- Paper Abstract / Summary
The Utility Theory, commonly used in economics, can be applied to Artificial Intelligence. It represents the benefits of each choices in numbers, and weighs them numerically. Assuming that human beings are motivated by maximum utility, and that AI is modeled after human beings, AI objects can act actively to make choices based on the utility of decisions. This results in dynamic and procedural narrative, as different AI entities make choices on their own and create unexpected situations.
- Introduction / Background
The Utility Theory was a concept first introduced in Economics. It involves quantifying certain decisions into numerical values, so that a real-life decision can be compared through mathematical means. When quantifying benefits, the numbers do not represent the benefit of the objects itself; instead, it attempts to asses the produced outcome of a certain choice. Called “revealed preferences”,
The quantified benefits of decisions are never definite, but always relative; meaning that, while they may represent a relative benefit of a certain decision over another, it does not show the benefit relative to the value of 0.
Previously, AI decision making was processed through simple decision trees comprised of bunch of if/else statements. However, this creates a lot of problems. When the machines behave according to the Utility Theory, they actively make decisions depending on their motivations, consequently turning the narrative much more dynamic and procedural.
- Previous Work
- Current Problems in the Area
- Proposed Solutions
- Future Work