Abstract

There is a need to conceptualize a multidimensional taxonomy for learner-AI interaction. This conceptual/perspective article shares recent work on AI learner education and further presents new conceptions for a multidimensional taxonomy for learner-AI interaction. A review of the literature is conducted (N = 11). Open coding is used to summarize an overview of work, challenges, and findings reported. The summarized work is then used to conceptualize considerations for a multidimensional taxonomy for learner-AI interaction. The contribution of this work is in identifying unforeseen limitations in characterizing human-AI interaction and presenting new conceptions for a multidimensional taxonomy for learner-AI interaction based on the synthesis of the reviewed literature. This work thus shares current findings and challenges reported by the literature and our conceptions. Four conceptions are introduced, namely the alignment between the learner and AI; diverse metrics for the learner, AI, and learner-AI interaction; feedback direction when summarizing interactions; and what works in human-AI interaction by using prior research. We find there to be challenges with the use of AI by humans. The more interaction time spent between humans and AI may not necessarily lead to enhanced learning and understanding. Humans may exploit and use AI in inappropriate ways such as plagiarism. This eminent threat begs the question to reconsider our evaluation methods in light of AI systems.

中文翻译：

---

学习者与人工智能交互的多维分类

摘要

需要为学习者与人工智能交互构建多维分类法。这篇概念/视角文章分享了人工智能学习者教育的最新工作，并进一步提出了学习者与人工智能交互的多维分类的新概念。对文献进行了回顾（N  = 11）。开放编码用于总结工作、挑战和报告的发现的概述。然后，总结的工作用于概念化学习者与人工智能交互的多维分类的考虑因素。这项工作的贡献在于确定了描述人类与人工智能交互的不可预见的局限性，并根据综述文献的综合提出了学习者与人工智能交互的多维分类学的新概念。因此，这项工作分享了文献和我们的想法所报告的当前发现和挑战。引入四个概念，即学习者与人工智能的一致性；学习者、人工智能以及学习者与人工智能交互的不同指标；总结交互时反馈方向；以及通过使用先前的研究，什么在人机交互中起作用。我们发现人类使用人工智能面临挑战。人类与人工智能之间花费的互动时间越多，不一定会增强学习和理解。人类可能会以抄袭等不适当的方式利用和使用人工智能。这种突出的威胁引发了我们根据人工智能系统重新考虑我们的评估方法的问题。