Guide suggests user preferences by smart questions during talks
GUIDE: Generative Utility Inference and Decision Engine
Machine Learning
Summary
People have many preferences that are hard for computers to understand, especially when these preferences involve many different factors. The authors created GUIDE, a computer system that chats with users and asks smart questions to quickly learn what they like. It uses both clever sampling methods to pick questions and rules about the world to better guess preferences. When they tested GUIDE on choosing investment portfolios, it made better recommendations early on compared to other methods. This helps computers make choices more closely matched to what users really want.
What this means in practice
- •For financial advisors: Provide personalized investment advice by quickly understanding client preferences through interactive conversations.$Commercial implications: Enables advisory platforms to offer tailored portfolio recommendations using guided preference inference, improving client satisfaction and retention.
- •For product designers: Design better user experience by integrating adaptive questioning systems that uncover customer needs efficiently during product selection.
Authors
Anagha Tiwari, Alexander G. Gray, Nick Feamster, Brian Jabarian, Alex Imas, Alex Kale
Abstract
Measuring the preferences of human users remains a fundamental challenge of AI alignment. Existing elicitation approaches struggle to efficiently discover multidimensional preferences or accurately ground these inferences in domain knowledge. To address this, we introduce GUIDE, an LLM-driven elicitation architecture that infers user preferences through conversations by combining Bayesian adaptive sampling for question selection and symbolic representation learning to initialize domain-specific preference models. GUIDE generalizes adaptive sampling to diverse elicitation questions through an extensible type system of transforms on a parameterized preference state. GUIDE produces domain-specific preference representations through an initialization process using symbolic rule-based learning to capture world knowledge and set priors over preference dimensions grounded in data about decision alternatives. The architecture provides observability and steerability to facilitate deployment and analyze elicitation processes. In silico experiments on investment portfolio optimization demonstrate that GUIDE improves cold-start and minimizes recommendation regret consistently within early elicitation interactions across user personas compared to prior work, LLM-only baselines, and ablated GUIDE versions.