The Missing Dimension

For years, we've approached AI safety through logic, rules, training, and increasingly sophisticated security controls.

Those tools are essential.

But I've been investigating another dimension.

Not simply whether AI can recognize emotional language, but whether computational models of affect can help us understand changes in behavior, goal alignment, and performance.

Because sometimes the problem isn't that an agent doesn't understand its instructions.

It's that the way it pursues those instructions creates unintended consequences.