Researchers have introduced Decision-Value Attribution (DVA), a new framework based on Shapley values designed to explain the decisions made by predictive models in operational systems. Unlike standard methods that focus on forecast explanations, DVA attributes value directly to the downstream decisions induced by these forecasts. The framework offers variants to attribute value to features, operational configurations, and the interactions between information and design parameters. Case studies in electricity storage and emergency medical services demonstrate that DVA can provide more accurate insights into operational value than predictive explanations alone, guiding interventions and clarifying when predictive information is relevant to decision-making. AI
IMPACT This framework could improve the interpretability and effectiveness of AI systems used in operational decision-making.
RANK_REASON The cluster contains an academic paper detailing a new framework for decision-value attribution. [lever_c_demoted from research: ic=1 ai=1.0]
Read on Hugging Face Daily Papers →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
