Reading Cognition as Decisions Unfold in Words: A Factorized Inverse Decision Model

2026-08-10

AI Summary

Existing inverse decision models rely solely on action trajectories and struggle to capture dynamic verbal responses—such as language production, interaction, and hesitation—in cognitive tasks involving speech. This work proposes the Factorized Inverse Decision Model (FIDM), which, for the first time, incorporates verbal response dynamics into the modeling framework by decomposing task execution likelihood into distinct action and effort factors, thereby enabling disentangled estimation of actions and cognitive effort. FIDM leverages a language model to extract structured execution trajectories from raw transcripts and performs inverse decision inference using individual-specific parameters. Evaluated on conversational shopping data from 400 older adults, FIDM not only recovers the underlying factor structure in a selective manner but also significantly outperforms baseline approaches—including clinical scores, trajectory summaries, and frozen linguistic representations—in binary cognitive status classification.

Abstract

Inverse decision modeling infers latent properties of decision processes from observed behavior, but existing formulations rely primarily on action trajectories. In verbalized cognitive tasks, task execution also produces response dynamics that action-only formulations leave unmodeled, such as verbal production, interaction, and hesitation. We propose a factorized inverse decision model (FIDM) that decomposes each individual's task-execution likelihood into an action factor and an effort factor, governed by separate individual-specific parameters. From raw verbal transcripts, a language model produces structured task-execution traces for factorized inference. On data from 400 older adults performing a grocery-shopping dialog task for cognitive screening, controlled recovery shows selective estimation of the intended factors, while matched semi-synthetic conditions show that FIDM preserves action-execution distinctions even when aggregate behavioral summaries are matched. Action evidence further localizes task-defined deviations across participants. In cognitive-status classification, FIDM provides information complementary to clinical scores, trajectory summaries, and frozen language representations, with consistent gains across all evaluated baselines in the binary setting.

Problem

• inverse decision modeling
• verbalized cognitive tasks
• response dynamics
• action trajectories
• cognitive screening

Innovation

• inverse decision modeling
• factorized model
• verbal response dynamics
• language model
• cognitive screening