regret

Terms from Artificial Intelligence: humans at the heart of algorithms

Human regret is the emotion we feel when (typically) something goes wrong and we think of ways in which we could have acted differently. It can be a crippling emption, but is actually a finely tuned learnibg mechanism that focuses out attention on the small changes in behaviour that would make a big difference in outcomes. As well as this more typical negative regret, there is also a variation, posituve regret when things turned out well, but could have been even better. Regret involves complex counter-factual reasoning as well as lower emotion and stimuus--response learning. Regret can used in AI as a success metric during machine learning: how well does the system behave compred with the best possible outcome. Models of human regret can also be used both to help understand human cognition, and to enhance machine learning.

Defined on pages 546, 546

Used on pages 104, 126, 379, 533, 544, 546, 548, 549