← About

How We Communicate

A quarterly account of how this site communicates and how it performed. A site arguing for transparency about persuasion techniques should be transparent about its own.

This page also demonstrates the approach it describes. Tap any underlined word to see its precise definition, and open the "Deeper" box for the full cognitive-science citations. The main text works on its own.

The reports

The first report covers 30 May to 30 September 2026: every correction we made, the hooks we did not use, and the money: none raised yet, the running costs carried personally. It has no reach figures. Our analytics counted properly only from 22 September, and nine days say nothing about reach. The next report, in January 2027, covers October to December and carries the first figures, including the disappointing ones.

What we publish quarterly

What we count. We do not test headline variants, and there is no newsletter yet. The site's self-hosted, cookieless analytics counts page views and a few anonymous events: how far people get through each demo and whether they finish it, a click on a demo's "one thing to do now" link, which links to other organisations are followed, a press of a share button and the visits a shared link brings, and the Yes or No under "Did this page help?". The privacy page describes each one. Nothing identifies a reader, and nothing records what anyone typed. If we ever use an activating headline, we say so — and we note whether we think it was justified or crossed a line.

What we do not do. We do not use outrage hooks — headlines like "This is destroying your children" or "The shocking truth about Facebook" — even though we know from the research documented on this site that they would perform better. We do not buy followers or engagement. We do not use dark patterns to inflate newsletter sign-ups.

Corrections. Every factual error corrected is logged publicly. Original claim, correction, date, and reason. See the corrections log.

How we explain complicated things

This site documents how attention-optimising platforms work. The mechanisms are real, but they are not simple. We face a design challenge: how do we explain a complex system to people with vastly different knowledge and time?

Research in cognitive psychology shows that presenting everything at once causes cognitive overload — the facts do not land. But oversimplifying is equally harmful: unnecessary scaffolding actually worsens understanding for people who already know the topic. So there is no single correct level of explanation. It depends on what the reader already knows.

We solve this with three layers:

Layer 1 — the main text. Every page explains the core idea in plain language, short sentences, and concrete words. You can understand the entire argument from the main text alone. No jargon, no parenthetical asides.

Layer 2 — term definitions. Where the main text uses a plain word for something with a precise technical meaning, that word is underlined and tappable. Tap it to see the academic term, a rigorous one-sentence definition, and sometimes a reference. The plain word stays in the prose; the precision is available without cluttering the page.

Layer 3 — evidence and depth. At the end of each section, collapsible "Deeper" boxes contain dense exposition, peer-reviewed research, contested points, and how we know what we claim. You do not need to open them to understand the argument. But if you are fact-checking or teaching, they are there.

The science behind this. This approach is grounded in cognitive load theory (Sweller), advance organisers (Ausubel), and scaffolding research (Bruner, Vygotsky). When you read the clear main text first, you build a mental schema. That schema lets you integrate details without overload. For novices, the spine is enough; the depth is available if needed. For experts, the spine is quick to read, and the depth is where rigour lives. No single level works for everyone — but this structure serves both.

Deeper — the cognitive science, and why one reading level can't serve everyone +

The reason for layering rather than picking a single reading level is a specific, well-replicated finding: the expertise-reversal effect. Kalyuga, Ayres, Chandler and Sweller (2003) showed that instructional support which genuinely helps novices — extra explanation, worked detail, scaffolding — becomes redundant and can actively worsen performance for people who already know the material, because processing the now-unnecessary support consumes working memory they would otherwise spend on the task. So "write it simply for everyone" is not a safe default: it taxes experts, while "write it rigorously for everyone" overloads novices.

Layering resolves the dilemma instead of splitting the difference. The Layer 1 spine is the advance organiser (Ausubel) that lets a newcomer build a schema; the Layer 2 terms and Layer 3 depth are there for the reader who wants or needs them, and invisible to the one who does not. This is the broader cognitive-load-theory programme (Sweller; Paas, Renkl & Sweller, 2003) applied to a public website rather than a classroom. It is also why we resisted the obvious "simple version / expert version" split: that forces every reader to self-classify and still serves each badly at the edges.

Sources

• Kalyuga, S., Ayres, P., Chandler, P. & Sweller, J. (2003), The Expertise Reversal Effect, Educational Psychologist 38(1).
• Paas, F., Renkl, A. & Sweller, J. (2003), Cognitive Load Theory and Instructional Design: Recent Developments, Educational Psychologist 38(1).

How this site is made

This site is made with the help of AI tools. They help find and read sources, draft and edit text, translate pages, and review the work: one AI model checks what another wrote, and automated checks compare every translation's numbers, links and sources with the English. The note at the foot of every page says so.

AI tools make mistakes, so the site is built to catch them. Every factual claim links to its source, and the claims that carry a number, a date, a named study or a quotation are tracked, with how each was checked, in the claim ledger — published on this site, so a reader can check the checking. Those source checks were themselves made with AI tools and re-checked by a second model; the pages a person has read carry the date in their footer. A person decides what is published and is responsible for it. A translation goes live only after a native speaker has reviewed it — with one exception, an edition published while its volunteer reviewers are still reading it, which says so on every page (how that works).

Mistakes still get through. When we find one, or a reader tells us about one, we correct it and log it on the corrections page. Tell us if you spot one.

How to use this site

If you have 5 minutes: Read the main text of one page. You will get the core argument.

If you have 15 minutes: Read the main text, and tap underlined terms that confuse you. The definitions will clarify.

If you are fact-checking: Start with the Sources block at the foot of the page. Every factual claim links to its source on its page — to peer-reviewed research, an official investigation or a primary document where one exists, otherwise to established journalism, with any industry estimate labelled as one — and every claim that carries a number, a date, a named study or a quotation is tracked in the claim ledger, which lists every checked claim with its source, so you can verify it.

Our engagement policy

We welcome: factual corrections (always, regardless of how they are delivered), genuine questions, substantive disagreement including from researchers who think we have got something wrong.

The objections we hear most often are gathered and answered, in public, on the Yes, but… page.

We do not engage with: personal attacks on the people running this site, coordinated harassment, messages that are factually false and decline to engage with evidence when it is provided, content using the manipulation techniques this site documents.

Comments and correspondence. This site has no comment system. Correspondence reaches us by email, and the policy above applies to it. That policy is published in advance and is publicly reviewable, which is what distinguishes it from the opaque, algorithmically applied moderation this site criticises in platforms: anyone who thinks we applied it wrongly can say so by writing to us. If comments are ever opened, the same commitments apply, and what was held, removed, or corrected will be reported here.

On using social media to discuss social media

This site uses social media platforms to share content about social media's harms. We acknowledge the irony. We do it because that is where people are. The argument for selective, informed use is more honest than abstinence. We try to demonstrate this in practice: using the platforms, understanding what they are doing, being transparent about the choices we make while doing so.

← About