Dignity as a layer, not a slogan

Dignity appears in almost every AI principles page and almost no actual systems. The difference between the two is whether dignity is a line the system cannot cross by design, or a word it prints while crossing it.

Can dignity be engineered into AI?

Dignity can be a design constraint, not a slogan. In most AI documents it is a value declared and then unconnected to anything the system does. Treated as a layer, dignity becomes concrete: specific things the system will not do to a person regardless of what the optimization wants, built into the values layer for AI rather than into the brochure. That is harder and far more honest than a principles page, and it is the kind of thing trustworthy systems are actually made of.

What dignity-as-a-layer means

It means encoding floors, not just goals. A goal can be traded away when another metric wins. A floor cannot: the system will not deny a person recourse, will not make an irreversible call about them with no human path, will not treat them purely as a number to be optimized. Those are design commitments you can check, not sentiments you can only assert.

How it differs from a values statement

A values statement describes intentions; a layer constrains behaviour. The test is whether anything the system does would change if you deleted the statement. If not, it was decoration. Dignity as a layer fails loudly when violated, which is exactly why it is worth building and why so few do.