COGAM: Measuring and Moderating Cognitive Load in Machine Learning Model Explanations

要旨

Interpretable machine learning models trade -off accuracy for simplicity to make explanations more readable and easier to comprehend. Drawing from cognitive psychology theories in graph comprehension, we formalize readability as visual cognitive chunks to measure and moderate the cognitive load in explanation visualizations. We present Cognitive-GAM (COGAM) to generate explanations with desired cognitive load and accuracy by combining the expressive nonlinear generalized additive models (GAM) with simpler sparse linear models. We calibrated visual cognitive chunks with reading time in a user study, characterized the trade-off between cognitive load and accuracy for four datasets in simulation studies, and evaluated COGAM against baselines with users. We found that COGAM can decrease cognitive load without decreasing accuracy and/or increase accuracy without increasing cognitive load. Our framework and empirical measurement instruments for cognitive load will enable more rigorous assessment of the human interpretability of explainable AI.

キーワード
explanations
explainable artificial intelligence
cognitive load
visual explanations
generalized additive models
著者
Ashraf Abdul
National University of Singapore, Singapore, Singapore
Christian von der Weth
National University of Singapore, Singapore, Singapore
Mohan Kankanhalli
National University of Singapore, Singapore, Singapore
Brian Y. Lim
National University of Singapore, Singapore, Singapore
DOI

10.1145/3313831.3376615

論文URL

https://doi.org/10.1145/3313831.3376615

動画

会議: CHI 2020

The ACM CHI Conference on Human Factors in Computing Systems (https://chi2020.acm.org/)

セッション: Models & measurement

Paper session
316C MAUI
5 件の発表
2020-04-30 23:00:00
2020-05-01 00:15:00
日本語まとめ
読み込み中…