What is it about?
We examine how companies and websites' choice architecture deployment when collecting consumer data can affect the quality of data they collect. In particular, we look at the tradeoff between two objectives when companies collect consumer data---maximizing data volume (getting more data) and minimizing data bias (having a more representative dataset), and characterize how effective choice architecture design can balance the two objectives when they compete with each other.
Featured Image
Photo by Mika Baumeister on Unsplash
Why is it important?
Existing choice architecture design often emphasize maximizing the volume of data collection as their end goal. What is often neglected is a second dimension of data quality—the representativeness of data. Biased input data often leads to biases in algorithmic predictions and inferences, which can in turn lead to biased insights and decision-making. Therefore, it is crucial to understand how choice architecture affects the bias in the data collected and how to design better choice architecture to collect more representative data.
Perspectives
A biased dataset compromises the quality of data-driven analytics and inferences. In fact, a large yet biased dataset can lead to overly precise yet wrong estimates, misguiding business and policy decisions. As such, the potential bias induced by a volume-maximizing design can counteract the benefits of collecting more data and decrease the value of shared data as a result. As empirical researchers and analysts in the digital economy, we need to think more about how to collect better and more representative data, not just more data.
Tesary Lin
Read the Original
This page is a summary of: Choice Architecture, Privacy Valuations, and Selection Bias in Consumer Data, July 2023, ACM (Association for Computing Machinery),
DOI: 10.1145/3580507.3597674.
You can read the full text:
Resources
Contributors
The following have contributed to this page
