Ethics & consent

Three separate things, three separate decisions

Most apps ask once and treat every later use as covered. We split it, because these are genuinely different questions and you might well answer them differently.

What we do

• Record which questions you answered, whether you were right, and how long you took.
• Record which specific thing you confused with which — 你 for 妳, a second tone for a third.
• Record how long the gap was since you last saw an item, and whether you still knew it.
• Store the trial-by-trial data from any cognitive task you choose to take.
• Show it back to you, and to your teacher if you are in a class — with one commercial exception, stated rather than hidden: on a free account four of the seven facets (tone, listening, speaking and writing) are withheld by plan rather than by evidence. Their names are still shown, so you always know what is being withheld, and your teacher sees no more of them than you do.
• Let Mandarin Temple staff reach your record where running the service requires it — a support request, a safety concern, or administering a study you joined. Your cognitive-task results stay pseudonymous even to us.

What we do not

• Use your camera or microphone to watch or listen to you.
• Track your eyes, your face, or your emotions.
• Log what you type outside of an exercise you are doing.
• Follow you around other websites.
• Sell any of it, ever, to anyone.
• Claim any of this diagnoses attention, intelligence, or a learning difficulty.

If you are under 18

If your account is managed by a parent or guardian, you cannot agree to research on your own — they have to, from their account. This is not a formality we can be talked out of: the server checks who is asking and refuses, whatever the page happens to show.

A guardian can withdraw at any time, and withdrawing removes the child from any future analysis.

The promises we can actually keep

We will not show you a number we cannot support.

Every figure in the Lab carries how much evidence is behind it, and refuses to render below a threshold. You will see “about 20 more retrievals and this becomes meaningful” rather than a confident percentage built on four answers.

We will not call a guess a measurement.

We say “task engagement stability”, not “attention”, because a browser cannot see attention. Where a pattern has several possible explanations, we say so instead of picking the flattering one.

A bad run is marked, not deleted.

If you were interrupted or pressing buttons at random, that run is stored and labelled unusable. Silently dropping it would make everything else look better than it is.

Withdrawing is a real thing, not a setting that does nothing.

Turn research off and your data stops being included. Delete your account and the whole record goes — including the research rows. We do not keep a copy under a research exemption.