The Decisions We Hand To Machines

By David Bruce

I was in the car this morning, driving my youngest child to primary school. Country roads.

Quiet. All very familiar.

Ahead, a heavy goods vehicle (HGV) pulled out to overtake a slower-moving tractor. I saw

two oncoming vehicles appearing in the distance, heading straight towards me.

For a split-second, scenarios unfolded in my mind that caused that hot, prickly sensation

down my spine. My grip tightened on the wheel and I quickly began to assess my next move,

my brain on high alert.

Fortunately, the vehicles passed safely. The tractor continued on its way and as quickly as

the danger appeared, it was gone.

But the mind of a parent does not remain calm for long after such a moment.

• What if there wasn’t enough time to react?

• What if this happened around a blind bend just as the HGV committed to the

• manoeuvre?

• Do I brake? Do I swerve? Do I rely on instinct or on training?

The moment passed uneventfully, yet the thought lingered.

Later that day, reflecting again, my mind jumped to a conversation about Artificial

Intelligence and ethics—a thought experiment we half-jokingly called the "Bruce Trolley

Problem"

The Hypothetical Scenario:

Imagine you are a principal software engineer responsible for designing the core decision

logic of a fully autonomous self-driving vehicle.

A heavy vehicle is approaching at speed on a narrow carriageway. A devastating collision is

unavoidable.

You must choose which logic to encode: Option One

The vehicle continues on its current trajectory. The impact is catastrophic and the

probabilistic outcome is that no occupants survive.Option Two

The system initiates an aggressive evasive manoeuvre. This reduces the severity of the

collision. Most occupants survive, but one will not.

On paper this may seem like a technical optimisation. In reality, it is a profound moral and

ethical dilemma. Introducing the Human Element

Inside the vehicle are:

• Two adults in the front

• A younger adult in the rear seat

• An elderly passenger beside them

Pause and consider the implications.

How should the system decide?

Do you:

• Minimise overall harm, knowing this requires the sacrifice of one life?

• Or avoid intervention, accepting the worst outcome for all occupants because it spares the system from making a value-laden choice?

And the questions grow more uncomfortable:

• If the system can infer age, vulnerability or survival likelihood, should it use that data?

• Whose values determine how such sensitive data is weighted?

• And who bears moral and legal responsibility? The developer? The corporation? The regulator? The machine that simply executes the logic it was given?

The Real Issue with AI

This is where AI discussions often fail.

We treat them as:

• Technical optimisation challenges

• Questions of data quality

• Model performance issues

But this is not a data problem.

It is an inherently human, moral problem.

Data can tell us what is. It cannot tell us what ought to be done.

Every automated decision system reflects human judgements and values—whether

acknowledged or not. As these systems operate at immense speed and scale, human value-

judgements become embedded, invisible and difficult to challenge or audit.

This is the true hidden risk.

The danger is not that machines will make all our decisions.

It is that we stop owning the ethical decisions we've already handed over to them.The

Critical Question

Senior leaders and policymakers should not ask whether AI can make life-or-death choices.

They must ask:

Are we prepared to be clear, honest and accountable for the human values we encode into

autonomous systems?

Because those values will guide the machine when the critical moment comes. A Question

for You

What would be your personal choice?

And more importantly:

Who should ultimately be held accountable when the machine executes that choice?

I am genuinely curious. Let us discuss the implications.