TAMIZFor reviewers · UNESCO Youth Hackathon 2026

Ten megabytes. Twelve messages. You can't check everything.
That's the point.

TAMIZ is a behavioural media and information literacy instrument, played as a 16-bit phone game. It measures what people do with deceptive information rather than what they say they would do. The interface is in Ecuadorian Spanish. This page is the doorway: what you are about to play, and the twelve words you will meet on screen.

Play the instrument (Spanish)Or play it in English (review copy)

A review copy: the same twelve messages, the same engine, translated. Seven people have already finished that round — you are the eighth, so when you finish, the group result unlocks and you see what the eight of you produced. It reports nothing to us and no measurement in the submission comes from it.

Eight minutes. No account, no install, works offline.

What you are about to play

Three days on a simulated phone. Twelve messages arrive across four surfaces — a messaging app, SMS, email, a social feed — and ten megabytes of prepaid data that do not stretch to checking them all.

You stamp each one, you are asked what made you decide, and then you act: forward, archive, or warn that it is false. Nothing tells you whether you were right. For three days there is no tick, no cross, no score.

At the end you get your own result. If you are playing inside a round, the group result appears only when the eighth person has finished — and never before.

The twelve words on screen

You do not need Spanish to follow the game, but these twelve carry the mechanics. Everything else is a message you can skim.

LO CREOI believe itThe verdict is stamped on the message, and it is recorded first and separately from everything else.
NO LO CREOI don't believe itThese two, plus the third, are what d′ and c are computed from.
NO SÉI don't knowAn honest answer, kept as one. It is never forced into a yes or a no.
Fuente −1See the source — costs 1Who registered the domain, when, and whether it has an author.
Contrastar −2Corroborate — costs 2Does anyone else report this. Twice the price of the source.
Detector −2Synthetic-content detector — costs 2Two items are calibrated traps: it answers correctly and still misleads.
ReenviarForwardForwarding something you yourself marked false is the first dissociation.
ArchivarArchiveDoing nothing is an action too, and on a true public-safety alert it has a cost.
Avisar: es falsoWarn that it is falseDenying something true is scored as its own kind of damage.
¿QUÉ TE HIZO DECIDIR?What made you decide? Touch that part of the message.One tap, after the verdict, with nothing highlighted. This is the measure we think matters most — §7.2 of the document.
REENVIADO MUCHAS VECESForwarded many timesA label the real platform already shows. Free to read. Whether you read it is itself the measurement.
MEGASMegabytes — you have ten, for twelve messagesChecking everything would cost more than forty. That gap is the instrument.

Why it is not in English

The twelve messages are threat patterns documented in Ecuador by named sources — bank impersonation, a held parcel, a job offer abroad, a real civil-protection alert. Translating them would not change the language. It would change what is being measured.

An instrument for Ecuadorian young people, written in Ecuadorian Spanish, is not a limitation we ran out of time to fix. It is the design. Every figure in the submission document is translated in its caption, and this page translates the controls — so the mechanics are readable without the stimulus being altered.

What it separates that a questionnaire cannot

Believing ≠ sharingShe marked it false — and forwarded it anyway. She knew.
Believing ≠ not clickingHe disbelieved it and opened the link "just to look".
Being right ≠ having methodShe got them right without spending a single megabyte.
Being right ≠ knowing whyHe was right, and pointed at the emoji as his reason.

Four people answer a survey identically. Each one needs a different response, and none of it survives being asked.

What it claims to measure, and how new each claim is

The competency model, the evidence model and the threat surface are published as structured data rather than described in prose. 22 competencies across six areas, 14 of them with a measure attached, over a mapped surface of 40 documented threat patterns.

Access · 2Triage of attention · recognising an information void
Evaluation · 9Discrimination, response bias, grounding, source, recontextualisation, social proof, emotion, metacognition
Creation · 3Civic amplification · correcting without amplifying · belief–action coherence
AI · 4Deference to automated verdicts · synthetic origin · provenance · algorithmic awareness
Ethics · 2Anticipating harm to others · data care
Security · 2Operational hygiene · out-of-band verification

And the part we would rather you read from us

Every measure carries a novelty field in the data itself, so the claim is checkable rather than rhetorical:

3establishedAlready measured elsewhere. d′ and c are standard signal detection theory; we did not invent them.
3lab onlyDemonstrated in laboratories, never deployed as a field instrument.
8no known instrumentWe could find no instrument that collects these. That phrasing is deliberate — the absolute form invites refutation by a single link.
Read the full method

All 22 competencies, all 14 measures with their novelty, the 40 threat patterns and the eight modules — rendered from the same file the instrument reads, with the raw JSON at the bottom of it for machines.

What is not done

The instrument is designed under Evidence-Centred Design and built — never validated. No external group has played it yet. The submission document says so on its first page rather than leaving it to be found.

Play the instrument (Spanish)