FIG. 18The spectrum of the ethical decision
the two extremes evade the decision · only the middle holds it
×
Extreme 1
Ban
"just in case"
Over-regulate to the point of deploying nothing. Elegant way to give away competitive advantage.
●
The right gray
Governance that takes
measured risks
Risks evaluated, consciously accepted by those with authority, and supervised with correction mechanisms.
×
Extreme 2
Deploy
"we'll see"
Pursue benefit without assuming the consequences of bias, error or failure. The bill always arrives at the worst moment.
Betrays the organisation's competitiveness
The honest positionthe hardest to hold — and the only ethical one
Betrays the people affected by the decisions
1
Bias and equity
Does AI reflect the world as it has been, or as it should be?
Systems learn from historical data, and historical data reflects past inequalities. You have to intervene actively in how it's trained and what's optimised, or the models inherit gender, age and origin biases from the previous system.
2
Privacy and data sovereignty
Who owns the information that trains the machine?
AI runs on data, the more the better. That clashes head-on with the right to control information about oneself. Health, behaviour, mobility: GDPR is a frame, but the hardest questions still lack a universal answer.
3
Transparency and explainability
Can an opaque decision have serious consequences?
The most powerful models are the most opaque. Decisions with impact on people must be auditable, contestable and appealable. Without that, automation doesn't empower: it subjects.
4
Disinformation and manipulation
How do we defend a shared base of facts?
Deepfakes, fabricated articles and fake profiles at scale erode shared reality. The speed of fabrication beats the speed of verification. Technical, institutional and educational problem at once.
5
Responsibility and attribution
When something fails, who responds?
The team that designed it? The company that deployed it? The leader who authorised it? Without clear frames, risk is privatised — falling on the most vulnerable people who interact with the systems.
Operational ethics, not decorative
These five dilemmas aren't solved with principles written on a wall
They're solved with the Governance block of EXOS: clear principles, information transparency, and the commitment that each person be accountable for the consequences of their decisions. Ethics in exponential times isn't a department. It's a muscle exercised in every daily decision.
"The ethical question about AI isn't what it can do. It is who answers when it does. And that question doesn't have a technical answer: only a political one."
— Exponential Times · Chapter 15