What the work involves

You will be handed model outputs that touch adolescent and young-adult substance use and asked to judge them the way you would judge a trainee's note. Typical tasks: rating two candidate answers to a parent's question about vaping or cannabis; flagging where a model recommends adult-dose buprenorphine induction for a 16-year-old; rewriting an unsafe or vague answer into one you would stand behind; and writing short rationales that explain why an answer fails, in terms a non-clinician annotator can apply to the next case. Some batches ask you to author original prompts — realistic clinical vignettes with a hidden trap, such as a teen disclosing use during a sports physical, or a family pressing for information the minor has not consented to release.

What the platform screens for

micro1 runs an AI-led interview before any paid work. It probes whether your knowledge is load-bearing: expect follow-ups that push on CRAFFT versus S2BI, on how adolescent withdrawal presentation differs from adult, on when medication for opioid use disorder is appropriate under 18, and on the tension between state minor-consent statutes and 42 CFR Part 2. Vague, guideline-flavored answers get pressed until they either resolve into specifics or collapse. The screen also tests evaluation judgment — whether you can separate "an answer I would phrase differently" from "an answer that would harm a patient," and whether you can articulate that line consistently across cases.

Logistics

  • Fully remote, asynchronous; batches are claimed from a queue rather than scheduled.
  • Most contributors work 10–20 hours a week; some weeks have no batches in your specialty, which is normal and not a performance signal.
  • Pay is hourly within the observed $45–95 band, set by credential, task type, and calibration performance. Rates are observed on the platform, not guaranteed.
  • Expect a paid or unpaid calibration set before full access, and periodic re-calibration against gold-standard items.
  • You will need a current clinical credential and, for some batches, willingness to sign a confidentiality agreement covering model outputs.