EAGER: Computationally and Socially Guided Self-Experiments

这项研究将开发一个完全自动化的系统，指导用户调整健康跟踪和行为改善信息技术系统，以满足他们的个人需求。 许多人在新奇感消失后对健康跟踪设备失去了兴趣。 虽然使用户能够通过数据改善自己的健康和健康的目标是有影响力的，但目前的方法严重依赖于显示图表并提供有关所采取的步骤和心率的汇总统计数据，从长远来看，对大多数人来说并不足够激励。个性化推荐是提高参与度的关键因素。 这个项目提出了一个新的模式，为社会的改善，还没有被测试过之前-计算引导的自我实验。 它使人们能够对行为变化进行严格的结论性分析，而无需了解统计分析和实验设计的细节，并且无需使用现有的移动的设备。其基础技术是一种基于干预的实验设计，观察和学习对每个人有效的方法。完整的周期包括通过可操作的行为改变建议生成干预措施，从结果中学习因果分析，然后返回生成新的干预措施。实验每月都在不断发展，因此用户总是被要求做出可操作的更改，并被告知正在进行的结果。实验设计遵循单病例干预研究方法的标准，但经过修改以利用贝叶斯统计来消除不确定或过长的实验。该研究包括一个调查的有效性，引导自我实验作为一种新的范式，改善人？s的生活。测试的一个条件是，将人们分组为匿名队列以分享正在进行的结果是否有助于激励他们继续进行漫长的实验。群体的社会性允许人们互相观察？的进步，自我激励，并获得新的自我实验运行的想法。

This research will develop a fully automated system that guides users in adapting health tracking and behavior improvement information technology systems to suit their personal needs. Many people lose interest in health tracking devices after the novelty wears off. While the goal of empowering users to improve their own health and wellness with data is impactful, current methods which rely heavily on displaying charts and providing summary statistics about steps taken and heart rate, are not sufficiently motivating for most people in the long term. Personalized recommendations are a key factor to improve engagement. This project poses a new paradigm for societal improvement that has not been tested before - computationally guided self-experiments. It enables people to conduct rigorous conclusive analysis of behavior change without having to know details about statistical analysis and experimental design, at no cost by using their existing mobile devices. The underlying technology is an intervention-based experimental design that observes and learns what works for each individual.The complete cycle includes generating interventions through actionable behavior change suggestions, causal analysis to learn from outcomes, and then back to generating a new intervention. The experiment is constantly evolving from month to month, so the user is always being asked to make actionable changes and informed of ongoing results. The experimental design is guided by standards from single-case intervention research methods, but modified to take advantage of Bayesian statistics to eliminate inconclusive or excessively long experiments. The research comprises an investigation into the efficacy of guided self-experiments as a new paradigm for improving people?s lives. One condition tested will be whether people grouped into anonymous cohorts to share ongoing results can help motivate them to continue lengthy experiments. The social nature of the groups allow people to observe each other?s progress for self-motivation, and gain ideas for new self-experiments to run.