Are Explanations Helpful? A Comparative Study of the Effects of Explanations in AI-Assisted Decision-Making
Are Explanations Helpful? A Comparative Study of the Effects of Explanations in AI-Assisted Decision-Making
复制标题
解释有帮助吗？
DOI：
10.1145/3397481.3450650
复制
发表时间：
2021
期刊：
影响因子：
--
通讯作者：
Yin, Ming
中科院分区：
文献类型：
--
作者：
Wang, Xinru;Yin, Ming
本文通过对一组已建立的XAI方法在AI辅助决策中的效果进行比较，为可解释AI（XAI）方法的经验评估提供了越来越多的文献。具体来说，基于我们对以往文献的回顾，我们强调了理想的AI解释应该满足的三个理想属性-提高人们对AI模型的理解，帮助人们认识到模型的不确定性，并支持人们对模型的校准信任。通过随机对照实验，我们评估了四种常见的模型不可知的可解释AI方法是否在两种类型的决策任务上满足这些属性，在这两种类型的决策任务中，人们认为自己具有不同的领域专业知识水平（即，累犯预测和森林覆盖预测）。我们的研究结果表明，人工智能解释的影响在很大程度上是不同的决策任务，人们有不同程度的领域专业知识，许多人工智能解释不满足任何理想的属性的任务，人们几乎没有领域的专业知识。此外，对于人们更有知识的决策任务，特征贡献解释被证明可以满足AI解释的更多需求，而被认为类似于人类如何解释决策的解释（即，反事实解释）似乎并没有改善校准的信任。最后，我们讨论了我们的研究对改进XAI方法的设计以更好地支持人类决策的影响。
This paper contributes to the growing literature in empirical evaluation of explainable AI (XAI) methods by presenting a comparison on the effects of a set of established XAI methods in AI-assisted decision making. Specifically, based on our review of previous literature, we highlight three desirable properties that ideal AI explanations should satisfy—improve people’s understanding of the AI model, help people recognize the model uncertainty, and support people’s calibrated trust in the model. Through randomized controlled experiments, we evaluate whether four types of common model-agnostic explainable AI methods satisfy these properties on two types of decision making tasks where people perceive themselves as having different levels of domain expertise in (i.e., recidivism prediction and forest cover prediction). Our results show that the effects of AI explanations are largely different on decision making tasks where people have varying levels of domain expertise in, and many AI explanations do not satisfy any of the desirable properties for tasks that people have little domain expertise in. Further, for decision making tasks that people are more knowledgeable, feature contribution explanation is shown to satisfy more desiderata of AI explanations, while the explanation that is considered to resemble how human explain decisions (i.e., counterfactual explanation) does not seem to improve calibrated trust. We conclude by discussing the implications of our study for improving the design of XAI methods to better support human decision making.
登录
查看更多内容
DOI：
10.1145/2030112.2030168
发表时间：
2011-09
期刊：
--
影响因子：
--
作者：
Brian Y. Lim;A. Dey
通讯作者：
Brian Y. Lim;A. Dey
可解释人工智能（XAI）中的反事实：来自人类推理的证据
DOI：
--
发表时间：
2019
期刊：
International Joint Conference on Artificial Intelligence
影响因子：
--
作者：
R. Byrne
通讯作者：
R. Byrne
基于特征的解释不能帮助人们检测在线毒性的错误分类
DOI：
--
发表时间：
2020
期刊：
Proceedings of the Fourteenth International AAAI Conference on Web and Social Media
影响因子：
--
作者：
Carton, Samuel;Mei, Qiaozhu;Resnick, Paul
通讯作者：
Resnick, Paul
DOI：
10.1016/s0020-7373(87)80013-5
发表时间：
1987-11-01
期刊：
INTERNATIONAL JOURNAL OF MAN-MACHINE STUDIES
影响因子：
--
作者：
MUIR, BM
通讯作者：
MUIR, BM
DOI：
--
发表时间：
2019-01
期刊：
ArXiv
影响因子：
--
作者：
Isaac Lage;Emily Chen;Jeffrey He;Menaka Narayanan;Been Kim;Sam Gershman;F. Doshi-Velez
通讯作者：
Isaac Lage;Emily Chen;Jeffrey He;Menaka Narayanan;Been Kim;Sam Gershman;F. Doshi-Velez