Ethics

Artificial Intelligence and Human Freedom

An option can remain on the screen after it has become impractical to choose.

A bright moving walkway runs straight through a transit hall beside open walking routes.
An imagined arrangement: one choice made easier than the others. AI-generated image

A useful recommendation can give someone a choice they did not know existed. It can also make the same narrow set of choices easier to encounter again. The presence of a recommendation tells us neither which has happened nor what the system was designed to reward.

Ask who set the objective. A librarian’s suggestion, a paid advertisement, and a model choosing the next item to prolong a session are different arrangements, even when they offer the same book.

Freedom in such an arrangement includes the capacity to understand enough of its terms, change them, or leave. A nominal opt-out is poor evidence of consent if refusal costs a person employment, care, or access to essential services.

The useful friction

Some friction protects a choice: a pause before publishing private information, a second look at a consequential recommendation, an easy way to reset a profile. Other friction prevents a choice: a cancellation process harder than signing up, an appeal route inaccessible to the person who needs it.

Calling both merely convenience or inconvenience loses their moral direction. Whose action is being made harder, and for whose benefit?

A person can try one small change without promising a transformation: choose the next thing before opening the feed, inspect a default, or keep a period free from recommendations. The Unclaimed Interval offers a form. It is not a remedy for institutional power and not a test for people whose access depends on a screen.

For builders, the stronger obligation is structural. Make the objective legible, make refusal usable, and do not punish the person who takes the alternative. A life need not be arranged around the metric that made the interface profitable.