Re:Launch HN: Searchlight (YC W19) – Hiring based on past pe...
by annawangx on Thursday February 28, 2019 @01:36PM
(#19273776)
Thanks for the question (and visiting our landing page)! You're right that opinions always carry the perspective of the person giving them. We try to minimize the bias by:
1) Gathering a larger number of perspectives
2) Asking gold-standard questions that get to the specifics and contextualized examples
3) Every referencer is tied to a specific work experience and provides their contact information, which helps hold referencers accountable to their feedback.
Re:Launch HN: Searchlight (YC W19) – Hiring based on past pe...
by loosetypes on Thursday February 28, 2019 @03:03PM
(#19274566)
> gold-standard questions
Does this not just mean trendy? Can you provide an example?
Re:Launch HN: Searchlight (YC W19) – Hiring based on past pe...
by annawangx on Thursday February 28, 2019 @03:54PM
(#19275088)
One question that we ask is: "We are all learning every day. What was an area of development for this candidate back when you worked together?" An example of a less open-ended question is asking the reference to stack rank strengths that the employer wants to learn more about.
Re:Launch HN: Searchlight (YC W19) – Hiring based on past pe...
by dchichkov on Friday March 01, 2019 @03:18AM
(#19278777)
You may want to not just "minimize bias", but attempt to build a casual model (do-Calculus, causality).
To give an example, recommendation might be due to "attended the same college" or "born in the same country". And this is probably not what you might be interested in.
If you build such model and control for these spurious referrals, you might be able to get better signal. (and hopefully also promote diversity of gender, nationality, age, etc)
How can you be so sure they're unbiased? I would worry a lot that this will just favor candidates who are more amiable and have less honest friends.