DDRIG in DRMS: Measuring Persuasion Without Measuring a Prior Belief: A New Application of Planned Missing Data Techniques

接受建议的实验通常要求接受建议的人在收到实际建议之前，表达他们对正在接受建议的话题的信念。然后，在给出建议后，被咨询者报告了一个新的信念。然后，研究人员可以通过观察被建议者的信念如何改变来研究建议的效果。然而，这种设置存在一个潜在的问题。让被咨询者在收到建议之前明确考虑他们的期望可以改变其效果，但如果事先不知道被咨询者的信仰是什么，就无法研究他们的信念是如何改变的。这项研究使用尖端的统计方法，根据已知的其他因素来估计被建议者的先前信念，使研究人员能够在更现实的条件下了解建议的影响，在被建议者被建议之前不必直接考虑他们相信什么。测量影响后续行为过程的情况的科学术语是“测量反应性”。这个项目解决这种测量反应性的方法是将被咨询者的先前信念视为丢失的数据。统计学方法的进步使研究人员能够在正确的一组情况下无偏见地归因于丢失的数据。这些方法通常在偶然丢失数据时使用，但该项目将在实验上随机化数据丢失时，确保正确的条件使补偿有效工作。由于这些被建议者从不直接报告他们先前的信念，研究人员可以研究在没有测量反应性的情况下建议利用是如何发生的。这项工作对于顾问如何在许多现实生活场景中帮助被咨询者优化他们的专业知识具有重要的意义。使用计划缺失数据实验设计来治愈测量反应性的方法创新也有可能推广到其他科学领域，在这些领域它是一个潜在的混杂。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Experiments on advice taking typically require a person receiving advice (an “advisee”) express their belief about the topic they are being advised about before receiving the actual advice. Then, after advice is given, the advisee reports a new belief. Researchers can then study the effects of advice by looking at how the advisee’s belief changed. However, there is a potential problem with this setup. Making the advisee explicitly consider their expectation before receiving advice can change its effects, but there is no way to study how the advisee’s belief changes without knowing what they believed beforehand. This research uses cutting edge statistical methods to estimate advisees’ prior beliefs based on other factors that are known about them, allowing researchers to understand the influence of advice under more realistic conditions, where advisees don’t have to directly think about what they believe before being advised. The scientific term for situations where measurement affects subsequent behavioral processes is “measurement reactivity”. This project’s approach to curing this measurement reactivity is to treat advisees’ prior beliefs as missing data. Advances in statistical methods allow researchers to impute missing data with without bias under the right set of circumstances. These methods are typically utilized when data are missing incidentally, but this project will experimentally randomize when data are missing, ensuring the correct conditions for the imputation to work effectively. Because these advisees never directly report their prior beliefs, the researchers can study how advice utilization occurs in the absence of measurement reactivity. This work has important implications for how advisors can help advisees optimize their expertise in many real-life scenarios. The methodological innovation of using a planned missing data experimental design to cure measurement reactivity also has the potential to generalize to other scientific domains where it is a potential confound.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.