Prolific wrapper: consent, instructions, comprehension check, debrief, mobile block

[htsukamoto5]

Phase 1 — unblocked. Split out of #7, which is now scoped to the completion-code plumbing only. This issue is the participant-facing wrapper: the largest single piece of unblocked work, and 100% absent today.

The only repo-wide matches for prolific|consent|debrief|comprehension are prose in README.md and the HTML header comments. Nothing is implemented.

Required	Status
Consent form	missing
Instructions	missing — only the one-line prompt on the game screen
Comprehension check	missing
Debrief	missing
Fullscreen / mobile block	missing
Timing estimate shown to participant	missing

Completion codes and the submission redirect are not here — see #7. This issue has no dependency on #7 and can be built in parallel with everything else in phase 1.

Dyad-specific, beyond the standard list

• Instructions must set partner expectations — that a real person is waiting, and that abandoning costs someone else their payment. In dyadic work this framing measurably reduces mid-task dropout, and it is free.
• Instructions and comprehension check must run BEFORE the lobby. Currently nameTrial then lobbyTrial leaves one participant reading nothing while the other waits. Anything slow goes before pairing — the files already apply exactly this reasoning to preloadTrial, so extend it.

Minor but real

• nameTrial collects a free-text display name, which is participant-entered PII shown to a stranger. Assign neutral labels ("Partner A") or filter. Cheaper than an IRB conversation.
• jsPsychPreload defaults to halting on error. One 404 in assets/tangrams/ leaves a paid participant dead in the water with no completion code. Set on_error / continue_after_error deliberately.
• Free-text chat between strangers is user-generated content. Worth a line in consent and a decision on moderation/reporting before IRB asks.

See READINESS.md section 4.

[htsukamoto5]

Suggested screening design and timeline ordering

Recording the design settled while working through the Prolific payment rules. Tentative, but concrete enough to build against. The completion-code and data-field plumbing lives in #7; this is the participant-facing side.

Ordering

consent
  → short honest description (duration, real partner waiting, cannot be paused)
  → prior-exposure item
  → commitment gate  ──[declines]──> screened-out exit
  → full instructions
  → comprehension check
  → lobby

Three constraints drive this, and they are not all obvious:

1. Consent first, or there is no permission to store the screening responses.
2. The gate goes after a short description, not after full instructions. An informed commitment decision needs only the duration and the partner obligation — not the task mechanics. Everything before the gate counts as screening time, so the description has to stay tight.
3. Full instructions and comprehension still precede the lobby, so nobody sits waiting while their partner reads. This is the same reasoning the files already apply to preloadTrial.

Why brevity before the gate is load-bearing

Prolific's screen-out reward is duration-scaled — $0.14 minimum for a screener of one minute or less, so ~$0.28 at two minutes, ~$0.42 at three. And there is a feedback loop: longer pre-gate content → higher required screen-out reward → more attractive to reward farmers → faster consumption of the screen-out slots → higher risk of the study auto-pausing mid-run.

Keeping the pre-gate section under about two minutes is what keeps the reward at the boring end of the range where farming is not worth anyone's time. This is the main reason full instructions moved to after the gate rather than before it.

The prior-exposure item

• Ask about prior experience with tangram / reference-game tasks, embedded in a short list of unrelated task types rather than asked alone. Spotlighting tangrams both primes the participant and signals what we care about.
• Screen on prior exposure to the stimuli or task, not on knowledge of the theory. Someone who has done a tangram task before may arrive with entrenched labels already formed ("the ice skater"), which starts them near asymptote and flattens the very slope we measure. Knowing that descriptions shorten is a much weaker threat and is hard to act on while genuinely trying to coordinate in real time.
• Ask a more specific version again at debrief as an honesty check, noting that payment is unaffected either way.

Note that this is the only available lever for prior exposure: Prolific's exclude-previous-participants feature covers our own past studies, not other labs'.

The commitment gate

• Ask directly whether the participant intends to complete the study now, with a genuinely costless paid exit.
• Framing determines whether this is a filter or a ritual. "Will you complete this study?" gets ~100% yes from social desirability. Something closer to "if now isn't a good time, you can leave with $X and no penalty — that genuinely helps us" reframes leaving as cooperative rather than as failure, which is the only version that catches the people it is designed for.
• Expect a small yield, and build it anyway: the exit costs ~$0.14–0.28, while one prevented mid-task abandonment saves two payments, a dyad, and a recruitment slot.

Why this is worth the effort at all: returns are not rejections. Participants may return a study at any time without penalty, and Prolific advises researchers to return rather than reject precisely to protect approval rate. So no approval-rate prescreener catches serial abandoners — the platform is structurally blind to the exact failure mode that costs us most, and an in-study gate is the only thing that addresses it.

Interaction with the partner-expectation framing

The gate and the "a real person is waiting, abandoning costs someone else their payment" framing already in this issue reinforce each other. Anyone who passes the gate knows that everyone else who passed it also committed, which adds a normative cost to bailing on top of the informational one. They should be written as one coherent sequence rather than two separate interventions.

See READINESS.md sections 4 and 5, and #7 for the codes and reward configuration.