Receive daily articles & headlines each day in your inbox with your free ERE Daily Subscription.

Not logged in. [log in or register]

Computers Aren’t Ready to Take Over Hiring

by Oct 11, 2012, 5:28 am ET

Scientist at Xerox

If you are into the use of technology to support the hiring process, read the recent Wall Street Journal article about algorithmic hiring.

It offers a very real glimpse into the future of hiring. To those companies who are looking for ways to increase the efficiency and effectiveness of the hiring process, the value returned by the newest wave of advanced technology can be significant.

But it would be wrong to blindly accept that computers are poised to take over the hiring process from human hands.

As a traditionalist who also embraces change and loves technology, I straddle two sides of this issue. I believe in the value of algorithms and data to help optimize and automate decision-making. However, the role of humans in the hiring process cannot and should not be replaced.

The last book I read, The Physics of the Future by Michio Kaku, provided me with some really good perspective on this issue. This is a fantastic book in which the author, a physicist, uses factual scientific information to predict what we can expect in the near future.

The author discusses the future of the workforce and suggests that by midcentury (2030-2070) almost all lower-level jobs will be automated. He goes on to suggest that the types of jobs that will not be automated will be those that require “the one commodity that robots cannot deliver: common sense.”

The inability of machines to think creatively and to have intuition creates a limitation to their use and value. So, while the wealth of information available to us will be staggering, it will still take a human brain to digest it, evaluate it, and make decisions that cannot be programmed or made using algorithms.

I could not agree more with Kaku and when it comes to machines and hiring, we need to keep a sense of realism about what we can expect machines to do. More than anything we need to see them as a helpful tool to make experts better, not as a substitute for human intelligence.

Know this about hiring by algorithm:

  • Algorithms must be fed quality post-hire performance data to be useful. Without a stream of quality outcome data, the brainiest computer in the world will be blind to the real picture of what predicts job performance. Companies are notorious for failing to care enough to capture and use performance data. This has been a bane for us I/O psychologists for decades. The value of algorithms will be limited by the current mindset held by many corporations.
  • Automated hiring has a power alley. High-volume hiring has been looking to automation for a decade or more now. High volume is the best place for automation in hiring because humans cant possibly evaluate all applicants, and have issues with consistency and objectivity. Automated hiring shines because it provides a way to quickly sort applicants and differentiate them well enough to remove the deadweight. As one deals with more complex jobs at lower volume levels, it will get harder to use algorithms and human judgment will become more important. Likely we won’t see deep penetration of algorithms beyond entry-level or mid-level jobs anytime soon.
  • It will always be about informed decision-making. There is no substitute for expert judgment by humans with training, experience, and motivation. This is what we do best. The goal is to support humans with better quality information, not to try and remove us from the entire process.
  • A process-based approach should be the goal. Any hiring strategy is about using the right tool for the job at the right time and collecting information that is timely, accurate, and useful. When using automated hiring, one must know where it best fits with the overall process and strategy.
  • Boundaries on what constitutes usable data will soon need to be established. While the type of data that is traditionally found on application blanks is fair game, why stop there? Data such as social network use or anything that is scrapable and trackable can add to computer models and identify patterns that may predict valued outcomes. But at what cost? Privacy is the new battlefront on the web and the data used to determine suitability for hire going to be an issue. What if deep personal data patterns of data show discrimination (a very real possibility)? Will the contributing data be ruled illegal? Even if it can be shown to be job related?  Our current government hiring regulations were created in 1978 (when Pong was high tech).  Something will have to give in the near future.
  • Our concept of validation will need to be expanded. Validation means verifying that a test (or any data used for prediction) accurately measures the job performance domain. Traditional validation is static, using a data-based snapshot in time. But these new algorithms are dynamic and ever-changing, requiring validation to be fluid. While the core concept of validation will never change, its execution will have to.

Change is not always easy, but it is always inevitable. With progress comes the need for change and the need to reshape our conceptions. We as a society are going to face many interesting battles that will test our boundaries in the near future. Hiring is but one of these battles. As with all of them the key will be to know how best to use technology while not losing sight of the value that the human mind brings to the table.

This article is provided for informational purposes only and is not intended to offer specific legal advice. You should consult your legal counsel regarding any threatened or pending litigation.

  • http://www.smarterhire.com.au John Millican

    I completely agree with you. There is an art and science to identifying a high quality hire. The use of technology in-interview and as part of the process is where I believe hiring will evolve.

    “iPad interview guides” are available already and bring structure and science to a very important process. There are too many variables to completely foresee replacing an experienced recruiter. Giving recruiters and hiring managers these tools affords the perfect mixture of technology and the human brain which optimises the art and the science.

  • Martin Snyder

    Another excellent, thought provoking item Charles. It’s so important to include a reference to scale in these discussions, as you did.

    Your closing line about the value of the human mind in the future reflects what I have observed in the last 20 years: the value of medium intensity brain work has fallen off a cliff in the USA and elsewhere, and in many ways, that is the work getting easier to automate.

    We are far, far from automated plumbers and roofers and nurses and painters and police officers.

    That means a major shift in who gets paid what in our society. As of now, managers and professionals are 1/3 of the workforce. If that proportion goes to 1/6 or 1/10 because of technology (like what happened with farming in the 19th / 20th century), its going to mean big changes to our economy.

    Its not really the skilled working people who have to worry- my plumber is a PhD in Mech E but makes more then twice the money not working as an engineer, and my carpet cleaning guy is near the same story (BS Mech E).

    That will change the culture in big, unknown ways. It speaks further to a higher ed bubble for one thing…

    As to scale and machine decision making in human affairs: we are yet well away from machines being able to game the emergent complexities of humanity when it comes to symbolism, motivation, politics, sex, power, humor, and style.

    All of those can be mapped and manipulated and I think it will happen faster than people think and in creepy ways- including for hiring.

    What it really means is more serious problems defining objective reality going forward. When machines can simulate people and scenes at full HD quality, and do so with a directed agenda and the power to influence people, just what is and what is not a fact is going to blur even more- and I find the current pace alarming to say the least….

    Is the ultimate luxury good your very own reality? We can count on the free market to deliver….

  • Richard Araujo

    “Algorithms must be fed quality post-hire performance data to be useful.”

    That is, I think, the key issue that will lead to most failures in this approach to hiring. I think these methods will force more reliable performance measures into use over time. However, I think it will be a rocky adoption process because in my experience bad management starts at the top, as does company culture, and company owners, executives, and financiers are loath to admit their lack of skill in any area, management included. More likely they will avoid the innovations that give them information they don’t want to receive while the more flexible, forward thinking will adopt it and modify their techniques to complement the process.

    I also wonder if the era of small business is going to come to an end, at least in the traditional sense. With the workforce trending towards temps and contractors, a future of near total process outsourcing might be possible. That would allow these kind of techniques to be used more efficiently as businesses increasingly specialized in functions as a product. Interesting times ahead in any event.

  • http://www.shakercg.com Joseph Murphy

    Charles – thanks for continuing to invite us forward.

    The decision to hire will most likely always be an act of personal judgment.  However, better data regarding the variables that impact the quality of the decision is what differentiates down-stream outcomes.  And in the case of hiring, that means on-the-job performance.

    Your first and last bullets are the ‘sit up and take notice’ elements to embrace.

    All the writing on Big Data is capturing the imagination of business and working its way into the business process called staffing.  To extract value from Big Data requires rigor and discipline.  This is the work of HR Analytics.

    Read more here

  • Keith Halperin

    Thanks, Dr. Handler.”It will always be about informed decision-making. There is no substitute for expert judgment by humans with training, experience, and motivation. This is what we do best. The goal is to support humans with better quality information, not to try and remove us from the entire process.”

    We may do it best, but increasingly research (http://en.wikipedia.org/wiki/Thinking,_Fast_and_Slow) indicates that people aren’t “rational actors” and are heavily and inherently prone to all types of cognitive biases that can negatively influence our decision-making. As somebody said, “Don’t waste your intuition on what an algorithm can do better.” IMHO, the key is finding out what is best left to algorithms and what is best decided by trained, experienced, and motivated intuition. (Maybe when I finish “Think Fast, Think Slow”, I’ll find out what those are….)

    Cheers,

    Keith

  • http://www.rolepoint.com Kes Thygesen

    High-volume hiring really is the one place that truly needs an automated system of some sort in the hiring process. But you’re right, there is NO replacement for common sense — at least not yet. (A little scary!) But if you have a high volume of open jobs and limited HR staff, automation via software, encouraging employee referrals and internal candidates, and common sense are key to successful hires.

  • Glyn Brokensha

    Yes, absolutely agree … with one small caveat … “Yet”. If people aren’t concerned about the pace of this change, then they should be.

    As remotely collected biometric data becomes ubiquitous and Moore’s “Law” grinds inexorably on, I’d predict the broader trend will mimic the “computers will never play chess” debate of the late sixties.

    I remember personally the certainty with which the pundits made the declaration, “That will never happen”. But it did. Grandmaster Gary Kasparov lost to IBM’s “Deep Blue” in the famous Game 1 in 1996.

    I’ve developed this notion in “The Final Position” at http://expr3ssion.com/2012/the-final-position if your readers would care to drop by.

  • http://www.rocket-hire.com charles handler

    The Singularity is near!!

  • Glyn Brokensha

    Yes, Charles, again I agree… it may indeed.

    Regardless, though I believe it’s not necessary to posit the Singularity for machine-assisted decision making in recruitment and the broader field of human relations to be a reality.

    Your article very properly highlights that it already is, albeit in a limited way. Currently this is largely very limited by its reliance on language (resume, screening question scores, psychometrics).

    My thesis is that there will be a qualitative shift in the types of data used for that decision making. It’s already possible to integrate non-verbal data like voice tones, blink rate and skin blood flow into the mix. And to do that in real time without the physical presence of the subject.

    I wonder, for example, how long it might be before a PET (Positron Emission Tomography) scanner might be scanning away during a remote interview? And how might we (or the post-singularity machines!) use that data? I hope that it will be wisely.

  • http://www.neorecruitermom.com Shannon Wagner

    I have a question: When you are dealing with high volume recruiting, at what point does efficiency and business need become more important than the candidate experience?

    In other words, if you automate as much of the process as possible for efficiency purposes, what is the risk of the candidate having a negative experience and/or not feeling valued?

    I am naturally a very “personal-touch”, customer service oriented person, so my instinct is to speak and interact with candidates personally as much as possible. However, when you are hiring in the thousands annually, realistically you have to leverage technology to streamline processes as much as possible.

    I truly am seeking thoughts and opinions on this–I’m not trying to make any sort of point or argument for one way vs. another!

    Thanks,
    Shannon

  • Glyn Brokensha

    @ShannonWagner

    Good question. It’s our experience (over > 200,000 applicants) that applicants are pretty sanguine about the realities of recruitment, especially if the role is likely to attract lots of people. So they are philosophical about automation, although they do appreciate reassurance that no applicant will be summarily rejected without human consideration.

    We have found that it’s very important to communicate with applicants every three days as recruitment progresses (email is sufficient).

    And it’s even more important to communicate with unsuccessful applicants and to do so at the earliest opportunity. Again a carefully worded email is acceptable.

    Applicants hate a vacuum much more than any amount of automation. It’s only the words, “Only Successful Applicants will be contacted” that they hate more!

    And our (Australian) research shows that more than 80% of applicants never receive any communication, not even an acknowledgement email.

    Hope this is helpful

  • Pingback: Workforce Development News – October 15, 2012 | Workforce Solutions Group | St. Louis Community College