3 ms·

The way I see it is you're trying to solve this problem in the wrong place, just like you said, the real issue lies elsewhere and this is precisely where we sho

by Aqua 8y ago

The way I see it is you're trying to solve this problem in the wrong place, just like you said, the real issue lies elsewhere and this is precisely where we should direct our efforts to fix it. If fewer women decide to pursue a technical career, then perhaps we should learn why and counteract. By implementing an artificial counter-bias you're merely hiding the real problem by obscuring the data, while the symptoms and root causes remain intact.

tomrod 8y ago
A counter-bias on the conditional step could indeed assure a correction. Let's formalize it a bit, I think this helps clarify. Let A be the hiring propensity, B be job preparation (education, experience, what have you), and C be the "blinding factors" Helsinki is trying to address. As a model we would like the joint propensity for everyone to be normalized. In other words, P(A,B,C) would be about the same. The joint propensity P(A,B,C) would be expressed as P(A,B,C) = P(A | B,C) * P(B | C) * P(C) [0] Helsinki is making the first term, P(A | B,C) independent of their "blinding factors", or, in other words, P(A | B,C) = P(A | B).[1] The remaining terms P(B | C) and P(C) are unaffected. It is P(B|C) differential that many arguing for what I call non-autarky "social corrections" (we do something as a society). Since Helsinki's decision affects _only_ the conditional probability P(A| B,C), the other factors are still at play, and P(A,B,C) remains at issue. There is nothing that can be done by Helsinki's hiring department for P(B|C) in the short term -- people are dealt the hand they have, life isn't fair, and so on. Completely true. By blinding themselves, they set P(A|B) to be flat on C and no "equalization" of P(A,B,C) can be observed. A truly egalitarian society would have P(B|C) = P(B). However, we aren't there. C, what Helsinki blinds themselves to, matters for qualifying life experiences. The model that is being argued for by you and others I believe is formalized as a conditional independence model, namely that the propensity of hiring is independent of race, conditional on life preparation: P(A,C | B) = P(A|B) * P(C|B).[2] Note though: bias still persists: P(C|B) = P(B|C)*P(C)/P(B) -- we can't get away from that P(B|C) term in either model, and we know qualitatively that P(B|C) != P(B), hence why a correction can be positively argued for (rather than solely from emotion or duty). Anyhow, that's the model in mind, and I'm open to being wrong/corrected. Thoughts? -------------------- After thinking about it more, it seems that the ultimate goal of Helsinki is to make P(A|B,C) = P(A|B), not the latter model. -------------------- [0] Chained conditional probabilites: https://en.wikipedia.org/wiki/Chain_rule_(probability) https://en.wikipedia.org/wiki/Chain_rule_(probability) [1] Independent variables definition: https://en.wikipedia.org/wiki/Independence_(probability_theory) https://en.wikipedia.org/wiki/Independence_(probability_theo... [2] Conditional independence: https://en.wikipedia.org/wiki/Conditional_independence https://en.wikipedia.org/wiki/Conditional_independence
- Aqua 8y ago
  Thank you for formalising this, I get your point and I see what you're signalling, but I disagree with the solution you're proposing. Let me explain why. Firstly, I never said Helsinki's approach will completely solve the problem of inequality and you're right about that, nonetheless I perceive it as an improvement over the standard recruitment process. Your math makes sense, although I think the initial conditions are incorrect, because if I understand correctly your assumption is that the attributes hidden by the blinding factors would have had a positive impact on the "land a job interview or not" decision. On the other hand, I believe the opposite is far more likely, and this is precisely where we disagree. Also, I completely condemn the notion of attempting to fix the problem with affirmative action, primarily because it attempts to solve inequality(evil) with deliberate and explicit discrimination(greater evil). Don't forget that discrimination and prejudice is where it all started. Edit: I should have been more precise when referring to blind hiring. I meant completely blind hiring, including the "on-site" interview stage(making it "blind" could be difficult). Obviously you're also correct about the earlier life experiences which could have been impacted by numerous factors, but my perception on that is the same as with law, it shouldn't be retroactive because you can't change people's past. Besides, each case is different and applying counter-biases here is like blind shooting.
  - tomrod 8y ago
    >On the other hand, I believe the opposite is far more likely, and this is precisely where we disagree. If I understand where you are coming from, "landing the job interview" are the direct factors Helsinki is choosing to be blind to, like name and so forth. Am I understanding that right? If so, let me clarify the point -- the factors that Helsinki _will_ judge resumes on -- experience and so forth -- are pre-conditioned on your name (etc.). So the fact that one makes the initial cut is still driven by the factors Helsinki chose to be blind to. > Also, I completely condemn the notion of attempting to fix the problem with affirmative action, primarily because it attempts to solve inequality(evil) with deliberate and explicit discrimination(greater evil). I get where you're coming from on this. The method I propose, if Helsinki uses the blinding methodology, is to impose any "corrective"/affirmative actions after the the blind evaluation of skills. That way every person is ensured to be qualified (avoiding the critique of "diversity hires"). One possible (and usually a default) "corrective" approach is autarky/doing nothing. And... that's my whole point! I appreciate the discussion.

by Aqua 8y ago

tomrod 8y ago

A counter-bias on the conditional step could indeed assure a correction. Let's formalize it a bit, I think this helps clarify. Let A be the hiring propensity, B be job preparation (education, experience, what have you), and C be the "blinding factors" Helsinki is trying to address. As a model we would like the joint propensity for everyone to be normalized. In other words, P(A,B,C) would be about the same. The joint propensity P(A,B,C) would be expressed as P(A,B,C) = P(A | B,C) * P(B | C) * P(C) [0] Helsinki is making the first term, P(A | B,C) independent of their "blinding factors", or, in other words, P(A | B,C) = P(A | B).[1] The remaining terms P(B | C) and P(C) are unaffected. It is P(B|C) differential that many arguing for what I call non-autarky "social corrections" (we do something as a society). Since Helsinki's decision affects _only_ the conditional probability P(A| B,C), the other factors are still at play, and P(A,B,C) remains at issue. There is nothing that can be done by Helsinki's hiring department for P(B|C) in the short term -- people are dealt the hand they have, life isn't fair, and so on. Completely true. By blinding themselves, they set P(A|B) to be flat on C and no "equalization" of P(A,B,C) can be observed. A truly egalitarian society would have P(B|C) = P(B). However, we aren't there. C, what Helsinki blinds themselves to, matters for qualifying life experiences. The model that is being argued for by you and others I believe is formalized as a conditional independence model, namely that the propensity of hiring is independent of race, conditional on life preparation: P(A,C | B) = P(A|B) * P(C|B).[2] Note though: bias still persists: P(C|B) = P(B|C)*P(C)/P(B) -- we can't get away from that P(B|C) term in either model, and we know qualitatively that P(B|C) != P(B), hence why a correction can be positively argued for (rather than solely from emotion or duty). Anyhow, that's the model in mind, and I'm open to being wrong/corrected. Thoughts? -------------------- After thinking about it more, it seems that the ultimate goal of Helsinki is to make P(A|B,C) = P(A|B), not the latter model. -------------------- [0] Chained conditional probabilites: https://en.wikipedia.org/wiki/Chain_rule_(probability) https://en.wikipedia.org/wiki/Chain_rule_(probability) [1] Independent variables definition: https://en.wikipedia.org/wiki/Independence_(probability_theory) https://en.wikipedia.org/wiki/Independence_(probability_theo... [2] Conditional independence: https://en.wikipedia.org/wiki/Conditional_independence https://en.wikipedia.org/wiki/Conditional_independence

Aqua 8y ago
Thank you for formalising this, I get your point and I see what you're signalling, but I disagree with the solution you're proposing. Let me explain why. Firstly, I never said Helsinki's approach will completely solve the problem of inequality and you're right about that, nonetheless I perceive it as an improvement over the standard recruitment process. Your math makes sense, although I think the initial conditions are incorrect, because if I understand correctly your assumption is that the attributes hidden by the blinding factors would have had a positive impact on the "land a job interview or not" decision. On the other hand, I believe the opposite is far more likely, and this is precisely where we disagree. Also, I completely condemn the notion of attempting to fix the problem with affirmative action, primarily because it attempts to solve inequality(evil) with deliberate and explicit discrimination(greater evil). Don't forget that discrimination and prejudice is where it all started. Edit: I should have been more precise when referring to blind hiring. I meant completely blind hiring, including the "on-site" interview stage(making it "blind" could be difficult). Obviously you're also correct about the earlier life experiences which could have been impacted by numerous factors, but my perception on that is the same as with law, it shouldn't be retroactive because you can't change people's past. Besides, each case is different and applying counter-biases here is like blind shooting.
- tomrod 8y ago
  >On the other hand, I believe the opposite is far more likely, and this is precisely where we disagree. If I understand where you are coming from, "landing the job interview" are the direct factors Helsinki is choosing to be blind to, like name and so forth. Am I understanding that right? If so, let me clarify the point -- the factors that Helsinki _will_ judge resumes on -- experience and so forth -- are pre-conditioned on your name (etc.). So the fact that one makes the initial cut is still driven by the factors Helsinki chose to be blind to. > Also, I completely condemn the notion of attempting to fix the problem with affirmative action, primarily because it attempts to solve inequality(evil) with deliberate and explicit discrimination(greater evil). I get where you're coming from on this. The method I propose, if Helsinki uses the blinding methodology, is to impose any "corrective"/affirmative actions after the the blind evaluation of skills. That way every person is ensured to be qualified (avoiding the critique of "diversity hires"). One possible (and usually a default) "corrective" approach is autarky/doing nothing. And... that's my whole point! I appreciate the discussion.