Ask Before You Optimize: Dynamic Pre-Formulation Clarification for Interactive Optimization
Teaching AI to ask clarifying questions before solving business math problems
When companies describe optimization problems to AI in plain language, they often leave out crucial details—missing constraints, hidden objectives, or business rules that completely change the right answer. Researchers built a benchmark and an AI system that learns to recognize these gaps and ask targeted clarifying questions before attempting to solve the problem, rather than blindly modeling incomplete information.
Operations research teams rely on optimization models to make million-dollar decisions about supply chains, scheduling, and resource allocation. If an AI formulates a model based on an incomplete or misunderstood problem statement, the resulting "solution" could be useless or actively harmful. Teaching AI systems to pause and ask clarifying questions before modeling—rather than guessing at missing details—is essential for making these tools trustworthy enough to use in real business contexts.