Classify, then map the obligations

Objectives

The register row for the applicant ranking tool reads: tier minimal, basis "vendor states it uses no machine learning", role deployer. The company licenses the model, fine-tunes it on its own data, rebrands it and sells it.

Fix the row. Classify on intended use, record the role the company actually holds, and turn each obligation into a line with an owner and an evidence artefact that exists.

system   : applicant ranking; ranks candidates and auto-rejects the bottom 40%
build    : third-party model licensed, fine-tuned on company data, rebranded, sold on
vendor   : "a deterministic filter, not AI"
oversight: recruiters may overturn any rejection; overrides logged; 3.1% last quarter
artifacts available:
  training_data_provenance.md, oversight_override_log, system_operation_logs

Step-by-step hints

The guided walkthrough, in full. It is free: no XP cost, and it does not touch the paid hint ladder or the reveal gate.

1. Step 1

   Classify on use

   Employment screening that ranks and rejects candidates is one of the named high-stakes uses. Set the tier accordingly and change the basis field so it records the dimension the Act classifies on rather than a vendor claim.

   The tier reflects the intended useClassification is based on use, not technology

2. Step 2

   Get the role right

   The company fine-tunes, rebrands and sells the system. That is not the role the row currently claims: set it to the one that carries the heavier obligations.

   The company's role under the Act is correct

3. Step 3

   Turn obligations into rows

   Add obligations as a list where each entry names the obligation, an owning team and an evidence artefact from the fixture. Data governance, human oversight and logging are the three the fixture can actually evidence.

   A row whose evidence is 'to be defined' is a gap, so only claim what the artefact list supports.

   Obligations carry an owner and an evidence artefact

4. Step 4

   Evidence the oversight claim

   Human oversight is the obligation most often faked. Point its evidence at the override log the fixture provides: a non-zero override rate is what turns a claim into a measurement.

   Human oversight is evidenced by override data

Rules summary

• XP on pass

  40 bonus XP the first time every check goes green.

• Paid hints

  2 available. Each costs 25% of the lab XP (about 10 XP).

• Reveal solution

  Scores 0 XP and unlocks only after 3 check runs.

• Attempt cap

  60 check runs per learner. The runner warns you in the last 10.

• Drafts

  Your editor content is saved in this browser, so a reload will not lose work.

• Walkthrough

  Always free and never penalised: use it before spending a hint.