# First Discovery concept-poll charter

**Round:** Art Direction 01
**Status:** Parent-preview advisory poll; adult aggregate collection approved
**Question:** Which two visual worlds make you most curious to meet Mosskip?

## What we borrowed

The process is structurally inspired by the transparency of long-running game
community polls, especially the official Old School RuneScape distinction
between an **opinion survey** and a **lock-in poll**, the proposal material shown
before a vote, a skip option, visible total participation, and public option
counts. See the official [Old School polls index](https://oldschool.runescape.com/polls/latest)
and an example [lock-in poll with question-level counts](https://oldschool.runescape.com/polls/2026/1758).

First Discovery is independent and not affiliated with or endorsed by Jagex.
This poll does not copy another game's visual design, wording, world, or brand.

## What this vote can decide

This is an **opinion poll**, not a binding production vote. Aggregate preference
may shortlist two contrasting directions for proper print, accessibility, and
human-art-direction studies. The highest count does not automatically become
the final style.

A direction can be held back even with strong support when it fails grayscale
printing, color-vision independence, small-card readability, originality,
provenance, cost, accessibility, or child-safety review. If the team does not
follow the leading preference, the eventual decision note must publish why.

The poll does not change gameplay rules, approve a public release, grant an art
license, authorize fundraising, or decide which artist is commissioned.

## What every voter sees

Every concept uses the same Mosskip subject, scene purpose, portrait crop, and
AI constraints so the main experimental variable is visual style. Every option
shows:

- its stable number and name;
- the unedited generated image used in this round;
- the design hypothesis;
- strengths and known limitations;
- the normalized generation prompt and shared constraints;
- the human visual-QA result;
- a visible **AI-generated concept, not final art** label.

The canonical record, prompts, file hashes, alt text, limitations, and decision
rule live in `art-concepts-round-01.json`.

## How participation works

1. A responsible adult reviews and facilitates the local page.
2. The participant sees all six concepts in an alternated order.
3. They may select one or two concepts, or skip the question.
4. The participant steps away before the adult records the ballot.
5. The page clears the individual selection and retains only temporary counts.
6. The adult reviews the temporary tally and deliberately presses **Submit this
   tally online** once. A green accepted receipt is required.
7. An offline facilitator may instead export the aggregate JSON and later use
   the separate reviewed-file handoff.

The page asks for no name, age, birth date, grade, school, location, account,
contact detail, device identifier, timestamp, photo, recording, handwriting,
or written comment. The hosted page makes no request on load, during selection,
or while a ballot is recorded. Only the explicit adult submit action sends the
fixed aggregate. Nothing is stored between page loads. The downloadable bundle
remains fully offline. This is not a representative scientific survey.

The separate adult handoff accepts only the fixed round ID, the six concept
counts, ballot total, and skipped total. Missing or extra fields fail. The
database adds accepted values directly into one cumulative round-total row; it
does not retain a submission row, timestamp, individual ballot, identity,
demographic, free text, child-authored content, IP address, or user-agent value.
MakeJoy does not read or log connection identifiers. Cloudflare necessarily
processes ordinary connection data to deliver and protect the site. See
`NETWORK-REVIEW.md` for the reviewed boundary and remaining provider gate.

Because there is no identifier or submission ID, the endpoint cannot detect the
same temporary tally or aggregate file being deliberately sent twice. The page
locks after an accepted direct submission, but reloading can bypass that local
guard. Facilitators must submit each session once. One request may represent at
most 200 facilitated ballots.

## How results must be published

Publish the round ID, exact question, opening/closing context, total ballots,
skipped ballots, raw count for every option, percentage of all non-skipped
ballots selecting each option, method, limitations, and the decision that
followed. Because each person may select two options, percentages do not need to
sum to 100%.

Never publish or retain individual ballots. Never describe a small convenience
sample as “what kids want.” Do not hide skipped ballots or options with zero
support. No result is valid until at least one real aggregate has been received;
the prototype contains no example votes that could be mistaken for community
data.