The number of referred candidates for any given role will be significantly higher than the number of candidates that are fit for this role. To help clients focus on the most relevant candidates, Job Protocol enables the vetting of candidates.
Candidate vetting is applying a ranking to all referred candidates. The higher the rank, the more likely a candidate is a fit for a given role.
Vetting relies on a lot of explicit and implicit information of the candidate, the referrer, the role and the client. Things to consider here are
- The background, skillset and trustworthiness of the credentials of the candidate;
- The referrer reputation;
- The requirements, responsibilities and compensation of the role;
- The culture, existing team and vision of the company;
Job protocol takes two approaches to vetting candidates: by relying on the referrer reputation and by introducing the ‘reviewer’ actor.
The referrer reputation is a measure of how well past referred candidates have done in the interviewing process: how many got an interview, how many got hired. A higher referrer reputation means a higher percentage of referred candidates have received interviews or were hired.
There are multiple ways to calculate the referrer reputation, and different frontends will use different calculations based on their specific needs.
An easy way to rank candidates is to give priority to candidates brought in by high-reputation referrers. Other ways to prioritize candidates is by giving preference to candidates that have more information available.
Any kind of candidate ranking will take referrer reputation into account, which creates a strong incentive for referrers to be cautious about which candidates they refer, increasing the overall signal to noise ratio of the system. Furthermore, the bounty a referrer can earn is dependent on the referrer reputation, further increasing its importance (more on this later).
To further increase the signal over noise ratio, we introduce the ‘reviewer’ actor to the network. Reviewers who correctly help predict successful candidates will be rewarded for that.
Anyone can become a reviewer by going through KYC (e.g. proof of humanity) and by putting down a ‘good behavior’ stake. Reviewers have access to anonymized profiles (e.g. without candidate contact details), and for each profile they access they are required to give a score between 1 and 10 indicating how well the candidate fits the role requirements (higher is better).
Reviewers are rewarded a part of the bounty for their work. The amount they earn depends on how well they reviewed for a role, e.g. the more candidates they score high and get interviewed, the more they earn. Every candidate can be reviewed by multiple parties, creating the opportunity to come to a ‘global consensus’ of what the best fitting candidates are for a specific role.
The challenge with this approach is that candidate information is confidential, and potentially valuable (e.g. a vetter can ‘steal’ candidate information to try to place the same candidate in another role). To prevent fraud, vetters have to put up a significant ‘good behavior’ stake that they risk to lose in case bad behavior is exposed (evaluating proof here is another challenge and probably requires a distributed court system like Kleros).
The beauty of this approach is that vetting becomes an open ecosystem where anyone - humans or AIs - can help increase the signal > noise of the network, making it better for everyone involved. Evaluating candidates is hard and multi-dimensional, creating opportunities for a lot of innovation (e.g. niche reviewers focussed on a specific geographic area, or reviewers incentivizing candidates to take certain standardized tests). Furthermore, this allows the vetting information to be used in determining the referrer reputation, significantly increasing the granularity of the reputation.