The critical parameters here are the probability of a malicious/non-malicious actor (or competent/non-competent, etc) accepting the job (let's call it Pma, Pna)

The critical parameters here are the probability of a malicious/non-malicious actor (or competent/non-competent, etc) accepting the job (let's call it Pma, Pna), and the probability of selecting a malicious/non-malicious actor (let's call em Pm, Pn). What you want is that the probability of selecting times the probability of acceptance be greater for non-malicious: Pn Pan >> Pm Pam If we assume we can't do better than random screening for malicious, we have a fixed Pn, Pm. So the crucial thing is to find incentives so that non-malicious actors are the least likely to take the job. A good pre-screening (if possible) would have to raise the RHS or lower the RHS. For example, if the major incentive for malice is monetary, sampling from wealthy citizens could lower Pm, but it might also sightly lower Pan, so you may have to spend more resources to guarantee Pan=Pam=1.