Live status · Read the latest
monad-knowledge Wi-Fi sensing lab · FIIT STU
open EXP-012 instrument

Visible Points and Dwell Compliance

In plain words

When participants can see their score, do more of them take part, and do they still stand still for the full 30 seconds each measurement needs?

scan hold still 30 s walk scan hold still 30 s walk scan hold still 30 s record duration and check movement independently
One probe: scan a card, then stand still for thirty seconds while the radios measure, then walk on. A visible score must not shorten the standing-still part, and the design watches for exactly that. Schematic of the design, not a measurement.
Why it matters
A reward can raise participation while quietly ruining the measurement: a hurried dwell scores the same as a proper one and destroys the data. That failure would otherwise read as success.
How it is done
Fix the before/after boundary now, before visible scores ship. Four measures from data the backend already records: dwell duration, dwell motion in the radio, completion rate and abandonment.
Where it stands
Preregistered with no attributed experiment data. The actual participant-visible score exposure boundary still needs verification; it cannot be inferred from the card's old 'not armed' text.

Asks, formallyMaking a participant's score visible and accruable changes participation volume and/or the quality of the 30-second stationary dwell that every probe quest rests on.

FoundThe preregistration fixes a before/after comparison for participant-visible scores.

StandingStatus not declared

RoomFIIT STU library — open and silent reading rooms, the ten-node AX210 fleet, participant handsets running the monad app

Visible Points and Dwell Compliance

Current assessment (2026-09-15)

The preregistration fixes a before/after comparison for participant-visible scores. No experiment data are attributed. The exact deployment/exposure timestamp was not verified in this review. Six revised quests were published September 15, changing instructions and identifiers; this is a separate protocol change that must be recorded in any participation comparison.

Do not infer that nobody has seen a score from a September 1 description of the application. The exposure record needs the deployed feature and the participant's app version. Quest revision is now a potential confound. CSI motion during a dwell can come from bystanders, and completion timestamps alone cannot identify which person moved.

Next evidence needed: Recover the actual exposure timestamp before dividing arms. Stratify by quest revision and app version, exclude operator and backfilled award records, and distinguish a dwell-motion proxy from independently verified participant compliance. Preserve the original preregistration boundary instead of backdating it.

Basis: Thesis Narrative, the September 10 synthesis, and the September 15 register review. See the review record for evidence sources and outstanding checks.

Earlier protocol and dated results

The methods and observations below retain their original scope. Read their readiness statements and proposed next steps as historical. The current assessment above records what has changed; frozen thresholds have not been retrospectively revised.

Why this card exists before the feature does

IP-145 Phase 6 makes a participant's score visible and accruable for the first time. This card is Phase 5, and it blocks Phase 6: the boundary is written down first, because a reward armed and then studied is not a comparison.

That is not a hypothetical caution. It already happened once. Points were armed on 2026-08-26, and templates.py records the cost in its own source: the CSEDU paper's §6 pre-declared a first session "with the points mechanic at zero throughout", and arming from day one removed the unarmed arm that M1, M4 and M6 were to be contrasted against.

Why a baseline still exists

The 2026-08-26 arming was partial. Points became a number printed on a quest card, and nothing else:

  • User has no points column and QuestEnrollment had none before IP-145.
  • QuestController returned points_earned in the completion response and stored nothing.
  • There is no profile, no total, no history and no leaderboard anywhere in the app.

So no participant has ever held a score. The period up to Phase 6 is therefore still a usable announced-but-never-delivered arm, and Phase 6 is the instant it ends.

The arming boundary

Arm instant: the deploy that ships IP-145 Phase 6 (GET /api/me/stats plus the app's profile screen). Record the exact UTC timestamp here on the day, from the deploy, not from the merge.

armed_at: <not yet armed — fill from the Phase 6 deploy>

Everything before that instant is arm A (announced). Everything after is arm B (delivered).

Measures, declared now

Split at armed_at. All four come from data the backend already records, so none of them needs new capture.

# measure source why it is the one to watch
M1 dwell duration per probe step quest_step_completions.completed_at − started_at The protocol asks for 30 s. A participant who hurries scores the same and destroys the measurement.
M2 dwell motion fleet CSI over the dwell window The independent check on M1. A dwell that was not stood still shows as a motion signature whatever the timestamps say.
M3 quest completion rate quest_enrollments.status Volume, the effect a reward is supposed to have.
M4 abandonment rate and step reached status = abandoned, quest_step_completions count Distinguishes "more people start" from "more people finish".

M1 and M2 together are the point of this card. M3 going up while M2 degrades is the failure mode, and it is the one that would otherwise be read as success.

Exclusions, declared now

  • Backfilled ledger rows. Version20260901150000 stamped points_awarded on already completed enrollments from the quest's current value, because nothing recorded what those walkers were promised. Their signature is awarded_at equal to completed_at to the second. Exclude them from anything that treats an award as a stimulus.
  • Operator takes. audience = operator. The operator knows the design, so they are not a participant observation.
  • Quarantined capture sessions, on the register's standing rule.

What the corpus could not tell us

Searched 2026-09-01 and re-run independently rather than taken on trust: claim_search("gamification incentive participation reward data quality") returns zero claims, and search_vault with graph_enhanced=True returns three passing mentions with no effect size: draghici2019_b7bf ↗, darsena2023_50b7, gubbi2013_fab5. coverage_report(axis="domain") has no HCI, behavioural-science or citizen-science bucket.

Two papers were acquired on 2026-09-01 to start closing that: mekler2013_169c ↗, a controlled experiment isolating points as a game element, and brhlmann2020_a0e6 ↗. Both are status: new and neither has been read. Neither may be cited until it has been.

Carry this caveat: even a well-stocked gamification corpus measures annotation or survey quality, not dwell compliance in a physical sensing protocol. The failure mode this card watches for may be genuinely unpublished, which is exactly why it is being measured here rather than looked up.

Owning work

  • IP-145 Phase 5 (this card) blocks Phase 6.
  • Feeds §5.3 and §6 of the CSEDU 2027 paper, whose contribution is the instrument and the participant protocol.

Provenance

—
not recorded

Data types

  • quest-enrollment
  • quest-step-completion
  • ble-rssi
  • csi-amplitude