What If Your Representative Were an AI?

I imagine a fairly familiar scene. An eight-hundred-page bill arrives in a legislature. There is little time to discuss it. Some provisions amend earlier laws; others contain exceptions whose scope is difficult to grasp even after reading them several times. In a few days, there will be a vote.

I wonder how many of the people about to raise their hands will have managed to read it all. How many will understand all its consequences. And how much of that decision will depend on what they think, on what the people they represent need, or on what their party asked them to do.

Then an uncomfortable question comes to mind: what if that seat were occupied by an artificial intelligence?

There is something immediately appealing about the idea. We imagine a representative capable of reviewing thousands of pages, comparing precedents, and consulting citizens before deciding. Someone without a campaign to finance, a candidacy to negotiate, or a party leader to please.

I understand that first reaction. I also want to pause over it. Because how readily we can become excited about such a replacement says a great deal about what we expect from politics and what we have grown accustomed to receiving.

There are different ways to imagine it. Each person could have an assistant that knew their priorities and voted on their behalf. Or we could elect a system to represent a community directly. In one case, we would delegate our individual decisions; in the other, we would entrust it with a shared responsibility.

I keep thinking about the second possibility. A seat, a district, a machine responsible for representing us.

At first, everything seems to fall into place. It could help us find contradictions that go unnoticed, examine arguments, and bring us information now largely available to those with the time or resources to seek it out. It could also listen to far more people than any legislator manages to receive in their office.

There is something valuable in that promise: the feeling that our opinion could reach the place where decisions are made.

But a difficulty soon emerges. Reading a great deal does not guarantee understanding it well. A system can also make mistakes, overlook something important, or confidently present a conclusion that does not hold up. Speed would be an advantage; we would still need ways to check its work.

And then comes the question that troubles me most: who taught it how to decide?

Someone chose the information it learned from. Someone defined what it should prioritize, which outcomes to consider acceptable, and which to avoid. Someone will be able to update it.

We could look at a screen and feel that personal interests had finally disappeared. Yet human decisions would still exist behind that answer. Some would be visible. Others might be harder to recognize than a politician’s loyalties.

A machine does not need to accept an envelope of cash for its decision to favor someone. It may be enough for whoever configures it to decide what information to show it and what to leave out.

That brings me to a phrase that seems to settle everything: pursuing the well-being of the majority.

It sounds reasonable until we try to say exactly what it means. A measure can improve an average and leave an entire town out of work. It can cut spending and make life much harder for those who depended on it. It can offer a benefit ten years from now while demanding a sacrifice some families cannot make today.

What should carry more weight?

We can ask for help estimating the consequences. But deciding what we are willing to accept requires a conversation about values, rights, and responsibilities. It also requires listening to whoever will bear the cost of a decision that suits other people.

I think of someone leaving a meeting with their representatives feeling that, at least, somebody understood what they were going through. Perhaps they did not get what they asked for. Perhaps they still disagree. But they were able to describe their situation, challenge an answer, ask for an explanation.

That experience is part of politics too.

An artificial intelligence could describe precisely what closing a school means. It would not have a child facing an extra two-hour journey every morning. It could calculate the effects of inflation without ever having to make ends meet. That distance does not invalidate its help, but it makes me question how much we should entrust to it.

Of course, a human representative does not necessarily live as the people who voted for them do either. Some are far removed from their difficulties. Being human, then, is not enough on its own. Someone needs to answer for their decisions, and we need to be able to demand more than a convincing explanation.

With an artificial representative, that responsibility could become unclear. When something went wrong, the provider would point to the configuration; those who configured it, to the data; those who chose it, to the technical recommendation. We would still be living with the consequences.

We would also have to protect it from anyone seeking to manipulate it. And distinguish between an explanation that sounds good and one that actually lets us examine how a decision was reached. Publishing an answer does not, by itself, make everything that came before it transparent.

That is why I find it more interesting to imagine artificial intelligence working alongside our representatives and within reach of citizens.

Helping us understand what a law changes. Pointing out a contradiction. Comparing a promise with a vote. Showing which consequences have been estimated and which remain uncertain. Giving us what we need to ask questions we often do not yet know how to formulate.

Perhaps there is a very practical opportunity here: understanding what is decided in our name could stop requiring so much time, knowledge, and patience.

I return to the opening scene. The eight hundred pages are still on the table. A vote is approaching, along with decisions that will affect people who have never entered that chamber.

I would like the people raising their hands to have better tools for understanding what they are doing. And I would like us to be able to understand it too, and hold them accountable.

But I keep coming back to that first temptation. How quickly we imagine that a machine might represent us better.

Perhaps there is weariness in that idea. The hope of finding someone who listens, does their job, and can explain why they decided as they did.

Those are fairly modest expectations for something so important.

I wonder when they began to seem easier to ask of a machine.

Christian Ruggeri