Ethics & the Good Life

Every lesson so far in this track asked what is true or what is real. This one asks a different kind of question: what should you do? The gap between those two kinds of question is the whole hinge of ethics, and it is the same gap that makes building a well-behaved AI hard.

Before you start. The last lesson looked at how arguments work, and where reasoning hits walls it can prove exist but never get past. This lesson is about a wall of a different kind: no amount of correct reasoning about what is the case tells you what you should do.

The idea in 30 seconds

There are three ways to judge the same act: by its outcome (consequentialism), by the duty it keeps or breaks (deontology), or by the character of the person choosing (virtue ethics).
Two famous cases, a runaway tram steered by a lever onto one person instead of five, and a stranger pushed off a bridge to stop it, trade one life for five, yet most people approve the first and refuse the second: steering a harm feels different from using a person as the tool that stops it.
Each way has a real weak point: judging by outcomes can justify killing one person to save five, judging by duties can forbid even a lie that would save a life, and judging by character can’t always say which act a good person would pick.
Hume’s is/ought gap says no set of facts alone forces a value, which is why the three lenses can disagree forever, and why an AI’s goals cannot be read off its abilities.
The veil of ignorance is a portable fairness test: would you accept this rule not knowing which side of it you would land on?

Read the story · Skip to the numbers