MIMS Final Project 2016

A User-Centered Perspective on Algorithmic Personalization

This study focuses on user attitudes toward algorithmic personalization, or the process by which user data is employed to tailor content to users online. Drawing from a large-scale survey which includes experimental vignettes, we aim to bring a user-centered perspective to existing policy discussions on online personalization practices and how the potential harms from these practices should be addressed. We observe how users feel about the use of inferred personal information to support personalization, and whether the accuracy of those inferences is important to them. We take a contextual approach to understanding user attitudes about these aspects of online personalization by looking at them in three different domains: advertisements, search results, and pricing, all with a range of types of personal data.

Our findings suggest that some personal data types (whether inferred or provided) should not be used as the basis for personalization at all. In cases where personalization may be seen as fair and useful to consumers, companies should implement data practices that are conscious not only of the process-based methods of the Fair Information Practice Principles (FIPPS), but also of the ways that the context of the personalization may violate cultural values, resulting in potentially negative effects on user attitudes.

This research will contribute to guidelines for the fair and responsible use of algorithms that the Center for Democracy and Technology is developing as part of their Digital Decisions project.

The results of this research will inform policy discussion around the need for transparency on the part of providers who use personalization tools in their products or services. Providers’ decisions around personalization based on information about individuals such as location, browsing history, or demographics should be driven by an empirical understanding of what factors are most influential in users’ perceptions of unfairness.

This project is supported by the Center for Technology, Society & Policy (CTSP) and by the Center for Long-Term Cybersecurity (CLTC).

Last updated: December 6, 2016