Ending Research Subject Overexploitation: Methods to Reduce Respondent Overuse and Privacy Violations while Increasing Insights from Data

Abstract

The low price of data collection and use in the Internet age has facilitated collective ir- responsibility, where private companies, academics, and governments all fail to internalize the costs to respondents and other researchers of inefficient data strategies. As a norm, we overexploit subjects by surveying people we do not need to, by extracting more information than we need to from those we do contact, and by failing to protect the privacy of respondents when we use their personal data. In any single case, the privacy risk or additional burden on subjects is small. But our actions compound. We now confront a bad equilibrium where individuals withhold their personal data for fear of its misuse, data owners restrict access for fear of legal repercussions, and we all lose out on the valuable insights our collective data could generate. To avoid this, and to ensure ongoing researcher access to high quality information, we require new methodological tools for data collection, sharing, and use that treat personal data as a scarce resource. At the core of these methods should be a principle of minimizing extraneous information from subjects through careful design. In this work I contribute to this agenda with new experimental designs for efficient policy learning, and new methods for uncovering insights about populations while protecting the privacy of individuals.