Year: 2020
Type: dissertation
Abstract: Social media is an important means for interaction, but one that is increasingly curated by automated systems that aim to intelligently infer user intentions. This dissertation studies how users experience different mechanisms to configure and control these automated systems. Three sets of experiments, combining surveys, interviews, and participatory design workshops, seek to triangulate how users experience these mechanisms for control. We find that one prevalent mechanism, control settings, improves user satisfaction, but they do so whether or not they are functioning. Despite that, however, users describe turning to improvisational behaviors rather than control settings to see what they want on their news feeds. A second mechanism for control, appeals of content moderation decisions, does not improve user perceptions of the fairness, trustworthiness or accountability of the system, despite user calls for human review of automated decisions. These existing mechanisms appear to fall short of users’ needs. As a result, we explore a newly proposed mechanism for control – contestability, whereby systems are designed so that users can shape and influence decision-making processes. We find that when everyday users envision this, they require substantial (and currently absent) support in the form of representation, argumentation, access and compassion from platforms. While this dissertation identifies few easy solutions for the challenges of automated systems for social media, it calls for renewed attention the importance of control as a design feature.
Author: Vaccaro, Kristen
Institution: University of Illinois Urbana-Champaign
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7036770929