Is It Manipulation or Is It Design? Drawing the Line

Every conversation about ethics in this industry arrives at the same place. Somebody says the ad is manipulative, somebody else says it is just good design, and both are describing the same screen. The argument does not resolve, because neither side has offered a test — they have offered a verdict and an intuition.

The standard response is a list of virtues. Be honest, respect the player, do not exploit. All correct, all unobjectionable, and none of it helps the person looking at a specific mechanic on a Tuesday trying to decide whether to ship it. A principle that cannot be applied to a case is not doing the work a principle exists to do.

We have written before that the line between demonstration and manipulation is a continuum rather than a boundary, and that reasonable people place it in different spots. That is true and it is also an evasion if we leave it there. So this piece tries to draw it — not by defining terms more carefully, but by taking one mechanic and grading it until it turns.

Three tests get offered constantly. Each captures something real and each breaks on cases that obviously matter.

Intent. Did you mean to manipulate? This fails because almost nobody does. The teams producing the most questionable work are optimising a number, not plotting against anyone, and a test that acquits everybody who was not being deliberately malicious acquits nearly the entire industry. It also has the wrong shape: it asks about the state of mind of the person who built the ad rather than about what happens to the person who sees it.

Outcome. Did the player end up worse off? Better, because it looks at the right party, but unusable at the moment of decision. You cannot measure it until the campaign has run, and plenty of manipulative ads produce satisfied players who happened to like the game anyway. A technique is not rescued by the product turning out to be good.

Whether it converts. The industry's habitual answer, and not a test at all. It answers a commercial question and is entirely silent on this one. Conversion tells you the technique worked, which is the premise of the question rather than an answer to it.

What these have in common is that they ask about the wrong moment. The interesting question is not what you meant, or what happened afterwards, or how well it performed. It is what the technique is doing to the player while it runs.

Here is the one we actually use. Would this technique still work if the player could see exactly how it works?

Imagine a caption at the bottom of the screen describing the mechanism in plain language while the ad runs. *This puzzle is the real first level.* Or: *this puzzle is unwinnable and was tuned so you would nearly finish it.* Then ask whether the technique still does its job.

Some survive easily. A satisfying mechanic remains satisfying when you know it was designed to be satisfying — a well-made puzzle labelled *this was carefully tuned to feel good* is still a well-made puzzle. Clear feedback, honest difficulty, a mechanic drawn from the real game: explaining any of these changes nothing, because they were operating on something true about the product the whole time.

Others collapse immediately. An unwinnable puzzle labelled *you cannot win this* has nothing left. A fake mechanic labelled *this is not in the game* stops working the instant the caption appears. The technique was not pointing at a real property of the product — it was working on a gap in what the player could see, and the caption closes the gap.

That is the line, and it is not arbitrary. It identifies a genuine structural difference. Design works on something true and therefore survives inspection. Manipulation requires a blind spot, which is why it is destroyed by disclosure. You are not testing whether a technique is nice. You are testing whether it needs the player not to notice.

The test also sorts cases that intuition handles badly. An open loop — stopping the ad mid-action so the player wants to finish — survives disclosure comfortably, because *you were interrupted on purpose* does not remove the desire to complete the thing; the interruption pointed at a real game they can actually go and finish. A countdown timer showing a deadline that does not exist does not survive at all. Both are commonly described as psychological tricks, and only one of them is doing anything the player would object to on being told.

Abstract tests are easy to agree with and hard to apply, so here is the same mechanic at five settings. The technique throughout is the same: an ad built so the player comes close to finishing. Only the tuning and the truthfulness change.

Read it top to bottom and notice that no single step feels like a large moral leap. That is the point — and it is why teams arrive at the bottom of this table without anyone having made a decision they would describe as a decision.

Two things are worth drawing out of that table.

The first is that the technique never changes. Almost-winning is present in all five rows, and it is not the villain — rows one to three use it defensibly. This is why arguments about whether a *technique* is manipulative go nowhere. What moved across those rows is whether the near miss reflected something true about the product or was manufactured despite it.

The second is where the threshold actually sits, which is not where teams usually assume. It is not at deception in the ordinary sense — row four contains no false statement, shows real art, uses a real mechanic. What it does is produce a belief about the player's own chances that the design knows to be false. The lie is not about the product. It is about the player, and that is easy to miss precisely because nothing on screen is untrue.

It is also worth saying why that distinction is hard to hold onto in practice. Nobody in the room is tracking which row they are on. What they are tracking is a completion rate, and the instruction that produces row four is not *make this unwinnable* — it is *the drop-off is too high, make the near miss land harder*. Every row in this table is reachable from the one above it by a request that sounds like tuning, and the language of tuning has no vocabulary for the threshold. That is why the test has to be applied deliberately: the ordinary process of improving an ad walks straight through the line without the line ever being mentioned.

Three limits, stated because a test presented as decisive invites more confidence than this one has earned.

It does not scale down to children. The whole test assumes a player who, if informed, could act on the information. Persuasion knowledge develops with age, and an argument resting on what an adult would do with disclosure does not transfer to an audience that has not built those defences. For that placement, assume the bar is higher and that the format's advantages are the wrong advantages.

It says nothing about aggregate effects. A single ad can pass on its own and still contribute to something worse — an attention economy that a person did not consent to in general, even if each individual instance was fair. This test is for the mechanic in front of you. It is not a defence of the industry.

Row three is genuinely contested. We place the threshold between three and four. Someone could reasonably argue that selecting your hardest honest configuration is already engineering the conclusion, and we would not think them unserious. What we would insist on is that the disagreement is now specific — a named row, with a stated reason — rather than two people exchanging intuitions about whether an ad felt manipulative.