Public life depends on more than access to information. It depends on people being able to recognize common problems, understand how others are responding to them, and judge whether collective action is possible. When algorithmic systems personalize not only the information people receive but also their perception of the social field surrounding it, they shape the conditions under which people interpret one another, form judgments, and decide whether to participate.

In digitally mediated public spaces, two people can encounter the same post while receiving different surrounding signals. The comments surfaced, reactions prioritized, and expressions of support or hostility may vary, along with the cues that suggest what is popular, acceptable, or controversial. Each person may then take a selective environment as evidence of public opinion, despite engaging with the same underlying content.

These differences can be subtle and difficult to detect. Users rarely know how a platform has ranked, filtered, or organized what appears around a post, and may not know that someone else is encountering a different arrangement. A selected set of reactions can therefore be mistaken for a representative picture of what the wider public thinks.

The question is not simply whether platforms distribute information. It is how systems designed to capture attention, learn from behavior, and maximize engagement influence what people believe others think, feel, and support.

To understand how fragmentation develops, however, it is necessary to look beyond platform algorithms alone. The information entering those systems has already been shaped by institutional, economic, and political forces.

The public sphere is shaped before the feed

The information environment begins before content reaches a platform. Ownership, funding, editorial practice, regulation, distribution, and institutional priorities influence which events are investigated, which voices are resourced, and which narratives enter public circulation.

Media capture describes situations in which powerful political or economic actors influence media institutions, funding pathways, regulatory conditions, or distribution systems in ways that weaken editorial independence and public accountability. Capture does not require every journalist or outlet to receive direct instructions. It can operate through ownership concentration, political relationships, state advertising, licensing arrangements, commercial dependence, philanthropic priorities, or pressure on institutions that provide public scrutiny.

Corporate funding does not automatically constitute media capture. The relevant question is whether financial or ownership relationships affect independence, pluralism, dissent, investigative capacity, or the ability of media institutions to scrutinize powerful actors.

These upstream conditions matter because platform systems do not create public narratives from nothing. Funding and ownership influence production; production influences narratives; narratives enter platform systems; platform systems shape visibility; visibility shapes perception; perception influences behavior.

Focusing only on the feed can therefore make the platform appear to be the beginning of the information system, when it is also a later stage in a wider chain of production, circulation, and influence.

Attention is a psychological resource

Human attention is limited. People therefore rely on signals that suggest what is important, urgent, threatening, rewarding, socially meaningful, or worthy of response.

Emotion is one of the strongest of these signals. Anger, particularly moral anger and outrage, can direct attention toward perceived wrongdoing, identify an object of blame, and create an impulse to respond. A person may comment, share, challenge, monitor, defend, seek confirmation, or return repeatedly to the same conflict.

Anger is not the only emotion involved. Fear, anxiety, curiosity, surprise, belonging, and social approval can also sustain attention. The wider point is that emotionally activating content is often more difficult to ignore than information that appears neutral or remote.

This creates an important distinction between what people consciously prefer and what systems learn to promote. A person may not want to spend an hour consuming hostile or distressing content, yet the platform may interpret continued viewing, reacting, or returning as evidence that similar material should be shown again.

Research examining engagement-based ranking has found that emotionally charged and hostile content can be amplified even when it does not reflect users’ considered preferences. Immediate reaction is not the same as reflective endorsement.

From emotional activation to behavioral feedback

When a person pauses, clicks, comments, shares, replays, or reacts angrily, that behavior becomes data. The system can use it to predict what may hold the person’s attention next.

A feedback loop can then develop: