Many real-world eligibility problems, ranging from medical diagnosis to tax planning, can be mapped to decision problems expressed in natural language, wherein a model must make a binary choice based on user features. Large-scale domains such as legal codes or frequently updated funding opportunities render human annotation (e.g., web forms or decision trees) impractical, highlighting the need for agents that can automatically assist in decision-making. Since relevant information is often only known to the user, it is crucial that these agents ask the right questions. As agents determine when to terminate a conversation, they face a trade-off between accuracy and the number of questions asked, a key metric for both user experience and cost. To evaluate this task, we propose BeNYfits, a new benchmark for determining user eligibility for multiple overlapping social benefits opportunities through interactive decision-making. Our experiments show that current language models struggle with frequent hallucinations, with GPT-4o scoring only 35.7 F1 using a ReAct-style chain-of-thought. To address this, we introduce ProADA, a novel approach that leverages program synthesis to assist in decision-making by mapping dialog planning to a code generation problem and using gaps in structured data to determine the best next action. Our agent, ProADA, improves the F1 score to 55.6 while maintaining nearly the same number of dialog turns.

The ability to derive useful information by asking clarifying questions (ACQ) is an important element of real life collaboration on reasoning tasks, such as question answering (QA). Existing natural language ACQ challenges, however, evaluate generations based on word overlap rather than the value of the information itself. Word overlap is often an inappropriate metric for question generation since many different questions could be useful in a given situation, and a single question can be phrased many different ways. Instead, we propose evaluating questions pragmatically based on the value of the information they retrieve. Here we present a definition and framework for natural language pragmatic asking of clarifying questions (PACQ), the problem of generating questions that result in answers useful for a reasoning task. We also present fact-level masking (FLM), a procedure for converting natural language datasets into self-supervised PACQ datasets by omitting particular critical facts. Finally, we generate a PACQ dataset from the HotpotQA dataset using FLM and evaluate several zero-shot language models on it. Our experiments show that current zero-shot models struggle to ask questions that retrieve useful information, as compared to human annotators. These results demonstrate an opportunity to use FLM datasets and the PACQ framework to objectively evaluate and improve question generation and other language models.