Psychological, economic, and ethical factors in human feedback for a chatbot-based smoking cessation intervention - Data and analysis code

This repository contains the data and analysis code for the paper "Psychological, economic, and ethical factors in human feedback for a chatbot-based smoking cessation intervention" by Nele Albers, Francisco S. Melo, Mark A. Neerincx, Olya Kudina, and Willem-Paul Brinkman.<br>The data and analysis code have previously been published in almost identical form as part of the PhD thesis by Nele Albers: https://doi.org/10.4121/1d9aa8eb-9e63-4bf5-98a3-f359dbc932a4. The main differences lie in the references to figures and tables due to different figure and table names.<br><strong>Study</strong>The paper is primarily based on data collected in a study conducted on the online crowdsourcing platform Prolific. In this study, daily smokers and vapers interacted with the text-based conversational agent Kai in up to five conversational sessions between 1 February and 19 March 2024. The Human Research Ethics Committee of Delft University of Technology granted ethical approval for the research (Letter of Approval number: 3683).<br>In each session, participants were assigned one of 37 preparatory activities for quitting smoking (e.g., envisioning their desired future self after quitting smoking/vaping, learning a breathing exercise, tracking their smoking behavior). Between each pair of sessions, participants had a 20% chance of receiving a feedback message from one of two human coaches, who were Master's students in Psychology. Out of 852 people who started the first conversational session, 500 completed all five sessions. 449 people further provided their preferences for allocating human feedback based on different principles in the post-questionnaire. There was also a follow-up questionnaire, but data from this questionnaire is not included in the analyses performed in this paper.<br>The study was pre-registered in OSF: https://doi.org/10.17605/OSF.IO/78CNR.<br>The implementation of the conversational agent Kai is available online: https://doi.org/10.5281/zenodo.11102861.<br>The 523 human feedback messages that were written can be found here: https://doi.org/10.4121/7e88ca88-50e9-4e8d-a049-6266315a2ece.<br><strong>Data</strong>This repository includes these types of anonymized data:<br>Data from the prescreening questionnaire (e.g., stage of change for quitting smoking/vaping),Data from people's Prolific profiles (e.g., age, gender),Data from the conversational sessions with Kai (e.g., effort spent on activities),Data from the post-questionnaire (e.g., preferences for allocation principles), andData on people clicking on the reading confirmation links in the human feedback messages.<br>The variable "rand_id" is a random participant identifier and can be used to link data from different data files.<br><strong>Analysis code</strong>All our analyses are based on either R or Python. We provide code to allow them to be reproduced.<br><br>In the case of questions, please contact Nele Albers ([e-mail]) or Willem-Paul Brinkman ([e-mail]).

本仓库包含论文《面向基于聊天机器人的戒烟干预的人类反馈中的心理、经济与伦理因素》（作者：Nele Albers、Francisco S. Melo、Mark A. Neerincx、Olya Kudina、Willem-Paul Brinkman）所使用的数据与分析代码。本数据与分析代码此前以近乎完全一致的形式发表于Nele Albers的博士学位论文中：https://doi.org/10.4121/1d9aa8eb-9e63-4bf5-98a3-f359dbc932a4。二者的主要差异仅在于图表引用部分，因图表与表格命名存在不同。

**研究概况**
本论文的核心数据来源于一项在众包平台Prolific上开展的研究。在2024年2月1日至3月19日期间，每日吸烟者与电子烟使用者可与基于文本的对话智能体（conversational agent）Kai进行最多5次对话会话。代尔夫特理工大学人类研究伦理委员会已为该研究授予伦理批准（批准函编号：3683）。
在每次会话中，参与者将被分配37项戒烟准备活动中的一项，例如：畅想戒烟/戒电子烟后的理想未来自我、学习呼吸训练法、记录吸烟行为。在每两次会话之间，参与者有20%的概率收到两位人类辅导员（均为心理学硕士研究生）所撰写的反馈消息。在852名启动首次对话会话的参与者中，共有500名完成了全部5次会话；其中449人在后续问卷中提供了其基于不同原则分配人类反馈的偏好。本研究还设置了追踪问卷，但该问卷的数据未纳入本论文的分析范畴。
本研究已在开放科学框架（Open Science Framework，简称OSF）上完成预注册：https://doi.org/10.17605/OSF.IO/78CNR。
对话智能体Kai的实现代码可通过以下链接获取：https://doi.org/10.5281/zenodo.11102861。
本次研究生成的523条人类反馈消息可通过以下链接获取：https://doi.org/10.4121/7e88ca88-50e9-4e8d-a049-6266315a2ece。

**数据集说明**
本仓库包含以下几类匿名化数据：
1. 预筛选问卷数据（例如：戒烟/戒电子烟的行为改变阶段）；
2. 参与者Prolific档案数据（例如：年龄、性别）；
3. 与Kai的对话会话数据（例如：参与活动所投入的精力）；
4. 后测问卷数据（例如：反馈分配原则偏好）；
5. 人类反馈消息阅读确认链接的点击行为数据。
其中变量`rand_id`为随机参与者标识符，可用于关联不同数据文件中的信息。

**分析代码**
本研究的所有分析均基于R语言或Python语言实现，我们已提供可复现分析过程的代码。

若有任何疑问，请联系Nele Albers（邮箱：[e-mail]）或Willem-Paul Brinkman（邮箱：[e-mail]）。