arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2203.12131cs.LGcs.CY
收藏

机器学习模型在毫无头绪时是否应该向我们报告?

Should Machine Learning Models Report to Us When They Are Clueless?

Yale Center for Medical Informatics(耶鲁医学信息学中心);VA Connecticut Healthcare System(美国退伍军人事务部康涅狄格医疗系统)

Roozbeh Yousefzadeh, Xuenan Cao

摘要:

人工智能可解释性的权利已经巩固为研究界和政策制定的共识。然而,可解释性的一个关键组成部分一直缺失:外推,它描述了人工智能模型在遇到不熟悉的样本时可能毫无头绪的程度(即,在它们的训练集的凸包之外的样本,如我们将解释的)。我们报告说,人工智能模型经常在不通知用户和利益相关者的情况下,在其熟悉的数据范围之外进行推断。了解模型是否已经外推是一个基本的见解,应该包括在解释人工智能模型中,以支持透明度和问责制。我们没有纠缠于负面因素,而是提供了一些方法来清除促进人工智能透明度的障碍。我们的分析评论伴随着实用条款,这些条款有助于纳入人工智能法规,如美国的《国家人工智能倡议法案》和欧盟委员会的《人工智能法案》。

英文摘要:

The right to AI explainability has consolidated as a consensus in the research community and policy-making. However, a key component of explainability has been missing: extrapolation, which describes the extent to which AI models can be clueless when they encounter unfamiliar samples (i.e., samples outside the convex hull of their training sets, as we will explain). We report that AI models extrapolate outside their range of familiar data, frequently and without notifying the users and stakeholders. Knowing whether a model has extrapolated or not is a fundamental insight that should be included in explaining AI models in favor of transparency and accountability. Instead of dwelling on the negatives, we offer ways to clear the roadblocks in promoting AI transparency. Our analysis commentary accompanying practical clauses useful to include in AI regulations such as the National AI Initiative Act in the US and the AI Act by the European Commission.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑