Synthesizing Images on Perceptual Boundaries of ANNs for Uncovering and Manipulating Human Perceptual Variability

2025-05-06

AI Summary

Human perceptual decision variability—particularly under uncertainty—exhibits substantial individual differences whose cognitive and neural origins remain poorly understood. Method: We introduce the Boundary Alignment and Manipulation (BAM) framework, enabling predictive modeling and controlled intervention of human perceptual boundaries. Leveraging artificial neural networks (ANNs), we synthesized boundary-optimized stimuli; conducted large-scale online behavioral experiments (N=246 participants, 116,000 trials); and constructed variMNIST—the first systematically annotated dataset of perceptual boundary images (19,943 samples). Using boundary sampling, subject-specific model alignment, adversarial generation, and ANN interpretability analysis, we precisely predicted and modulated pairwise perceptual disagreements. Contribution/Results: We empirically demonstrated that boundary-aligned images significantly amplify perceptual decision variability. BAM establishes a novel paradigm for individualized perceptual modeling, providing both theoretical foundations and methodological tools for probing the mechanisms underlying human perceptual heterogeneity.

Technology Category

• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Computer Vision: Generative Adversarial Networks (GANs) for Vision

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Human decision-making in cognitive tasks and daily life exhibits considerable variability, shaped by factors such as task difficulty, individual preferences, and personal experiences. Understanding this variability across individuals is essential for uncovering the perceptual and decision-making mechanisms that humans rely on when faced with uncertainty and ambiguity. We present a computational framework BAM (Boundary Alignment&Manipulation framework) that combines perceptual boundary sampling in ANNs and human behavioral experiments to systematically investigate this phenomenon. Our perceptual boundary sampling algorithm generates stimuli along ANN decision boundaries that intrinsically induce significant perceptual variability. The efficacy of these stimuli is empirically validated through large-scale behavioral experiments involving 246 participants across 116,715 trials, culminating in the variMNIST dataset containing 19,943 systematically annotated images. Through personalized model alignment and adversarial generation, we establish a reliable method for simultaneously predicting and manipulating the divergent perceptual decisions of pairs of participants. This work bridges the gap between computational models and human individual difference research, providing new tools for personalized perception analysis.

Problem

• Investigates human perceptual variability in decision-making tasks
• Develops a framework to predict and manipulate perceptual decisions
• Bridges computational models and human individual difference research

Innovation

• Combines ANN perceptual boundary sampling with human experiments
• Generates stimuli along ANN decision boundaries for variability
• Uses personalized model alignment to predict and manipulate decisions