The Manipulation Matrix

What it is

Hooked is a build manual. Four chapters explain how to engineer a product people return to without deciding to, and then one chapter asks whether you should. The Manipulation Matrix is that chapter's instrument, a pre-launch self-test a maker runs before shipping, and it occupies pages 165 to 177 of a 246-page book about how to do the thing it is assessing.

Eyal starts by refusing to treat manipulation as an accusation. The definition is flat by design: manipulation is "an experience crafted to change behavior" (p. 165). Weight Watchers arrives immediately afterwards as the normalising case, a manipulation almost nobody objects to, because the behaviour it engineers is one the participant already wanted for themselves. That move does real work in the argument. It establishes that the interesting question is not whether a designer is shaping behaviour, since every designer is, but whose ends the shaping serves.

From there the two axes. Does the product materially improve the user's life. Would the maker use it. Cross them and you get four positions. The facilitator improves lives and uses the product. The peddler believes the product improves lives but does not use it. The entertainer does not improve lives and uses it anyway. The dealer neither improves lives nor uses the thing they built, and Eyal names that quadrant with the comparison intact, a casino operator or a drug dealer. Exploitation gets its own definition, and it is the dealer quadrant restated: "creating a product that the designer does not believe improves users' lives and that he himself would not use" (p. 174).

The framework's most interesting claim is not ethical at all. Eyal argues that facilitators have "the highest chance of success because they most closely understand the needs of their users" (p. 176), which makes the top-right quadrant both the honourable one and the profitable one. A test that promises the right answer is also the winning answer is a test that costs its taker nothing.

Now the problem, which is structural rather than incidental. Both axes are self-report. A maker asks themselves whether their product helps people and whether they would use it, and self-report has one predictable failure mode: it cannot catch anyone who believes their own pitch. The dealer quadrant is defined as a place no honest respondent will ever locate themselves, since the definition requires the respondent to admit they think their product is bad and that they avoid it. The vault's reading of Hooked puts this plainly, that the matrix is unfalsifiable by construction.

Eyal's hedges are where the framework quietly concedes its own reach, and they are worth reading rather than skipping. He cites a roughly 1 per cent addiction rate among slot-machine players (p. 170), a figure that arrives in the middle of an ethics chapter and does the work of making engineered compulsion feel like a small-tail problem. And on accountability, when the test returns a poor answer, the book puts the weight on the person at the other end: users bear "ultimate responsibility" for their own use, while the maker has to "live with the moral repercussions" (p. 176). No audit, no reviewer, no consequence that exists outside the maker's conscience.

This page carries not-a-research-claim, which on this site means a framework rather than a testable effect. That is not a demotion. A design ethic proposed by a practitioner is a legitimate thing to cite, attributed by name, and replication is not the standard it should be held to. The standard it should be held to is whether it does the job it claims, and the objection above is about that rather than about evidence.

In effect

Weight Watchers, and manipulation as a neutral word

Eyal's opening example is the one that sets the terms of the whole chapter. Weight Watchers is an experience crafted to change behaviour, which makes it a manipulation by his definition, and hardly anyone treats it as an ethical problem. The reason is that the outcome is one the participant wanted before the programme arrived. Fix that as the baseline and a great deal of persuasive design looks unremarkable, which is the point of putting it first. The move also narrows the question the matrix will go on to ask, from "should this exist" to "does the maker think it helps".

What the makers do at home

The closest thing to an external test of the first axis comes from a book Eyal did not write. Adam Alter reports in Irresistible that some technology leaders limit the use of the products they build at home. Read against the matrix, that is the "would I use it" question answered by the people with the best available view of the product, and answered no. It is reporting rather than an experiment, and it should be cited as reporting. It is still the sharpest available check on an axis that otherwise has none, and the axis fails it.

Jake Harriman, Nuru International, and the profitable quadrant

Eyal's own exemplar is Jake Harriman, a former Marine who founded Nuru International, an anti-poverty nonprofit, placed squarely in the facilitator quadrant (pp. 168-169). The choice is instructive in a way the chapter does not acknowledge. An anti-poverty organisation is an easy case for both axes, and product teams shipping engagement features are the hard one. Pair the exemplar with the claim that facilitators succeed most often and the test starts to read less like an ethical constraint and more like an argument that the ethical path is also the commercially correct one. Any product manager who runs the matrix on a Friday afternoon will finish it in the top-right quadrant and ship on Monday.

What it does not say

It is not a finding, and it cannot be one. The vault's reliability audit of Hooked records the Manipulation Matrix as conceptual scaffolding built from case studies and consulting practice, alongside the Hook Model, the Habit Zone and Habit Testing. No study tests it, because there is nothing in it to test. Cite it as Eyal's map, named in the sentence, never as evidence that a product is or is not exploitative.

It does not provide an outside check, and it does not claim to. The maker is the only respondent, the only judge, and the only person who sees the result. A test with no proctor is a test people pass by default, and the quadrant designed to catch the worst behaviour is defined so that nobody who behaves that way will tick it.

It does not survive contact with a structural account of the industry. Tim Wu argues that the attention merchants of the past 190 years were not villains but ordinary operators responding rationally to a business model that pays for harvested attention. That account has no place to sit inside a two-by-two, because the matrix locates the entire ethical question inside one person's private judgment about their own product. If the incentive is doing the work, individual conscience is the wrong unit of analysis.

The 1 per cent figure is Eyal's, not ours. The slot-machine addiction rate at p. 170 has not been traced to a source by us, and it appears in his prose without a named researcher or year, which the vault flags as the recurring pattern in this book. Do not repeat it as an established rate.

Eyal himself has moved. Indistractable, published in 2019, is in effect the counter-manual to the book this matrix appears in, written by the man who placed himself in the facilitator quadrant. Its introduction concedes that the ideas are now used to distract people as well as to build healthier habits. That is not a repudiation of the matrix. It is the author noticing what the matrix did not catch.

Who developed it

Nir Eyal, in Hooked (2014), pages 165 to 177. One correction belongs on the record: the vault note's frontmatter listed Tristan Harris as originator, at medium confidence with no year, while the body of the same note correctly described the matrix as Eyal's. Harris is a design ethicist whose position is close to the opposite bet, that the fix has to be structural rather than self-administered, and his name appears to have been pulled from a neighbouring note rather than from any source. It is the same famous-name-nearby error the vault has now logged seven times, and it is corrected here. Nir Eyal has a page on this site. Tristan Harris does not yet.

---