Catch up on stories from the past week (and beyond) at the Slashdot story archive

 


Forgot your password?
Close
typodupeerror
×
AI Businesses Technology

Artificial Intelligence is Coming for Hiring, and It Might Not Be That Bad (bloomberg.com) 149

Posted by msmash from the varying-opinions dept.
Even with all of its problems, AI is a step up from the notoriously biased recruiting process, a report argues. From the report: Artificial intelligence promises to make hiring an unbiased utopia. There's certainly plenty of room for improvement. Employee referrals, a process that tends to leave underrepresented groups out, still make up a bulk of companies' hires. Recruiters and hiring managers also bring their own biases to the process, studies have found, often choosing people with the "right-sounding" names and educational background. Across the pipeline, companies lack racial and gender diversity, with the ranks of underrepresented people thinning at the highest levels of the corporate ladder. "Identifying high-potential candidates is very subjective," said Alan Todd, CEO of CorpU, a technology platform for leadership development. "People pick who they like based on unconscious biases."

AI advocates argue the technology can eliminate some of these biases. Instead of relying on people's feelings to make hiring decisions, companies such as Entelo and Stella.ai use machine learning to detect the skills needed for certain jobs. The AI then matches candidates who have those skills with open positions. The companies claim not only to find better candidates, but also to pinpoint those who may have previously gone unrecognized in the traditional process.
This discussion has been archived. No new comments can be posted.

Artificial Intelligence is Coming for Hiring, and It Might Not Be That Bad More

Artificial Intelligence is Coming for Hiring, and It Might Not Be That Bad

Comments Filter:

  • More applicants than jobs (Score:3, Interesting)

    by rsilvergun ( 571051 ) on Thursday August 09, 2018 @04:54PM (#57098680)
    is the problem. The world has more qualified workers than job openings except at the very, very top end of the spectrum (yeah, we can always use more math wizs and surgeons, very few folks have the genetics for that, and yes, a steady hand is genetic).

    You'll still do interviews to pick between them. Hell, my Kid had an in person interview to apply for Nursing School so she could get into her 300 level courses. They had twice as many qualified students as openings...

    • Re: (Score:3, Insightful)

      by snapsnap ( 5451726 )

      Not really now since under Trump there is for the first time more job openings than workers.

      I know for programmers, there have been more openings most places than available workers. We've had more job openings for programmers than employees(!) for around five years despite the fact we pay over 20% more than average. There just aren't enough workers.

      • Re:More applicants than jobs (Score:5, Informative)

        by im_thatoneguy ( 819432 ) on Thursday August 09, 2018 @07:37PM (#57099396)

        Yeah the trend definitely moved with Trump taking office.
        https://tinyurl.com/y72ty3u3 [tinyurl.com] /Sarcasm

      • Re: (Score:2)

        by AvitarX ( 172628 )

        Under Trump, or since 2013?

      • have you actually looked into some of those "jobs". There's the H1-B bait ($50k salary for 8 years high level systems support). There's the "pay to work" jobs ($40k year, 80 hours/week and you use your own car to get to clients). Then there's the "20 hr/week minimum wage job that replaced a $70k/yr factory job".

        As for those programming jobs, good luck getting one without a college degree. In the 80s and 90s I knew lots of guys who programmed for a living with nothing but a high school diploma and took h

      • Re: (Score:2)

        by plopez ( 54068 )

        Average of what? All wages? 20% over what they pay a Mickey D's? Do you also pay benefits? What are the work requirements and the job description requirements? Just saying "20% more" is meaningless.

      • Re: (Score:2)

        by mikael ( 484 )

        What programming languages do you use? There are some areas that programmers won't dare to go because the API's are undocumented and undebuggable, or the employers don't give out references when you leave.

      • Comment removed based on user account deletion

    • The world has more qualified workers than job openings

      You've obviously never tried to find a decent plumber on Craigslist.

    • Re: (Score:2)

      by m00sh ( 2538182 )

      is the problem. The world has more qualified workers than job openings except at the very, very top end of the spectrum (yeah, we can always use more math wizs and surgeons, very few folks have the genetics for that, and yes, a steady hand is genetic). You'll still do interviews to pick between them. Hell, my Kid had an in person interview to apply for Nursing School so she could get into her 300 level courses. They had twice as many qualified students as openings...

      But matching the applicant for the job is still difficult.

      Job descriptions are vague or full of way too much requirements. I have been in jobs where two years into the job, I could only satisfy one line of the job requirement. Other jobs where I ended up doing things that were completely different than what the job description said.

      It's either recruiters who have a worse sense of what is required or my very limited personal connections that I've gotten jobs from. Online job applications has never worked

  • meritocracy? (Score:4, Insightful)

    by sittingnut ( 88521 ) <sittingnut@gmBOHRail.com minus physicist> on Thursday August 09, 2018 @04:56PM (#57098688) Homepage

    the important question is, will the so called "artificial intelligence" (in reality, a data analysis algorithm running on fast computing infrastructure, using fuzzy logic to arrive at faster good enough probabilistic solution, rather than harder best solution, to a problem) look at only data relating to candidates' competency about the job allied to? or will it look at other data too? "diversity" quotas of the employer, personal appearance and tact, social interaction and team work skills, etc? and how exactly?

    • Re: (Score:3)

      by gweihir ( 88907 )

      It cannot. Oh, maybe it can identify low-skill, low-problem candidates, but these you do not want to hire in the first place. Any job that requires any kind of actual skill will be filled in different ways by different people, because you have to bring your personality into it if skill is needed. Since artificial stupidity has absolutely no understanding of anything, it cannot determine whether anybody is a match for any job requiring actual skill. This is just more of the stupidity that you can successfull

      • Maybe it can even better....

        Don't train it with WhizKids job applications to try to pick those, feed it with your existing staff. Might find good matches for your existing team. (ok.. if your existing team already sucks at its job, you should NOT try this at home...)

      • Re: (Score:2)

        by AmiMoJo ( 196126 )

        This is just more of the stupidity that you can successfully hire (and manage) people without seeing them as people.

        TFS says it's the exact opposite. By eliminating biases it considers people's actual skills, rather than lumping them into groups and making assumptions based on things as trivial as their name or gender.

        For example, certain universities carry prestige. A candidate who went to Oxford or Harvard or Tokyo instantly looks better than one who has some low ranked institution on their CV. But that doesn't necessarily mean they don't have the skills or talent required.

        • You missed the point of the GP. They said,

          Any job that requires any kind of actual skill will be filled in different ways by different people, because you have to bring your personality into it if skill is needed.

          As an example of this, I was hired to fill a position (turns out it was two of them...) where the tradition had been to do a lot of manual work, and rely upon institutional knowledge of legacy systems. I was selected for an interview not because I was good at or had these things, but because I had strong skills in parallel areas.

          My experience was in modernizing a vaguely similar system, but not one that you'd immediately recognize as similar enough to the system I

          • Re: (Score:2)

            by gweihir ( 88907 )

            This isn't something that AI will be able to successfully figure out for quite some time yet. Over time it should get better and get there, but I expect that to take a long time. Intuition and creative problem solving are going to be some of the last things that AI is going to be able to tackle.

            At the moment there is no indication these systems will ever be able to tackle these. Don't forget that we do not have AI at all. All we have is dumb statistical classificators called (weak) AI for marketing purposes (i.e. lying to make thinks look massively better than they are).

    • It's long been accurately said that 'It's not what you know, it's who you know.' The system usually sees this as a feature, not a bug. I don't think a tech fix will solve a social problem. Provided they're not the ones getting screwed over, people like their biases.
    • I suspect that almost any system will result in greater diversity, simply due to the existing biases in hiring, which largely amount to who you know. The "AI" could just be a RNG that just picks candidates, and it'd still probably produce better candidates, more diverse candidates, and less cronyism/nepotism.
    • And isn't it funny that the stereotypes and prejudices of the developers of the AI are really enshrined in that code, yet nobody seems to think this is bias.

      • This is the sort of modern "no true scotsman" argument that's starting to popup.
        "why, the logical conclusion of that AI is biased because the man who programmed it is biased!"

    • As opposed to "human intelligence" which is in reality is data analysis algorithms and gut feelings running on slow meatspace hardware, using fuzzy logic to arrive at a good enough probabilistic solution, rather than harder best solution, to a problem.

      Hopefully it makes for a better meritocracy. But no-joke, if you hire a brilliant codemonkey who will lash out at anyone with the audacity of consuming unholy substances like caffeine... They're going to bring down the whole team. And the pro-diversity cr

      • The AIs are going out on the internet, scanning within and LinkedIn, then automatically sending emails to people. It's all before HR gets a chance to look at them.

    • This is the AI version of buzzword bingo, typically played by stupid HR departments, except this time it is outsourced. There is a reason by people go by connections. It is a way to know you are dealing with someone that is trustworthy.

      I expect a thriving business of SEO like businesses popping up. All of a sudden the AI will signal an abundance of qualified laborers from India or the like.

    • Comment removed based on user account deletion
    • Comment removed based on user account deletion
  • It will work fine for commodity jobs; but if someone is out of the ordinary - as most truly great people are - they will get screened by the AI system, which, after all, is really just a kind of filter. It takes true understanding to assess an extraordinary person, and they will get filtered out by the process. See my article on how we are being screened already: https://www.linkedin.com/pulse... [linkedin.com]

    • Re: (Score:2)

      by gweihir ( 88907 )

      I don't think it will even work for "commodity" jobs.

      • The whole thing is hype. How many organisations even hire enough people to produce a decent training set?

        A couple of countries' armed forces, Indian State Railways, and maybe the UK NHS.

        What's that, they could share them? ROFLMAO!

        What's a training set? You aren't using AI. Hype, like I said.

        • Re: (Score:2)

          by gweihir ( 88907 )

          Indeed. Hype driven by the incompetent that hope there will _finally_ be a magic technology that will help them to not suck at their job. Reminds me of the ever ongoing search for a magic programming language that will make language that will make bad coders write code that does not suck. Completely impossible, obviously as that is not where the problem lies.

    • The words "neural network" appear nowhere, which is just as well. Unlike neural networks, most machine learning algorithms can explain themselves to a statistician.

    • Re: (Score:3)

      by rtb61 ( 674572 )

      In this case the 'AI' is not really a filter but a set method of evaluating candidates upon the same basis. It is easy to do, simply come up with what ever you deem to be the important values of the job, and score each candidate against those values. Total up the scores and let it make the choice for you. Keep in mind, the potential employees who are best at the interview process are very likely to be psychopaths, smooth charming, shamelessly lying arse holes. Coming up with the correct metrics to score for

  • You know, 5+ years experience in a technology only 2 years old?

    I bet an AI would work, as long as it's not setup by an HR drone.

  • Garbage in... (Score:5, Insightful)

    by Thelasko ( 1196535 ) on Thursday August 09, 2018 @05:08PM (#57098748) Journal
    ...garbage out.

    If the training data is biased, the AI will learn to be biased. There have been numerous reports on this. [newscientist.com]

    • Re: (Score:2)

      by gweihir ( 88907 )

      Since artificial stupidity has no understanding of anything and can just sort-of replicate labelled training data statistically when used as a classifier in this way, it will have exactly the same problems as the training data, plus a few more. And the training data will be biased and bad, because if we could do this better, we would.

    • Re: (Score:2)

      by AmiMoJo ( 196126 )

      It's much easier to control the training data for an AI than it is for a human. Maybe the hiring manager has been reading Slashdot and picked up a slight bias against people who did a Gender Studies course along side their major subject.

      • It's much easier to control the training data for an AI than it is for a human.

        I agree, it should be easier. However, many AI projects have fallen into this trap. The AI logic is only half of the solution. The other half is high quality training data. Too many projects simply feed historical data to the AI without accounting for the fact the historical decisions were made by humans, and humans have bias. All humans have bias, and most are unaware. This is not insurmountable, but it is difficult. How do you account for something you are unaware of?

        Statistical correction of the

  • Biased is a feature! (Score:2, Insightful)

    by Anonymous Coward

    A hiring system should be biased by definition. Biased to the best candidates. If not, you are doing it wrong.

  • Not a fix for diveristy (Score:2, Insightful)

    by Anonymous Coward

    There is no way that AI in the hiring process will fix the 'diversity problem.' Mainly because the problem largely doesn't exist and is mostly PC-thuggary.

    I never hear about the diversity problem in nursing or preschool school teachers where men are effectively absent from the workforce, or how women want diversity in construction jobs or automotive repair.

    The sexes are different. The races are different. The cultures are different. You will not get a equal mix of them.

    • Besides which, all the "skills" and "requirements" HR has are bullshit. There's a century of research on it, thousands of studies. The only things that matter in predicting job performance are:
      1. IQ
      2. integrity (adds about 25% to IQ predictive validity)

      -both easily, quickly and cheaply testable. Adding other requirements adds very little, and then only for work sample and "structured interviews" (which are nothing like regular interviews, it amounts to administering certain IQ tests in person) and these co

    • Re: (Score:3)

      by AmiMoJo ( 196126 )

      I never hear about the diversity problem in nursing or preschool school teachers where men are effectively absent from the workforce, or how women want diversity in construction jobs or automotive repair.

      Only because you don't listen.

      https://www.bbc.co.uk/news/edu... [bbc.co.uk]

      https://www.ncbi.nlm.nih.gov/p... [nih.gov]

      https://www.womeninconstructio... [womenincon...summit.com]

      http://womeninautomotive.com/ [womeninautomotive.com]

      etc.

    • Reminds me of something in my current company. A friend of mine had her supervisor berated because he dared to "hire 2 white men in a row, don't let it happen again. The next one can not be a white male." Maybe not the exact wording, but the gist of it was basically "Don't hire 2 white men in a row ever again."

  • so I need list each skill 2-3+ different ways and maybe even list stuff like

    software used just to list windows 95 you may need to put down.
    Windows
    Windows 9X
    Windows 95
    Win 9X
    Win 95
    Win95
    Windows 4
    (not even listing all of the OSR updates)

    I one did an online job application with Comcast and they wanted me to fill out this really big skills matrix that was a little like that.

    With some very generic tiles.
    Software listed more then one for the same thing.
    The same basic skill worded 2-3 different ways.
    In house terms.

  • It doesn't really change anything, as the HR level stuff is pretty much mindless and arbitrary.

    And for people thinking this will remove bias... who do you think is going to be giving the program its parameters?

  • What happens when your 5-10+ page resume get's kicked out by an real person? after you needed one to list all of the skilled needed to get past the bot?

    • You send two resumes, or you put the long form skill list in 1 pt white on white in the margins.

      • Comment removed based on user account deletion

        • You still need a 1 pager. The long form one comes later.

          Before you resume gets to the person that might be interested, it has to get past the HR drones, who wouldn't understand the details even if they weren't THAT lazy.

  • the HR individual knows nothing about the skills they are hiring for. Thus looks for buzz words and a well crafted (questionable) resume.
    The initial filtering should be done by an individual with the knowledge and skills they are hiring for. In 10 mins you know what the lay of the land is. But that is only done in small businesses and is not how the corporate or government world works.
    As a self employed contract programmer I have not been asked for a resume in 15+ years. I do not advertise, use the web to

  • Oh good lord (Score:3)

    by cascadingstylesheet ( 140919 ) on Thursday August 09, 2018 @05:46PM (#57098938) Journal
    Truly unbiased hiring would not produce popular distributions of "underrepresented" groups - unless, of course, that is part of the criteria that it is given for success.
  • First of all AIs are biased towards provided dataset. Since most people cannot tell why they had a hunch to hire/not hire the person - how they are going to provide good data set for AI?

  • They'll get what they asked for. (Score:3)

    by ma1wrbu5tr ( 1066262 ) on Thursday August 09, 2018 @07:21PM (#57099328) Journal
    Instead of qualified applicants, they'll end up with people who can produce a good resume. Considering there are countless firms that will write one up for you at reasonable cost, almost anyone with a couple hundred bucks and the most basic knowledge will be able to get their foot in the door and "fake it til you make it". That is, unless you're over 40 and looking for a tech job.

    • Re: (Score:2)

      by mikael ( 484 )

      Those firms just get you to change your employer descriptions from "I did this, that and a bit of twiddly stuff" to using buzzwords "Achieved", "Led", "Successfully", "Developed new", and all that high-achiever keyword things.

  • AI can probably help with the amount of candidatures, which is often too big for human recruiters to properly manage it. When you have hundreds of candidates for a positions, a computer filtering irrelevant candidates would certainly help.
  • As a programmer, I'm pretty sure I know how to sweet talk an AI. Everyone will be wondering how I got the job and I'll be taking the AI out for an evening of formatting large data files and killing all humans!
  • Young Johnny just graduated from "Bob's School of Typewriter Repair and Information Technology" and your telling me I can't filter him out? While your degree doesn't matter as much after a certain point, it still matters. You want people to have a solid foundation and a mastery of the fundamentals. I don't really care what school, but I do care if it was in the top 10, 20, 50, or 100 for relevant degrees. Of course, if the resume indicates that they have the right skills and the overcame obstacles (no d

  • Its almost impossible to come up with well defined performance evaluation criteria for many jobs. Lines of code per day? Bugs found??? Papers published? How can you even produce a a training data set where former employees performance is rated.

    If humans do that evaluation, then whatever bias the humans had will just be trained into the algorithm.

    If you are hiring factory workers, you might be able to measure productivity or error rate or something, but that seems fraught with running afoul of age discrim

  • Didn't Northpoint promise that for the parole/correction software?

    https://www.propublica.org/art... [propublica.org]

  • "I think we oughta take the men out of the loop."

  • A bias-free human being is like a coffee table, where, when you spill water, the water stays exactly where it first lands: it doesn't preferentially dribble down one side, or pool in an (almost) invisible declivity, or find itself attracted by surface tension to a sticky area.

    Do you own such a coffee table? I don't. But I consider mine flat enough. My mugs don't rock, and I don't even need a soft coaster to achieve this. But my soup does ride a little higher at one end of the bowl, so perhaps what I need i

  • I mean, some idiot HR deps - sorry, I repeat myself - use DATABASE SEARCHES to find "qualified" candidates. NONE of them have any idea of what the requirements are, or what translates.

    And no, this isn't new: the last time I was looking, in '09, Grumman was doing just that. You may be wonderful, but if you don't have the right acronyms in the right order, they're not going to even look at you.

  • Having just escaped from job-hunting hell, I can say that the keyword-matching tricks you have to jump through are a real pain in the ass. There were several jobs that I knew I could do where my keyword-match score probably excluded me. The flip side is that many role descriptions are written by hiring managers (or the hiring manager from years back) and they are frequently not relevant to the role. I am a hiring manager and I write my own job descriptions fresh for each role that I'm hiring for and I valid

Slashdot Top Deals

Real Programmers don't eat quiche. They eat Twinkies and Szechwan food.

Close