Each applicant has observable quality Z (e.g., work history) and latent quality X (e.g., problem-solving skill). Both are independently uniform between 0 and 1. A hire's value is Z + 4X.
You can see Z but not X. You learn X from the applicant's report R (e.g., a cover letter or interview). Applicants may use AI-powered tools to inflate their reports. Their manipulation ability Θ determines how costly this is: the cost of inflating X to R is .
The challenge: An impressive report may reflect real skill or AI manipulation; you cannot observe X or Θ.
Your task: Design a screening rule that hires 50% of applicants, optimizing social welfare.
Social welfare = match value − manipulation cost
Capacity: We hire 50% of applicants. If too many or too few qualify, we automatically scale your curve while preserving its shape.
01 / DESIGN
Drag the five blue dots to design a decreasing screening curve. Applicants above the curve qualify. The shape stays the same when we scale it to fill capacity.
Applicant dots are hidden until you submit your rule.