Weighted Reward Allocation Mechanism for Data Collection Quality

Ayato Kitadai, Sinndy Dayana Rico Lugo, Sangjic Lee, Masanori Fujita, Nariaki Nishino

Abstract: Data utilization, which offers several benefits in real-world contexts, has caught the attention of several researchers and organizations. However, obtaining high-quality data is difficult. Moreover, it usually requires a large amount of compensation when the data are intended for private use. Recently, the development and refinement of methods to incentivize people to provide useful data, called mechanisms, has received special attention, consequently increasing demand for more efficient models. This study was conducted to propose a mechanism, “Weighted Reward Allocation Mechanism (WRAM)”, for promoting high-quality data supplied by rational data submitters considering budget constraints. The mechanism calculates and distributes the total budget in accordance with a weighted contribution evaluation vector of each item in a questionnaire. After mathematical details of WRAM are introduced, a simulation experiment is presented by which the data quality is compared with that of a widely used uniform reward mechanism. Results demonstrated that the submitter behavior changes depending on WRAM weights. Overall, our proposed mechanism performs better than the uniform reward mechanism in terms of the quality of the collected data.