Collaborative Research: HNDS-I:SweetPea: Automating the Implementation and Documentation of Unbiased Experimental Designs

合作研究：HNDS-I:SweetPea：自动化无偏实验设计的实施和记录

• 批准号: 2318548
• 负责人: Jonathan Cohen
• 金额: $ 32万
• 依托单位: Princeton University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2023
• 资助国家: 美国
• 起止时间: 2023-09-01 至 2026-08-31
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2318548&HistoricalAwards=false
• 关键词: Collaborative; Research; HNDS; SweetPea; Automating

Two important issues facing modern empirical research are those of transparency (how well others can figure out how the research was done) and replicability (whether the outcomes will be the same if someone else does the same experiments). Lack of transparency and failures of experiments to replicate stifle scientific progress and lead to a mistrust of the scientific method. This project develops an open-source programming language, SweetPea, that automates experimental design. It will help researchers understand and replicate the results of others, minimize errors and bias, and increase the efficiency and accuracy of the entire experimental process. Consequently, this technology will contribute to advancing scientific discoveries, make them more accessible and reliable, and provide researchers in empirical sciences with a valuable tool to aid their work.Many replication problems in the behavioral sciences arise because of the challenges encountered in implementing accurate and appropriately balanced experimental designs while avoiding confounding factors. Additionally, the lack of clear and transparent documentation reduces transparency and the ability to replicate experimental results. The SweetPea programming language is designed to facilitate reproducible experimental design. This project extends SweetPea's core functionality to support various design strategies, automating the documentation process, and expanding the community of users and contributors. SweetPea uses an intuitive interface for the declarative expression of experimental designs, and advanced computational algorithms for sampling and analysis. The software ensures that experimental designs are properly implemented without introducing unexpected confounds. In addition, the project leverages large language models for robust documentation of experimental designs. The project includes outreach activities to engage psychologists, neuroscientists, behavioral economists, and machine learning experts. Overall, this project improves the accuracy, transparency, and replicability of experimental designs, offering researchers an accessible and powerful tool for scientific investigation.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

现代实证研究面临的两个重要问题是透明度（其他人如何能理解研究的完成方式）和可复制性（如果其他人进行相同的实验，结果是否会一样）。 实验缺乏透明度和失败，无法复制窒息的科学进步并导致对科学方法的不信任。 该项目开发了一种开源编程语言SweetPea，该语言可自动化实验设计。 它将帮助研究人员理解和复制他人的结果，最大程度地减少错误和偏见，并提高整个实验过程的效率和准确性。因此，这项技术将有助于提高科学发现，使其更容易获得和可靠，并为经验科学的研究人员提供有价值的工具来帮助他们的工作。行为科学中的许多复制问题出现了，因为在实施准确和适当平衡的实验设计的同时避免了混乱的因素时遇到了挑战。 此外，缺乏清晰和透明的文档降低了透明度和复制实验结果的能力。 Sweet -pea编程语言旨在促进可再现的实验设计。该项目扩展了Sweetpea的核心功能，以支持各种设计策略，自动化文档流程，并扩大用户和贡献者的社区。 SweetPea使用直观的界面来表达实验设计的声明性表达，以及用于采样和分析的高级计算算法。 该软件可确保在不引入意外混杂的情况下正确实施实验设计。此外，该项目还利用大型语言模型来实验设计的强大文档。 该项目包括参与心理学家，神经科学家，行为经济学家和机器学习专家的活动。总体而言，该项目提高了实验设计的准确性，透明度和可复制性，为研究人员提供了一种可访问且有力的科学研究工具。该奖项反映了NSF的法定任务，并认为值得通过基金会的知识分子优点和更广泛的影响审查标准通过评估来进行评估。