In 2016, a Facebook data scientist named Andrew Bosworth wrote an internal memo that became famous after it leaked. The relevant line: “Maybe someone gets exposed to bullying… maybe someone dies in a terrorist attack coordinated on our tools. And still we connect people.” The document laid bare what researchers had long suspected — that the algorithms optimising for engagement were indifferent to consequences.
In 2026, those consequences are better understood. This article explains the mechanisms clearly, without alarmism, so you can engage with political information more deliberately.
Mechanism 1: The Engagement Loop
Social media algorithms do not show you what is true. They show you what keeps you on the platform. Anger, outrage, and tribal affirmation are the most effective retention mechanisms. A 2021 MIT study found that false news stories spread six times faster than true ones on Twitter — not because people prefer falsehoods, but because false stories tend to be more emotionally provocative.
The algorithm learned this pattern from billions of interactions and optimised for it. Nobody told it to spread misinformation. It discovered that inflammatory content drove engagement, and engagement was the objective.
Mechanism 2: The Filter Bubble
Personalisation is useful when you are shopping for shoes. It is dangerous when you are forming political views. When every article, video, and post you see confirms what you already believe, you stop encountering challenges to your thinking. You stop knowing what the other side actually believes (as opposed to what your side says they believe).
Research by Eli Pariser, who coined the term “filter bubble,” found that even Google searches return different results to different users based on their history. Two people searching “immigration” can see fundamentally different information.
Mechanism 3: AI-Generated Political Content
This is the 2026 addition to an already difficult problem. Generative AI makes it trivial to produce large volumes of politically targeted content — articles, social media posts, images, even video — at near-zero cost. Political actors who previously needed a team of writers to run an influence campaign now need a single person and a laptop.
The key characteristics of AI-generated disinformation: it is fast to produce, cheap to distribute, personalised to specific demographics, and increasingly difficult to distinguish from genuine content.
What a Well-Informed Citizen Can Do
This is not an argument for avoiding news or social media. It is an argument for consuming them with deliberate strategies:
1. Primary Source First
Before sharing or forming a strong opinion on a political claim, find the primary source. If it is about a politician’s statement, find the original quote in context. If it is about a study, find the study. AI tools like Perplexity.ai can help you find primary sources quickly.
2. Seek Out Disagreement Intentionally
Allsides.com rates news sources across the political spectrum and shows the same story from left, centre, and right perspectives side by side. Reading all three versions of a story takes 10 minutes and produces a much more accurate picture than reading 10 articles from a single perspective.
3. The SIFT Method for Evaluating Claims
- Stop — before reacting, pause.
- Investigate the source — who published this, and what are their incentives?
- Find better coverage — is this reported by multiple independent outlets?
- Trace claims — follow statistics and quotes back to their origin.
4. Detect AI-Generated Content
Tools like Hive Moderation, GPTZero, and Google’s SynthID detector help identify AI-generated text and images. None are perfect, but an image that scores 95% AI-generated probability deserves scepticism before sharing.
The Structural Problem
Individual media literacy helps, but it does not solve the structural issue: billions of people are not going to change their media consumption habits based on a blog post. The deeper questions — about platform accountability, algorithmic transparency, and AI content regulation — are political questions that require political answers. Being informed about how the system works is the prerequisite to participating in those debates meaningfully.
Key Takeaway: Algorithms are not neutral. They are optimised for specific objectives — and those objectives often conflict with an informed electorate. The first defence is understanding the mechanism. The second is building deliberate habits that counteract it.
Be the first to respond