Identifying and Measuring Manipulative User Interfaces at Scale on the Web

Abstract

In this dissertation, I present measurement methods to automatically identify manipulative user interfaces—colloquially known as “dark patterns”—at scale on the web. Using these methods, I quantify the prevalence of dark patterns in three studies and show how dark patterns are rampant on the web, thus a pressing concern for society. First, I examine whether social media content creators, or “influencers,” disclose their relationships with advertisers to their audience. Analyzing over 500K YouTube videos and 2.1M Pinterest pins, I find that only about 10% of all advertising content is disclosed to users. Second, I examine various types of dark patterns in shopping websites. Analyzing data from 11K shopping websites, I discover over 1,800 dark patterns on over 1,200 websites that mislead users into making more purchases or disclosing more information than they would otherwise. Third, I examine dark patterns in political emails from the 2020 U.S. election cycle. Through an analysis of over 100K emails, I find that over 40% of emails from the median sender contain dark patterns that nudge recipients to open emails or make donations they might otherwise not make. I further outlay the conceptual foundation of dark patterns and articulate a set of normative perspectives for analyzing the effects of dark patterns. I conclude with how the lessons learned from the studies can be used to build technical defenses and to lay out policy recommendations to mitigate the spread of these interfaces.