Abstract

Effective optimization is essential for interactive systems to provide asatisfactory user experience. However, it is often challenging to find anobjective to optimize for. Generally, such objectives are manually crafted andrarely capture complex user needs accurately. Conversely, we propose anapproach that infers the objective directly from observed user interactions.These inferences can be made regardless of prior knowledge and across differenttypes of user behavior. Then we introduce: Interactive System Optimizer (ISO),a novel algorithm that uses these inferred objectives for optimization. Ourmain contribution is a new general principled approach to optimizinginteractive systems using data-driven objectives. We demonstrate the higheffectiveness of ISO over several GridWorld simulations.

 

Introduction (beta)

None

 

Conclusion (beta)

None