Platform Design Is Not Neutral. Democracy Pays the Price.

In March 2018, reporting in The Guardian and The New York Times documented that Cambridge Analytica had harvested the personal data of millions of Facebook users without their explicit consent and used it to build psychographic profiles for political targeting during the 2016 US presidential election. Facebook's own internal rules had allowed the initial data collection through a personality quiz app. The company took years to act on what it knew. The story became public not through regulatory disclosure but through a whistleblower.

In October 2021, Frances Haugen provided internal Facebook documents to Congress and the press. Those documents, which became known as the Facebook Papers, showed that the company's own researchers had documented harms from its platform including damage to teenage girls' mental health from Instagram, the amplification of divisive and extreme content by recommendation algorithms, and internal awareness of how the platform was being used for ethnic incitement in countries like Ethiopia and Myanmar. In each case, internal researchers identified the problem. In each case, business considerations limited the response.

I keep coming back to these two episodes because they are not anomalies in an otherwise well-functioning system. They are evidence about how the system works. And the core mechanism in both cases is a platform design choice: optimize for engagement. When a platform's recommendation algorithm is designed to maximize time-on-site and interaction, it will systematically surface content that provokes strong emotional responses. Anger and outrage provoke strong emotional responses. Fear does too. Content that confirms existing beliefs and intensifies existing hostilities tends to be more engaging than content that introduces doubt or complexity. The algorithm does not prefer outrage because anyone designed it to prefer outrage. It prefers outrage because outrage keeps people on the platform.

Eli Pariser named the related phenomenon "the filter bubble" in his 2011 book of the same name. When recommendation systems personalize content based on your prior behavior, they create an information environment where you see more of what you already believe and less of what would challenge it. This is not primarily a political problem in Pariser's telling. It is an epistemological one. The filter bubble changes the informational conditions under which people form beliefs, make judgments, and reach conclusions about political reality. I think Pariser got the basic mechanism right even if the empirical research on filter bubbles has found mixed results when trying to quantify effects.

Evgeny Morozov has pushed further with what he calls "techno-solutionism," the tendency to frame complex social and political problems as engineering problems that better-designed systems can fix. His critique is that this framing is itself ideological. It treats the values embedded in platform design as technical rather than political, which makes them harder to subject to democratic scrutiny. When Facebook (now Meta) frames its content moderation problem as a machine learning challenge rather than a political governance challenge, it is making a claim about who has legitimate authority over the choices involved. The engineers and the company, not democratic institutions.

This is directly relevant to IS research. IS researchers who study social media platforms are not just studying technology adoption or user behavior. They are studying systems that shape political information environments, that affect how people form beliefs about elections and public health, and that operate under governance structures with almost no democratic accountability. That is a bigger set of stakes than most IS research explicitly acknowledges.

The IS research angle I find most pressing is the one about platform design choices and their political consequences. Infinite scroll, autoplay, notification systems, engagement-optimized algorithmic ranking: these are specific technical design decisions with documented behavioral effects. They are not inevitable features of digital communication. They are choices. Some of them were made to maximize advertising revenue. Some were made to increase platform stickiness. Some were made without much consideration of political consequences at all. IS researchers studying social media design have the tools to examine how these choices produce specific outcomes, and to ask what alternative designs would look like.

The governance question is also an IS question. What oversight mechanisms, regulatory or voluntary, would make platform design choices more accountable? What transparency obligations would allow researchers to study how recommendation algorithms affect political content? What institutional arrangements would give democratic oversight bodies access to the data they would need to regulate effectively? These are not purely legal questions. They are questions about information architecture, data access, accountability systems, and organizational governance. IS researchers have theoretical frameworks for all of them.

The Cambridge Analytica case and the Facebook Papers are now part of the public record. The causal question, how much did any of this actually change specific election outcomes, is harder to answer and I would not claim to have an answer. What I do think is clear, from the internal documents that became public and from the broader pattern of platform behavior, is that these design choices had real-world political consequences that the companies themselves knew about and made partial decisions not to address fully. That is the starting point for IS research, not the ending point.