Navigating Digital Worlds: Empirical Studies of Choice and Behavior in Sociotechnical Systems

Abstract

This dissertation empirically examines the nature and consequences of three types of choices that individuals, firms, and communities make (or might make) against the background of complex sociotechnical environments, new technologies, and competing goals. First, in Chapter 1, I present research about consumer choice and price comparison behavior in the context of the market for ridesharing services Uber and Lyft. Combining several novel data sources along with benchmark evidence from existing literature, the work calibrates a simple sequential search model to benchmark observed search behavior against theoretical predictions. The work finds that consumers compare prices substanially less than canonical models would predict given observed levels of price dispersion and benchmark estimates of consumer search costs. While the individual benefits of searching are modest, the aggregate implications are large; in a back-of-the-envelope calculation, we find that New York City-based rideshare customers collectively leave over $300 million per year on the table by not comparing prices (about 6% of platforms' gross booking volume), illustrating how small frictions can have substantial aggregate implications for the distribution of surplus in digital markets. Second, in Chapter 2, I present research that studies conflict and disagreement in online conversations. The study develops and experimentally tests a novel AI-based mediation tool designed to facilitate constructive online dialogue across political divides. Powered by a large language model (LLM), the tool automates mediation interventions grounded in communication and conflict resolution principles such as paraphrasing, identifying agreement, and encouraging perspective-taking. In a randomized controlled trial, the system successfully generated context-sensitive interventions in contentious conversations; however, effects on participants' attitudes towards people they disagree with were limited. The findings highlight both the promise and the challenges of using LLMs to promote healthier online discourse at scale. Note that results presented in the present chapter are based on a partial sample of data from the study; future drafts of this work will include results from the full sample of data. Finally, in Chapter 3, I present emerging research that studies workers' beliefs about the future labor market impact of emerging digital technologies, with a focus on AI and large language models (LLMs). As AI tools like LLMs become increasingly capable, workers must make decisions about how to respond. Importantly, these decisions are shaped not only by the actual rate and direction of technological change, but also by workers' beliefs about these future trajectories. Importantly, these beliefs may differ across workers and may not align with expert forecasts about the likely impact of AI, while nonetheless shaping economic behavior. In this research, we design and conduct a survey and randomized experiment to study the role of worker beliefs in shaping labor market responses to generative AI, with a focus on two professional fields: law and management consulting. There are three main findings. First, we find that beliefs vary substantially within and across professions. Second, we find that workers expecting larger AI impacts are somewhat more likely to report AI training, degree enrollment and workplace LLM use (though results are noisy and not significant based on current sample). Third, we find that exposure to contrasting expert narratives shifts stated beliefs about the likely impact of AI on the labor market; we do not find evidence of a significant impact on behavioral outcomes in current sample (small effects possible with more data). The findings in this chapter are based on preliminary pilot data from an ongoing project; findings may change as additional data collection continues. The title of this dissertation, "Navigating Digital Worlds", emphasizes that, at least collectively, we have agency in the digital worlds that we create and inhabit. Sociotechnical spaces are not fixed or inevitable, but are instead imagined, constructed, and shaped by people, technologists, experts, and policymakers. The choices we make about the structure of these spaces then shapes the opportunities, constraints, and consequences we face in subsequently navigating them.