What Makes a Good Example? Modeling Exemplar Selection with Neural Network Representations

2026-02-03

AI Summary

This study investigates how humans balance representativeness and diversity when selecting exemplars for teaching. By embedding novel visual categories into a one-dimensional morph continuum derived from pretrained convolutional and Transformer-based models, the authors systematically compare multiple subset selection strategies—including prototypicality, joint representativeness, and diversity—in their ability to model human exemplar choices. For the first time, neural network representations are integrated with principled subset selection criteria, revealing that joint representativeness—particularly when combined with diversity—best accounts for human judgments. Moreover, representations from Transformer models significantly outperform those from convolutional architectures, thereby validating dataset distillation as an effective computational model of pedagogical reasoning.

Technology Category

• Computer Vision: Representation Learning for Vision
• Machine Learning: Representation Learning
• Knowledge Representation and Reasoning: Common-Sense Reasoning

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Teaching requires distilling a rich category distribution into a small set of informative exemplars. Although prior work shows that humans consider both representativeness and diversity when teaching, the computational principles underlying these tradeoffs remain unclear. We address this gap by modeling human exemplar selection using neural network feature representations and principled subset selection criteria. Novel visual categories were embedded along a one-dimensional morph continuum using pretrained vision models, and selection strategies varied in their emphasis on prototypicality, joint representativeness, and diversity. Adult participants selected one to three exemplars to teach a learner. Model-human comparisons revealed that strategies based on joint representativeness, or its combination with diversity, best captured human judgments, whereas purely prototypical or diversity-based strategies performed worse. Moreover, transformer-based representations consistently aligned more closely with human behavior than convolutional networks. These results highlight the potential utility of dataset distillation methods in machine learning as computational models for teaching.

Problem

• exemplar selection
• teaching
• representativeness
• diversity
• computational modeling

Innovation

• exemplar selection
• neural network representations
• joint representativeness
• dataset distillation
• transformer vs CNN