is what a good person does, not because of what it gets you.

And the fourth was the one that took me the longest to understand. Check your own biases before you judge someone else’s. My grandfather believed that the first person you audit is yourself. Before you evaluate another person’s character, another community’s worth, another student’s potential, you look inward and ask: what am I carrying into this judgment that does not belong there?

Truth. Fairness. Humility. And the discipline to check your own blind spots.

I carried those ideas through school, through university, through the early years of my career without giving them much explicit thought. They were just how we did things. Background values. The operating system you do not notice until something crashes.

The crash happened when I started working with AI.

I remember the first time I looked at a training dataset and realised what it was doing. Not in the abstract. Not as a concept I had read about. In practice. A model trained on historical data was learning the patterns of the past and presenting them as the logic of the future. And the patterns of the past, in India, carry caste, carry class, carry gender, carry the entire architecture of who was given opportunity and who was not.

The model did not know it was doing this. It has no concept of fairness or unfairness. It is a machine. It finds patterns and it optimises for them. But the output, the decisions it makes, carries every inequality that its inputs carried, amplified by the mathematics of pattern recognition and scaled by the reach of the system.

And I heard my grandfather’s voice. Not literally. But the four ideas were suddenly not background anymore. They were front and centre, urgent, and I did not have a way to apply them.

Because here is the problem. My grandfather’s values were designed for human relationships. For a classroom. For a neighbourhood. For the kind of decision making where you look someone in the eye and you know, in your gut, whether you are being fair.

AI does not have a gut. AI does not look anyone in the eye. AI makes decisions about millions of people simultaneously, and the person who built the model is usually nowhere near the person who bears the consequences of its output.

How do you take values designed for human scale decisions and make them work at machine scale?

That question consumed me. It consumed me through two years of building AI systems, of watching ethics frameworks pile up on desks without changing how decisions were actually made, of seeing the same failures repeated across industries and continents.

And eventually, I found an unexpected bridge.

In 1932, a man named Herbert J. Taylor was handed a bankrupt cookware company in Chicago. Two hundred and fifty employees were about to lose their jobs. The company owed more than it owned. Taylor looked at the company’s existing ethics code. It was long, detailed, and in his own words, “almost impossible to memorize and therefore impractical.”

So Taylor did something extraordinary. He sat at his desk and wrote twenty four words. Four questions.

Is it the truth? Is it fair to all concerned? Will it build goodwill and better friendships? Will it be beneficial to all concerned?

He put that card under the glass top of his desk and applied those four questions to every business decision the company made. The company survived. It paid off all its debts. Rotary International adopted the test in 1943. Today, it is translated into over a hundred languages.

When I first encountered Taylor’s Four Way Test, I felt something click. Not because it was new. Because it was old. Because the values underneath it, truth, fairness, goodwill, benefit, were the same values my grandfather had carried to a classroom in Maharashtra for forty years. The same values my grandmother had lived through her unconditional service. The same values two hundred modern AI ethics frameworks are trying to articulate with varying degrees of jargon and varying degrees of success.

The values are universal. What Taylor proved, and what I needed for my work, was that you can make them operational. You can turn them into a test that a person under pressure, facing a real decision, can actually use.

That is what I built.

I took my grandfather’s four ideas. I took Taylor’s test. And I built a framework for AI ethics that is simple enough to remember, rigorous enough to apply, and specific enough to change actual decisions at nine in the morning on a Tuesday.

Truth became a transparency requirement. Can you explain what this AI system does, how it was trained, and what its known limitations are? If you cannot, you have not met the first test.

Fairness became a bias audit. Have you tested this system against the populations it will affect? Have you checked for proxy variables? Have you looked at whether the system’s errors fall disproportionately on vulnerable people? If you have not, you have not met the second test.

Humility, the one about not thinking you are above anyone, became a goodwill principle. Does this system serve the people it affects, or does it serve only the people who built it? Does the benefit flow in one direction, from users to platforms, from citizens to governments? Or is it genuinely mutual?

And the fourth, checking your own biases, became accountability. Have you built mechanisms to catch your own mistakes? Do affected people have a way to challenge the system’s decisions? Is there a human in the loop when the stakes are high?

Four principles. Four questions. Any AI decision.

I want to be honest about something. When I tell people the framework is rooted in my grandfather’s values, I sometimes see a flicker of scepticism. The technology industry expects frameworks to come from labs, from published papers, from institutions with three letter acronyms. Not from a schoolteacher in small town Maharashtra.

But here is what the labs and the papers and the institutions have given us: two hundred frameworks and a gap between knowing and doing that is widening, not narrowing. Here is what my grandfather gave his students: four ideas that stuck. That shaped how they lived. That their children carry to this day.

The problem with AI ethics was never a deficit of principles. It was a deficit of principles that people could actually carry in their heads and apply under pressure. Taylor understood this in 1932. My grandfather understood this without ever needing to articulate it.

I wrote a book to bridge the two. To take the values of a schoolteacher in Maharashtra and make them work in a world where machines make decisions about your loan, your resume, your health, your freedom.

The book is called Beyond Bias: The Four Way Test for Ethical AI. It is the most personal thing I have ever written. And it is the most practical.

If you want to know when it is ready:

I’m have written a book about exactly this  how AI and automated systems make decisions about your life, where accountability disappears, and what we can do about it. If you want to know morea about this book or order a copy, you can do it here: https://akshaywalimbe.com/beyond-bias/