Open Lab method
How an Open Lab trial is constructed
Every experiment in the Open Lab shares the same skeleton: commit before collection, lock before feedback, version everything, and report the aggregate rather than the anecdote.
Trial sequence
Step 1
A target is committed
The server draws one image at random from the active pool and records it against your trial before you begin.
Step 2
You describe your impressions
Free text plus a small set of structured impressions — indoor or outdoor, natural or artificial, water, people, movement.
Step 3
Your response is locked
The text is hashed with SHA-256 and stored immutably. This is the point of no return.
Step 4
You judge four candidates
The committed target and three decoys are shown in a randomised order. You pick the closest match.
Step 5
The target is revealed
Only after your decision is recorded. Your trial then joins the aggregate record.
What is recorded
Each trial stores an opaque trial code, the protocol version, the target pool version, the committed target, the locked response and its SHA-256 hash, and the judging decision with its randomised display order. Responses and judging decisions are immutable at the database level once written.
Participation is pseudonymous. A random identifier is stored in your browser so a trial can be continued and attributed to one session. No account, name or email is required, and exports for analysis carry a derived pseudonym rather than that identifier.
Known limitations of this first iteration
- One trial cannot distinguish ability from chance; only large aggregates carry information.
- Participants are self-selected and unsupervised, so this design is exploratory and cannot support confirmatory claims.
- Judging is done by the participant, which introduces a known bias that later protocol versions will address with independent judges.
- The target pool is finite; repeated participation can lead to familiarity with the pool.
How results will be reported
Chance performance in this design is 1 in 4 (25%). Aggregate hit rates will be published with confidence intervals and the total number of trials, including trials that were started and abandoned. Outcomes at or below chance are reported in the same place and with the same prominence as anything else.
Taking part does not require believing in anything. Chance alone produces hits at the expected rate, and a correct trial is exactly what chance looks like part of the time.