Social media platforms have gained a firm foothold in the everyday and are increasingly important as work and civic spaces. At the same time, they are not neutral conduits for user creativity and expression; rather, they channel behavior and organize information through design and data-driven techniques. Today, algorithms have taken center stage in the discussion of platform power. In this dissertation, I interrogate the continually (re)emerging empirical landscape of algorithmic social media platforms, and the scholarly debates concerning them, at various moments and from different angles. The central puzzle motivating the investigation is that on one hand, autonomous and powerful algorithms are depicted as an exogenous force, controlling and manipulating social media users for the benefit of corporations, while on the other, algorithms and machine learning models are presented as simply reflecting human behaviors and consumer demand. In encountering their outputs, it is claimed, we are in fact coming face-to-face with ourselves and our societies.

The dissertation is comprised of four original publications and employs a mixed methods approach. Taking my methodological cue from work that situates the social relevance of algorithms in how they shape and are shaped by social action, I engage with algorithms as embedded in broader socio-technical assemblages and in the contexts in which they are used. I use computational methods, close reading of social media data, and interviews to analyze in context how platform affordances and algorithms activate public debate and subjectivity in platform-specific ways (Articles I and II). I access users’ algorithm-related experiences and practices through qualitative analysis of online forum data (Article III), and engage in conceptual exploration to better understand the hybrid forms of agency emerging in and through the interactions among people, data, and algorithms (Article IV).

The four original publications enrich discussions that lie at the center of social media research, such as disinformation, social media activism, algorithmic content moderation, and human agency in relation to systems of automated decision making. The synthesized contribution of this thesis is then less about generalizing and theorizing based on a single empirical case and more about drafting a perspective to algorithmic social media platforms that applies across different cases and levels of analysis, from platform policy to everyday emotional responses to algorithms. Avoiding structural determinism on the one hand and praxeocentrism on the other, this perspective foregrounds how algorithmic social media platforms and their use(r)s mutually condition one another. However, this co-shaping takes place in an uneven landscape: if users are involved in molding algorithmic phenomena, they are only involved selectively, because they have little control over how their behavior becomes datafied and the purposes for which the algorithmic system optimizes it. A key argument of the thesis is that in light of the intimate and recursive human-algorithm relations that characterize today’s digital media landscapes, we need to take seriously the generative nature of use without assuming or exaggerating its empowering qualities.