A Rigorous Behavior Assessment of CNNs Using a Data-Domain Sampling Regime

TL;DR

This paper introduces a data-domain sampling regime to evaluate CNNs' graphic perception, revealing that CNNs can outperform humans in ratio estimation tasks and their biases depend on training-test distribution differences.

Contribution

The authors develop a novel sampling regime for assessing CNNs' perception behaviors and provide extensive analysis comparing CNNs and humans across multiple trials.

Findings

CNNs can outperform humans in ratio estimation tasks.

CNN biases are influenced by training-test distribution discrepancies.

The sampling regime enables detailed analysis of CNN perception behaviors.

Abstract

We present a data-domain sampling regime for quantifying CNNs' graphic perception behaviors. This regime lets us evaluate CNNs' ratio estimation ability in bar charts from three perspectives: sensitivity to training-test distribution discrepancies, stability to limited samples, and relative expertise to human observers. After analyzing 16 million trials from 800 CNNs models and 6,825 trials from 113 human participants, we arrived at a simple and actionable conclusion: CNNs can outperform humans and their biases simply depend on the training-test distance. We show evidence of this simple, elegant behavior of the machines when they interpret visualization images. osf.io/gfqc3 provides registration, the code for our sampling regime, and experimental results.