Not All AI are Equal: Exploring the Accessibility of AI-Mediated Communication Technology
Not All AI are Equal: Exploring the Accessibility of AI-Mediated Communication Technology
复制标题
DOI：
10.1016/j.chb.2021.106975
复制
发表时间：
2021-08-12
影响因子：
9.9
通讯作者：
Hancock, Jeffrey T.
中科院分区：
文献类型：
--
作者：
Goldenthal, Emma;Park, Jennifer;Hancock, Jeffrey T.
虽然人工智能技术和工具为其用户提供了各种潜在好处，但尚不清楚所有人是否都有平等的机会获得这些好处。我们研究了可用性和可访问性之间的差距，因为它涉及到采用人工智能介导的通信（AI-MC）工具，它使人际沟通，其中智能代理代表通信。在定义了六种功能性AI-MC类型（语音辅助通信，语言纠正，预测文本建议，转录，翻译，个性化语言学习）后，我们对519名美国参与者进行了一项在线调查，该调查结合了封闭式和开放式措施。我们的定量结果揭示了AI-MC的采用如何与软件，设备和互联网接入工具（如语音辅助通信）相关;人口统计因素，如翻译和转录工具的年龄，教育和收入;以及特定功能工具的AI-MC素养的一些组成部分。我们的定性分析为这些发现提供了额外的细微差别，我们阐明了访问，理解和使用AI-MC工具的一些障碍，我们认为这些障碍阻碍了传统上因一刀切的技术工具而处于不利地位的用户群体的AI-MC可访问性。最后，我们呼吁广泛解决数字技术行业内的可访问性问题。
While AI technologies and tools offer various potential benefits to their users, it is not clear whether opportunities to access these benefits are equally accessible to all. We examine this gap between availability and accessibility as it relates to the adoption of AI-Mediated Communication (AI-MC) tools, which enable interpersonal communication where an intelligent agent operates on behalf of a communicator. Upon defining six functional AI-MC types (voice-assisted communication, language correction, predictive text suggestion, transcription, translation, personalized language learning) we conducted an online survey of 519 U.S. participants that combined closed- and open-ended measures. Our quantitative results revealed how AI-MC adoption is related to software, device, and internet access for tools such as voice-assisted communication; demographic factors such as age, education and income in the case of translation and transcription tools; and some components of AI-MC literacy for specific functional tools. Our qualitative analyses provide additional nuance for these findings, and we articulate a number of barriers to access, understanding, and usage of AI-MC tools, which we suggest hinder AI-MC accessibility for user groups traditionally disadvantaged by one-size-fits-all technological tools. We end with a call for broadly addressing accessibility concerns within the digital technology industry.