SEED-SET: Scalable Evolving Experimental Design for System-level Ethical Testing

Researchers propose SEED-SET, a new Bayesian experimental design framework for ethical testing of autonomous systems like drones in high-stakes environments. The system uses hierarchical Gaussian Processes to model both objective evaluations and subjective stakeholder judgments, generating up to 2x more optimal test candidates than baseline methods.

• →SEED-SET addresses the critical need for automated ethical benchmarking of autonomous systems deployed in human-centric domains.
• →The framework combines domain-specific objective evaluations with subjective stakeholder value judgments using hierarchical Gaussian Processes.
• →Testing shows the method generates up to 2x optimal test candidates compared to baselines with 1.25x better coverage of high-dimensional search spaces.
• →The approach provides an interpretable trade-off between exploration and exploitation for ethical AI testing.
• →Current ethical benchmarking is understudied due to lack of well-defined metrics and stakeholder-specific subjectivity challenges.