[3I5-OS-4c] 信頼と文脈のインタラクションデザイン

[3I5-OS-4c-03] A Trial of Selecting Bounding Boxes to Show While Obtaining Reliance

〇Yosuke Fukuchi¹, Seiji Yamada^1,2 (1. National Institute of Informatics, 2. SOKENDAI)

Keywords:Explainable AI, Trust calibration, Selective explanation

Explanations on AI decision-making play an important role in building human trust and increasing reliance on AI. However, an excessive amount of explanations can lessen the effects or even bring negative outcomes. This paper considers a method for reducing the number of presenting explanations while keeping human reliance in a scenario where a human decides whether to rely on an autonomous vehicle while observing its object detection results as explanations. In an experiment, we found that our proposed method can reduce the number of bounding boxes to show while keeping human reliance.