Recon: Reconstruction-Guided Reasoning Synthesis for User Modeling

2026-05-26

AI Summary

Existing user modeling approaches often rely on post-hoc rationalizations of generated reasoning traces, which struggle to faithfully capture users’ latent decision-making processes. This work proposes a reasoning synthesis paradigm guided by action reconstruction: given contextual information and candidate reasoning, a reconstruction model predicts user behavior, and the fidelity of this reconstruction serves as a reward signal to optimize reasoning generation. This ensures that the synthesized reasoning naturally leads to the observed behavior rather than merely providing a retrospective explanation. The method requires no human annotations and achieves a 54.7% win rate over the Backward Synthesis baseline across four domains. Moreover, the generated reasoning demonstrates cross-model transferability and enhances downstream user modeling performance, yielding a 70.0% win rate.

Abstract

User modeling aims to use language models (LMs) to mimic an individual's behavior from a corpus of past context-action pairs (e.g., conversation turns), enabling the simulation of users in settings like behavioral science, human-AI collaboration, and market research. Recent approaches augment these corpora with synthesized reasoning traces, typically generated by conditioning on both context and action. However, such conditioning constitutes post-hoc rationalization rather than reasoning: the trace is guaranteed to justify the action, but may not encode the underlying latent causal decision paths. We propose Recon, which uses action reconstruction to score reasoning traces by their predictive power: given a context and candidate reasoning, a reconstruction model predicts the action, and reconstruction fidelity determines reasoning quality. Across four domains, Recon achieves a 54.7% win rate over Backward Synthesis, a standard post-hoc rationalization baseline. Further, we find that training a reasoning synthesis model with rewards derived from Recon improves downstream user modeling performance, achieving a win rate of up to 70.0% over baselines. We further show that Recon-synthesized reasoning transfers across models, and improves user modeling beyond the reconstruction model. Our work demonstrates that post-hoc rationalization is insufficient for reasoning synthesis, and that useful and interpretable reasoning should naturally elicit the action from the context.

Problem

• user modeling
• reasoning synthesis
• post-hoc rationalization
• action reconstruction
• causal decision paths

Innovation

• reconstruction-guided reasoning
• user modeling
• reasoning synthesis
• post-hoc rationalization
• action reconstruction