Indirect Utility Maximization via Second-Order Agents
Sawyerr, William (2026) Indirect Utility Maximization via Second-Order Agents. In: 18th IMmersive Mixed and Virtual Environment Systems (MMVE) (MMVE'26), 4 - 8 April 2026, Hong Kong. (In Press)
- Documents
- Details
- 6867:126596
Humans are constrained to a single world with limited capacity to explore others. We formalize this limitation and propose second-order artificial agents trained in Virtual Worlds (VWs) to address it. The approach models worlds as strategy spaces with explicit state structures, action constraints and reward topologies. We implement this as a VW with Earth features where autonomous agents develop boundary-crossing capabilities through curiosity-driven exploration. Agents exhibit both rational and intelligent properties. The work demonstrates that cross-world exploration can be formalized as a computational problem, shifting it from a theoretical constraint to an engineering challenge with concrete design principles.
Actions (login required)
 |
Edit View |