Labs: trying out new ways to show and compare smells

The 11 experiments

1. Status: EXPERIMENTAL

   Representation Lab

   How could a smell record be set out as open, structured data?

   What it does: Puts each smell's existing public fields side by side in one open record: its descriptor words, settings, feelings, colour, texture, temperature and sound words, and related smells. It is a way of laying out what is already written, not a scientific code for a smell.

2. Status: EXPERIMENTAL

   Cross-Sensory Lab

   Which colours, textures, temperatures and sounds do SmellBook's records link to each smell?

   What it does: Lines up the colour, texture, temperature and sound words already written in each record, so you can browse smells by them. These are editorial and cultural associations, not measurements.

3. Status: EXPERIMENTAL

   Prediction Lab

   Which pairs of smells might be related but are not linked yet?

   What it does: A fixed, published rule suggests a pair when two records that are not linked share at least two descriptor words and also match on at least two other listed fields. Each suggestion is PREDICTED: nobody has confirmed it, and it is not a probability.

4. Status: EXPERIMENTAL

   Interface Lab

   Which layouts make smell information easier to look through?

   What it does: Tries out cards, grids and side-by-side views built from the same public fields, keeping every label and limit visible.

5. Status: EXPERIMENTAL

   Evaluation Lab

   Does the suggestion rule give the same result every time it runs?

   What it does: Publishes a record of each run of the rule and counts of what it produced. The counts describe the output only. There is no answer key to compare against, so there is no accuracy or performance figure.

6. Status: EXPERIMENTAL

   Review Lab

   How would an editor look at a suggested pair?

   What it does: Shows each PREDICTED pair on a read-only review page with both records, the words they share and the questions an editor would ask. Looking at a page never approves or rejects the pair.

7. Status: EXPERIMENTAL

   Challenge Lab

   Which sets of suggestions are useful for testing the review pages?

   What it does: Sorts suggestions by fixed rules (pairs that only just meet the rule, number of matches, category and kind of match) to help test the review pages. These are test sets, not benchmarks.

8. Status: EXPERIMENTAL

   Review Context Lab

   What checked evidence already exists for the two smells in a suggested pair?

   What it does: Links each review page to the VERIFIED claims and registered sources already recorded for each smell, and marks where there are none. That evidence is about each smell on its own, not proof that the pair is related.

9. Status: EXPERIMENTAL

   Review Lenses Lab

   Can the same suggestions be grouped without ranking them?

   What it does: Groups the review pages by category pair, by whether evidence exists, by number and kind of matches, and by test set. Groups are listed in a fixed order, not by importance.

10. Status: EXPERIMENTAL

    Review Coverage Matrix

    How are the suggestions spread across pairs of categories?

    What it does: Counts the suggestions for each pair of categories and notes where checked evidence exists. A higher count is not better or more important.

11. Status: EXPERIMENTAL

    Change Audit Lab

    What changed in Labs from one release to the next?

    What it does: Keeps a fixed copy of the Labs records at each release (a snapshot) and compares snapshots record by record. Where no copy was kept, it says the comparison is not possible instead of guessing.