Dataset and Analyses for Using a conversational agent for thought recording as a cognitive therapy task: feasibility, content, and feedback

This dataset contains all data and analysis scripts pertaining to the research conducted for the frontiers paper: "Using a conversational agent for thought recording as a cognitive therapy task: feasibility, content, and feedback." Following a literature review that we conducted in 2017 and 2018 on the technological state of the art of e-mental health for depression, we saw an opportunity to use technology in a more dialogical way than was being done to date. We therefore developed a conversational agent to support people in regularly recording their thoughts. This thought recording is a common technique in cognitive therapy. The cognitive approach to psychotherapy aims to change patients' maladaptive schemas, that is, overly negative views on themselves, the world, or the future. To obtain awareness of these views, they record their thought processes in situations that caused pathogenic emotional responses. We recruited 308 participants through Prolific, a crowd-sourcing platform for research participants. The participants interacted with our chatbot in two sessions, one practice session of two thought records based on scenarios and one actual session in which we asked to complete at least one personal thought record but as many additional ones as they wanted. We assessed the feasibility of completing the task with the agent, the content of the personal thought records, and whether the agent providing feedback on the content of the thought record (using natural language processing) had a positive e ect on the number of voluntarily completed thought records and participant's engagement in self-reection. We here deliver: a natural language dataset: the thoughts delineated by participants in the scenario-based and open thought records the coding of all personal thought records on their content by two independent coders: all thought records of the second session were labeled with respect to their content on the DIAMONDS and on three additional categories (COVID, Achievement/Competence, and Comprehensibility) analyses to test the hypotheses related to whether the feedback of the agent can increasemotivation to complete thought records additional materials (scenarios, qualtrics surveys, data management plan) that could assist in the replication of the study.

本数据集涵盖了为发表于《前沿》（Frontiers）期刊的研究所产出的全部数据与分析脚本，该研究的论文标题为"将对话式AI智能体（AI Agent）用于认知治疗任务中的想法记录：可行性、内容与反馈"。2017至2018年间，我们针对抑郁症领域电子心理健康技术的前沿现状开展了系统文献综述，期间发现了相较于既往方案更具对话性的技术应用契机。据此，我们开发了一款对话式AI智能体，用于辅助用户定期记录自身想法。此类想法记录是认知治疗中的常用技术。心理治疗的认知取向旨在改变患者的适应不良图式——即对自身、周遭世界或未来过度消极的认知。为察觉此类认知偏差，患者需在触发负性情绪反应的情境下记录自身的思维过程。我们通过科研众包平台Prolific招募了308名参与者。参与者需与我们开发的聊天机器人开展两次交互：一次为练习会话，基于预设情境完成2次想法记录；另一次为正式会话，要求参与者至少完成1次个人化想法记录，亦可自愿完成更多次数。我们针对以下维度开展评估：使用该智能体完成任务的可行性、个人化想法记录的内容，以及通过自然语言处理对想法记录内容提供反馈的智能体，是否可对自愿完成的想法记录数量与参与者的自我反思参与度产生积极影响。本研究公开的资源包括：1. 自然语言数据集：参与者基于预设情境与开放式任务完成的全部想法记录文本；2. 两名独立编码员对所有个人化想法记录的内容编码结果：第二次会话的全部想法记录均基于DIAMONDS编码框架，以及额外三个类别（COVID、成就/能力与可理解性）进行内容标注，用于检验智能体反馈能否提升完成想法记录的动机这一研究假设；3. 辅助该研究复现的附加材料：包括情境预设脚本、Qualtrics调查问卷与数据管理计划。