Active Feature Acquisition for Cost-Efficient Temporal Prediction with Reduced Participant Burden

2026-10-05

AI Summary

This study addresses the dual challenges of data collection burden and model interpretability in longitudinal psychological prediction. We propose a tree distillation method, introducing it for the first time into the Long-horizon Active Feature Acquisition (LAFA) framework to extract interpretable decision policies from deep neural networks. This approach enables dynamic selection of optimal feature subsets, effectively balancing predictive accuracy against acquisition costs. Evaluations on simulated data and ecological momentary assessment (EMA) experiments for alcohol consumption prediction demonstrate that the proposed method substantially reduces the number of items collected per session while incurring only marginal declines in predictive accuracy. By overcoming the deployment bottleneck associated with black-box models, this work establishes an interpretable paradigm for high-accuracy, low-burden psychological measurement.

Abstract

Accurate forecasting of pathological outcomes is a central problem in psychology. To do so, psychologists often collect intensive longitudinal data. However, in such studies, the desire to acquire a large number of variables for the sake of accurate prediction is often counteracted by the need to minimize participant burden. Acquiring more variables per occasion can yield better predictions, but having too many acquisitions increase the risk of non-response and attrition. Longitudinal Active Feature Acquisition (LAFA) is a principled approach to resolve this conundrum. Instead of requiring responses to every item at every acquisition occasion, LAFA produces a policy that seeks to optimally select dynamic subsets of items to be acquired at each timepoint while preserving our ability to forecast a specific outcome. However, existing LAFA methods are mostly based on Neural Networks (NN) that are difficult to interpret in practice. In this work, we introduce a tree distillation method for learning an interpretable policy from NN-based LAFA networks. We validated our method through both a simulation and an empirical EMA dataset on forecasting daily alcohol consumption. In both cases, we find that we can meaningfully reduce the number of items acquired at each occasion with minimal loss in accuracy. Networks (NN) that are difficult to interpret in practice. In this work, we introduce a tree distillation method for learning an interpretable policy from NN-based LAFA networks. We validated our method through both a simulation and an empirical EMA dataset on forecasting daily alcohol consumption. In both cases, we find that we can meaningfully reduce the number of items acquired at each occasion with minimal loss in accuracy.

Problem

• Longitudinal Active Feature Acquisition
• participant burden
• temporal prediction
• interpretability
• intensive longitudinal data

Innovation

• Longitudinal Active Feature Acquisition
• Tree Distillation
• Interpretable Policy
• Temporal Prediction
• Participant Burden