What the work actually involves
You will spend most of your time labeling and explaining human behavior in recorded material. That means watching or listening to an interaction, marking where tone shifts, where a gesture contradicts the words, where a speech pattern signals hesitation or rehearsal — and then writing the explanation a model can learn from. The written rationale matters as much as the label; a tag with no defensible reasoning behind it is low-value data.
The second half of the work is evaluative. You will read or watch AI-generated responses to interpersonal scenarios and judge whether the model's behavioral read is accurate and appropriate — whether it inferred an emotion the evidence does not support, whether its suggested response would land badly with a real person. Some contributors also build scenario-based training material drawn from their own practice, and most are asked at some point to help tighten the annotation guidelines when edge cases pile up.
What the screen looks for
micro1's process is AI-led and leans on verifiable specifics. Expect to be asked what population you worked with, for how long, under what supervision or credential, and what observational framework you used — then pressed on the details of one case. The platform is not screening for AI experience; it is screening for whether you can describe a subtle behavioral cue precisely enough that a stranger could recognize it, and whether you can separate what you observed from what you inferred. Writing clarity is assessed directly, since annotations are the deliverable.
Logistics
- Contractor engagement, remote, part-time or full-time
- Eligible regions: North America, Latin America, UK, Europe, Oceania
- Largely asynchronous, with periodic calls with project leads on guideline changes
- Requires reliable equipment for audio/video review — headphones, stable connection, a screen you can actually study video on
- $21–70/hr observed; the band reflects differences in credential level, track, and region, and is not guaranteed