EFaST: An Explainable Framework for Fair Sequential Task Assignment

In this demo, we present EFaST, an interactive framework for exploring and understanding fairness in multi-stakeholder sequential task assignment problems. EFaST allows users to iteratively define and modify stakeholder soft constraints, execute a fairness-aware optimization process, and observe how adjustments to these constraints influence fairness outcomes. Fairness is evaluated at two complementary levels: local fairness captures the satisfaction of individual stakeholders, while global fairness jointly measures overall satisfaction and balance across stakeholders. Users can inspect intermediate solutions, explore fairness trade-offs arising from conflicting constraints, and request natural-language explanations grounded in structured outputs. These explanations summarize constraint satisfaction, highlight critical violations, and provide guidance on potential constraint refinement, making fairness dynamics observable and controllable.