tl;dr
tl;dr
Why do we think having machines doing our hiring is better than having humans do it?
Guys, hey, guys. Math is not inherently objective, machines are not inherently objective, algorithms are totally the opposite of objective. You have to decide what biases you want built in and select for those. Either you are biased for diversity of gender, race, education, etc., or you are biased against. There is no fairy la-la land where gender, race, unequal access to education and so forth suddenly cease to exist. Stop lying to yourselves and everyone else.
Meanwhile, the Reuters article contains these choice bits:
Some 55 percent of U.S. human resources managers said artificial intelligence, or AI, would be a regular part of their work within the next five years, ... Employers have long dreamed of harnessing technology to widen the hiring net and reduce reliance on subjective opinions of human recruiters.
I don't quite know how to put this, but if you are hiring humans who might have to work with other humans, there is no AI that is ever going to be that good at figuring out how to do this. Either you end up with everyone being basically the same, so no one to think differently or challenge the culture, ever, or you end up with so much randomness and disparity that it is inefficient if not impossible to get people to work productively together. Humans are the best gauge of working with other humans.
I cannot imagine any machine method of sorting through resumes that humans do not already do, but they could be faster. Things like what school did they go to (inherently discriminatory but widely used), what test score did they get on some standardized test (still widely used even though it has been shown that there is no validation for most of these applications), do they have relevant job experience.
Ah-ha! But even that could be tricky to program. What if the relevancy of someone's experience is not having a degree in the specific sub-field, or a prior job doing exactly one thing? There are the must haves, and then there is the question of do we want to only have one skill set we ever hire, or do we want people with different sets of additional skills? How does a machine determine whether what is on a resume is even legitimate or applicable?
https://www.reuters.com/article/us-amazon-com-jobs-automation-insight/amazon-scraps-secret-ai-recruiting-tool-that-showed-bias-against-women-idUSKCN1MK08G
https://www.theverge.com/2018/10/10/17958784/ai-recruiting-tool-bias-amazon-report
https://www.theverge.com/2018/10/10/17958784/ai-recruiting-tool-bias-amazon-report
Why do we think having machines doing our hiring is better than having humans do it?
Guys, hey, guys. Math is not inherently objective, machines are not inherently objective, algorithms are totally the opposite of objective. You have to decide what biases you want built in and select for those. Either you are biased for diversity of gender, race, education, etc., or you are biased against. There is no fairy la-la land where gender, race, unequal access to education and so forth suddenly cease to exist. Stop lying to yourselves and everyone else.
Meanwhile, the Reuters article contains these choice bits:
Some 55 percent of U.S. human resources managers said artificial intelligence, or AI, would be a regular part of their work within the next five years, ... Employers have long dreamed of harnessing technology to widen the hiring net and reduce reliance on subjective opinions of human recruiters.
I don't quite know how to put this, but if you are hiring humans who might have to work with other humans, there is no AI that is ever going to be that good at figuring out how to do this. Either you end up with everyone being basically the same, so no one to think differently or challenge the culture, ever, or you end up with so much randomness and disparity that it is inefficient if not impossible to get people to work productively together. Humans are the best gauge of working with other humans.
I cannot imagine any machine method of sorting through resumes that humans do not already do, but they could be faster. Things like what school did they go to (inherently discriminatory but widely used), what test score did they get on some standardized test (still widely used even though it has been shown that there is no validation for most of these applications), do they have relevant job experience.
Ah-ha! But even that could be tricky to program. What if the relevancy of someone's experience is not having a degree in the specific sub-field, or a prior job doing exactly one thing? There are the must haves, and then there is the question of do we want to only have one skill set we ever hire, or do we want people with different sets of additional skills? How does a machine determine whether what is on a resume is even legitimate or applicable?
https://www.reuters.com/article/us-amazon-com-jobs-automation-insight/amazon-scraps-secret-ai-recruiting-tool-that-showed-bias-against-women-idUSKCN1MK08G
https://www.theverge.com/2018/10/10/17958784/ai-recruiting-tool-bias-amazon-report
https://www.theverge.com/2018/10/10/17958784/ai-recruiting-tool-bias-amazon-report
The thing that attracts people to AI for hiring is that no existing hiring practices actually work well.
ReplyDeleteBut it seems like we're just baking the same things that seem not to work when people do them into what the machines/AI do. That does not seem like an improvement.
ReplyDeleteIt's definitely not an improvement.
ReplyDeleteI just think it's worth putting it in context: hiring is bad, and people are trying to make it better. But their ideas on how to make it better are also bad.