We use navigation systems without giving them much thought.
You enter a destination, the system knows the map, sees traffic and suggests a route. In the end, you choose which route to follow.
It sounds as if you are making the decision.
But imagine the navigation system has already decided:
- which map is being used;
- which roads are considered relevant;
- which routes seem most promising;
- which three options you get to see at all.
You are still “in the loop”.
But how much have you actually decided yourself?
Something similar happens with AI
We often discuss whether AI should be allowed to perform a task or make a decision. And as long as a human is still involved, that can feel reassuring.
But human control does not begin with the final approval click.
AI can influence the process much earlier.
It can determine which information is selected.
What the system optimises for.
Which patterns are highlighted.
Which options are put on the table.
And which alternatives remain out of sight.
The human may still formally have the final say, while much of the choice has already been shaped beforehand.
Human in the loop is not enough
That is why human in the loop is not enough on its own.
The better questions are:
Which map is the AI using?
Where is the system trying to steer us?
Which routes do I get to see?
And which ones do I not?
Only then comes the question:
Who makes the final decision?
Perhaps that is one of the real challenges of AI.
Not merely keeping a human somewhere in the process.
But making sure the human still has sight of the map, the destination and the alternatives.
Whoever chooses the map influences the route before anyone takes a turn.