Acquiring Semantic Knowledge for User Model Updates via Human-Agent Alignment Dialogues: Transcriptions

The data is the transcriptions for the focus groups underlying the paper "Acquiring Semantic Knowledge for User Model Updates via Human-Agent Alignment Dialogues: An exploratory focus group study" by Pei-Yu Chen, Myrthe Tielman, Dirk Heylen, Catholijn Jonker, and Birna van Riemsdijk. The goal of this paper is to explore what potential users like or dislike about certain aspects of the dialogues and identify dimensions that are important for designing good alignment dialogues.<br><strong>Study</strong>We performed an exploratory focus group user study, in which we showed participants six scenarios with different variants of how we envision such alignment dialogues might look like. The scenarios and dialogues are in textual form. After they finished reading, we asked them to discuss and compare the dialogues, and then we moved on to the next scenario and discussion. The process continued until all six scenarios were discussed. For the structure of the discussion, we prepared the following questions to guide the participants:Which version of the dialogue do you prefer, or which part of which dialogue do you prefer? Why?Is there a certain part of the dialogue that you particularly like/not like? Why? How would you want to do it instead?Which dialogue is more ‘intelligent’, as in has more capability in providing support?Do you feel one dialogue is more supportive than the other?After which dialogue do you think the agent would be more ‘on the same page’ as you?<br><strong>Data &amp; analysis</strong>We transcribed the focus group sessions and analyzed the transcriptions using inductive thematic analysis with the addition of triangulation with literature.本数据集为论文《基于人机对齐对话获取用户模型更新所需语义知识：一项探索性焦点小组研究》（英文原名：Acquiring Semantic Knowledge for User Model Updates via Human-Agent Alignment Dialogues: An exploratory focus group study；作者为Pei-Yu Chen、Myrthe Tielman、Dirk Heylen、Catholijn Jonker与Birna van Riemsdijk）所依托的焦点小组（focus groups）转录文本。该论文旨在探索潜在用户对人机对齐对话（human-agent alignment dialogues）各维度的好恶，并识别出设计优质对齐对话的关键维度。<br><strong>研究</strong>本研究开展了一项探索性焦点小组用户研究，向参与者展示六种不同变体的人机对齐对话场景设想，所有场景与对话均以文本形式呈现。参与者阅读完毕后，我们引导其对对话展开讨论与对比，随后进入下一情景与讨论环节，直至六个情景全部讨论完毕。为规范讨论流程，我们准备了以下引导问题供参与者参考：<br>你更偏好哪一版对话，或是偏好某一对话的哪个部分？理由为何？<br>是否存在某段对话内容是你特别喜欢或反感的？理由为何？你会如何调整该部分内容？<br>哪一版对话更具“智能性”，即能够提供更优质的支持服务？<br>你是否认为某一版对话相比其他更具支持性？<br>你认为经过哪一版对话后，智能体（agent）能够更贴合你的想法，与你达成共识？<br><strong>数据与分析</strong>我们对焦点小组会话进行了转录，并采用归纳式主题分析（inductive thematic analysis）结合文献三角验证法（triangulation with literature）对转录文本展开分析。