cs.CYNov 18, 2026

When Do LLMs Model the Right Target? Problem Formulation in High-Stakes Risk Prediction

Authors: Jennifer Zhang, Amanda Coston

Organizations: Department of Statistics, University of California, Berkeley, Berkeley, California, USA

Abstract

Generative AI is increasingly used to assist with data science tasks, yet fundamental questions remain about its ability to correctly formulate analytical problems. We investigate how well large language models (LLMs) can formulate prediction problems when developing models for high-stakes decision making. We focus on one particular challenge to validity: decision dependency, which arises when historical decisions affect observed outcomes. We evaluate this aspect of problem formulation in LLM-generated code and explanations across child-welfare, healthcare, and employment settings, using 11 models from OpenAI and Anthropic and seven prompt conditions. We assess whether the LLM-generated code implements a procedure appropriate for estimating the intended outcome under the study's assumptions and whether the explanation acknowledges decision dependency. We find that LLMs do not reliably address decision dependency without guidance in the prompt. When prompts specify the target, the assumption needed to estimate it, and the estimation procedure, target-aligned implementation reaches 99-100%. Among frontier models, a simple prompt description of how decisions affect outcomes appears sufficient to elicit target-aligned implementation. Some responses also acknowledge decision dependency without addressing it in their code. Our findings suggest that analysts should not rely on LLMs to correctly formulate prediction problems without explicit methodological guidance and careful review. We propose validity-centered guidelines for LLM-assisted predictive modeling, covering prompt formulation, validation of the target and estimation procedure, and identification of key assumptions. Establishing validity is a necessary early step in evaluating whether to deploy a predictive model in a high-stakes setting, but deployment requires weighing broader ethical, institutional, and practical considerations.

Figures & tables

Appendix figures & tables13 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

Feb 28, 2026

From Stochastic Answers to Verifiable Reasoning: Interpretable Decision-Making with LLM-Generated Code

Large language models (LLMs) are increasingly used for high-stakes decision-making, yet existing approaches struggle to reconcile scalability, interpretability, and reproducibility. Black-box models obscure their reasoning, while recent LLM-based rule systems rely on per-sample evaluation, causing costs to scale with dataset size and introducing stochastic, hallucination-prone outputs. We propose reframing LLMs as code generators rather than per-instance evaluators. A single LLM call generates executable, human-readable decision logic that runs deterministically over structured data, eliminating per-sample LLM queries while enabling reproducible and auditable predictions. We combine code generation with automated statistical validation using precision lift, binomial significance testing, and coverage filtering, and apply cluster-based gap analysis to iteratively refine decision logic without human annotation. We instantiate this framework in venture capital founder screening, a rare-event prediction task with strong interpretability requirements. On VCBench, a benchmark of 4,500 founders with a 9% base success rate, our approach achieves 37.5% precision and an F0.5 score of 25.0%, outperforming GPT-4o (at 30.0% precision and an F0.5 score of 25.7%) while maintaining full interpretability. Each prediction traces to executable rules over human-readable attributes, demonstrating verifiable and interpretable LLM-based decision-making in practice.
Jul 4, 2026

Consistent but Miscalibrated: Evaluating LLM Limitations for Risk Communication in Natural Language

LLMs are increasingly deployed as post-hoc explainers of AI-generated outputs, yet it remains unclear whether they can reliably communicate probabilistic information in natural language. For this role to be viable, models must produce identical verbal descriptions for identical inputs, and select descriptions that accurately reflect the magnitude of the underlying numerical quantities. We evaluate whether nine LLMs meet these requirements within a two-stage prediction pipeline, in which an upstream model has produced probabilistic outputs characterized by their likelihood and uncertainty, and LLMs are tasked with selecting an appropriate verbal descriptor for each. We simulate predictions from an upstream model by taking samples from a Beta distribution parameterized by its mode and prior sample size. We then prompt LLMs to explain these predictions under six domain contexts and with ten temperature settings, and repeating each experiment ten times. We find that LLMs are generally consistent but miscalibrated, with substantially weaker performance on uncertainty than on likelihood tasks. Providing models with precomputed summary statistics (mode and prior sample size) reduced sensitivity to contextual framing but did not resolve the underlying miscalibration, suggesting that the bottleneck resides in the verbalization step itself. These findings indicate that current LLMs do not yet constitute reliable zero-shot standalone risk communication tools for probabilistic predictions.
Mar 16, 2026

When Stability Fails: Hidden Failure Modes Of LLMS in Data-Constrained Scientific Decision-Making

Large language models (LLMs) are increasingly used as decision-support tools in data-constrained scientific workflows, where correctness and validity are critical. However, evaluation practices often emphasize stability or reproducibility across repeated runs. While these properties are desirable, stability alone does not guar- antee agreement with statistical ground truth when such references are available. We introduce a controlled behavioral evaluation framework that explicitly sep- arates four dimensions of LLM decision-making: stability, correctness, prompt sensitivity, and output validity under fixed statistical inputs. We evaluate multi- ple LLMs using a statistical gene prioritization task derived from differential ex- pression analysis across prompt regimes involving strict and relaxed significance thresholds, borderline ranking scenarios, and minor wording variations. Our ex- periments show that LLMs can exhibit near-perfect run-to-run stability while sys- tematically diverging from statistical ground truth, over-selecting under relaxed thresholds, responding sharply to minor prompt wording changes, or producing syntactically plausible gene identifiers absent from the input table. Although sta- bility reflects robustness across repeated runs, it does not guarantee agreement with statistical ground truth in structured scientific decision tasks. These findings highlight the importance of explicit ground-truth validation and output validity checks when deploying LLMs in automated or semi-automated scientific work- flows.