What the work involves

You will spend most of your time reading model output and deciding whether it holds up as psychology — not whether it sounds fluent. That means judging clinical reasoning in AI-drafted case formulations, catching ethically unsafe advice (crisis handling, diagnostic overreach, confidentiality assumptions), and flagging where a Japanese response is a literal translation of an American clinical frame rather than something a practitioner in Japan would actually say. The second half of the work is generative: writing case vignettes, assessment items, and reference answers in both languages, with the cultural specifics intact — school refusal (不登校), hikikomori presentations, family-involved treatment decisions, the way distress gets somatized and reported in Japanese clinical settings.

Every judgment is documented. Reviewers are expected to write short, defensible rationales that another psychologist could audit — citing the construct, the guideline, or the cultural norm behind the call, not just "this feels off." Terminology fidelity matters: DSM-5-TR and ICD-11 labels, Japanese diagnostic conventions, and the standard Japanese renderings of instruments and constructs all need to be handled consistently.

What the screen looks for

micro1's screening is AI-led and conversational. It probes the doctorate and where it was earned, actual clinical hours versus purely academic work, and how you acquired each language — self-reported "fluent" collapses quickly under a follow-up asking you to explain a construct like 甘え or to render "differential diagnosis" and "therapeutic alliance" naturally in Japanese. Expect at least one scenario where a plausible-sounding AI answer contains a subtle ethical or cultural error and you are asked what you would do with it. No AI or annotation background is required; unexamined confidence is the thing that fails.

Logistics

  • Contractor engagement, fully remote, asynchronous within weekly batch deadlines.
  • Volume is project-driven — commonly 10–20 hours per week, sometimes bursty around delivery windows.
  • Observed band is $100–200/hr, set by task complexity and credential depth; not guaranteed.
  • Requires a reliable connection, comfort with browser-based review tools, and Japanese input on your machine.