How are real ethical dilemmas handled in AI-assisted projects?
Ethics is not a list of maxims: it is what you do when the shortcut is tempting, oversight is unlikely, and a justification is ready. Here you learn from cases, not from principles.
Ethics begins where rules end
Rules cover foreseeable cases; real dilemmas arise in unforeseen ones, where two right things conflict or where no one would ever check. A slightly inflated figure that no one will recalculate. An AI that wrote almost everything, and the temptation not to say so. A disappointing result that could be framed better than it actually is. Ethics in practice is how you behave in these moments — when getting it wrong is the convenient choice.
The three-light test
A simple method for everyday dilemmas. Sunlight: would this choice hold up if it were public, recorded in the register with full name attached? Others' light: would you find it fair if a competitor made the same choice towards you? Time's light: in five years, would you be glad you made it? A choice that passes all three lights may still be difficult; one that does not is already decided, even if it seems convenient.
The typical dilemmas of the AI era
Attribution: how much of the work is mine if AI did most of it? Simbial's answer: it is yours what you can defend, and transparency about tool usage is not optional. Synthetic truth: AI produces plausible but fabricated numbers and citations; using them without verification is not naivety, it is negligence. Manipulation: AI knows how to write persuasive texts on any thesis; using it to make something appear green when it is not is called greenwashing, and it is the exact opposite of our work.
Honesty about results: the hardest case
The most common dilemma in impact projects is not about AI: it is about numbers falling short of expectations. The target said minus 20 percent, the field says minus 12. The temptation to adjust the narrative is human; Simbial's rule is that the register tells the truth, always. An honest minus 12 builds reputation; a doctored minus 20, sooner or later discovered, destroys it — and with it the credibility of the entire community. Honesty about results is not heroism: it is the only stable equilibrium.
Ethics as an advantage, not a constraint
In the short term, the shortcut seems to win. In the medium term, the impact economy systematically rewards those who are reliable: Impact Controllers sign off more readily on projects with a clean track record, companies call back those who have never put them in an awkward position, and the merit scale weighs the rating. We built the system so that doing the right thing pays — but the system only works if the people within it truly believe in it.
The test that awaits you
These are not topics to read about: they are muscles to train. The AI Ethics Licence — the final certification of this programme — puts you in front of concrete dilemmas and evaluates your choices. There are no perfect answers: there are defensible choices and choices that are not. The next two pills give you the final tools, then it is your turn.
Frequently asked questions
What is greenwashing and why does it concern AI?
It is making something appear sustainable when it is not. AI makes this easier — it knows how to write persuasive texts on any thesis — and that is why those who work on impact have an extra responsibility: to use that same power in service of measurable truth.
How do you make a decision in an ethical dilemma without clear rules?
A good method is the three-lights test: would the choice hold up if it were made public? Would you find it fair if it happened to you? Would you be happy with it in five years? What doesn't pass the three lights is already decided.
What happens if a project misses its target?
You declare the true result. The Simbial register always tells the truth: an honest result below expectations builds reputation, a fabricated number destroys it.
← The full course · Next pill: What AI can really do (and what it can't) →