Why Artificial Intelligence for Recruiting and HR is Really Stupid.

fourfour-small-wonder-so-bad-its-triflingI’m sure you’ve heard by now, that artificial intelligence is coming to Human Resources. Of course, the manifold marketing materials and click baiting content dedicated to this growing, uh, phenomenon are predicated on the assumption that there was actual intelligence in the HR function to begin with.

This point can probably be debated, which brings us to the larger question: why are we talking about AI in HR, anyways?

The fact of the matter is, the problems most endemic to HR, the biggest challenges facing our profession are inherently the holes in even the most sophisticated AI solutions. Whether in reality or in speculative Science Fiction (or somewhere in between, like your provider’s “product roadmap”), true AI is the HR Technology equivalent of tilting at windmills.

Here are 3 reasons why we have at least a small wonder about why this is such a big deal in our industry.

1. Irrational Actors.

Artificial intelligence is engineered to work exclusively as a rational actor; that is, there’s an underlying logic to AI that requires some form of codified, consistent and standardized inputs in order to produce linear and directly correlated outcomes.

As any HR professional, can tell you, there is no such thing as a “rational actor” when you’re dealing with human beings. Take any HR function, like employee relations or performance management (and especially recruitment) and there are literally an infinite amount of variables which preempt rational action, which is why we still struggle so much with process and productivity.

2. The Breakdown of the Bicameral Mind.


GET MORE TECHNOLOGY POSTINGS LIKE THIS

Want the latest hr tech industry news and talent trends? Sign up for RecruitingDaily and stay in the know.

For intelligence, artificial or otherwise, to be achieved, cognition must occur – and so far, while we’ve gotten deontology down pat, you can’t ever achieve any sort of simulated neurological functioning without ontology – which, we’re finding out, is much harder to code into our core stacks.

In terms of cognition, AI can achieve the necessary first step of deontology – which, similar to the rule of rational action, is simply the existence of rules which govern both machine learning and its ostensible outcome, which is artificial intelligence – quite easily.

Bots are a great example of this – ask them a question, and they’ll tell you an answer, assuming that you structure the question correctly and that it can access the data necessary to provide a response.

This isn’t so different from Boolean search, only using natural language instead of keywords and a question mark as the main modifier for returning searches, only unlike Boolean, that historic information will automatically inform (and assumedly improve) future results. If this is the future of HR Technology, these capabilities have been available to organizations for almost a decade – or back when “Watson” was just plain old Kenexa (which, by the way, had a killer matching engine that no one used back then, either).

quote-before-we-work-on-artificial-intelligence-why-don-t-we-do-something-about-natural-stupidity-steve-polyak-296515

Let’s be clear that deontology is not artificial intelligence, although that seems to be the accepted basis for buzzword usage. Again, intelligence requires cognition, and once deontology is achieved, you’ve got to nail ontology, too.

Ontology is how one deals with and conceives reality – fundamentally, our quest for the meaning of life and reason for our respective existence are both major ontological considerations that even humans haven’t totally figured out.

Human resources obviously still struggle with that “seat at the table” thing because our reality and that of the C Suite are obviously extremely divergent, and it’s that dynamic perspective that makes “artificial intelligence” so unlikely, considering that without a consistent ontology, you can’t ever achieve phenomenology, which is how learning occurs – with or without a machine.

3. All The Feels.

tumblr_ma672asc9i1rxqqtjo1_400AI cannot possess EI – although it can fake empathy, it can never actually achieve it. And without empathy, HR is basically a back-office backwater that should be replaced with automation, not augmented with AI.

So the next time someone tries to sell you on artificial intelligence, use some real smarts for once and realize that no technology on earth can fix what’s really broken in your HR organization. And real people, not artificial intelligence, are what HR should really be focusing on, not the latest tech trend or BS buzzword.

PS: I’m really excited about AI, and believe that it’s not only achievable, but pretty damn close to being a reality in many instances, business or otherwise.

Just not in a function where friggin’ Oracle and Workday are considered cutting edge solutions, frankly. Let’s start with the small stuff, first – like figuring that whole “system of record” thing out.

Editor’s Note: The author has no background in any of this stuff and this is probably mostly speculative, but we hope he sounds as smart as he thinks he is using words about “phenomenology.”

Matt Charney is the Executive Editor of Recruiting Daily. Follow him on Twitter @MattCharney or connect with him on LinkedIn.



  • Andrew Gadomski

    I will be the first to bite 😉

    So I will start by saying that almost all 2017 predictions will likely include something about AI, and unless the prediction says “AI will continue to develop, somebody will have a case study that is barely relevant and transferable, and there will be more failed or inconclusive results than successes” – its probably just smoke.

    However…AI does have a place in recruiting, as data analysis and use of data has.

    It MAY be that AI can be added to the recruiting tech stack to help fill a much needed void (or one that creates lots of waste)…VERIFICATION.

    Verification is basically a process where a system / process is checked to meet a condition. Its the “does it have, does it not” evaluation. We do lots of this in recruiting. Its not judgment…think check boxes.

    There are clearly algorithms and AI that allow for verification, and I think they will have their place.

    The empty promise that I hear about (and Matt outlines above) is VALIDATION. This is more like grading. Its where we ask if something is meeting the need, not simply a condition. It takes cognition and more consideration of variables….and thats why we doubt AI. How could it “think” like us?

    Well…it can. It will verify…but validation? Harder.

    What AI may…and I say that 25x over…is it may greatly reduce a recruiters need to verify within the recruiting process, leaving an opening to increase our investment in how much we validate.

    Nice job MC

    • Matt Charney

      Since you’re the expert on this stuff I can’t tell you how much I appreciate this particular kind of validation. Thanks for taking the time to read and respond, Andrew!

  • John Harney

    Great post Matt. All recruiting products we see today use Semantic Search to provide the underlying data used in AI. Semantic Search is invisible, inaccurate, hard coded, shallow and analyzes to the phrase level, or possibly an entire sentence. But in recruiting, we’re searching for the best overall resume or professional profile. We’ve learned that language-independent (non-Semantic) parsing, indexing, scoring and searching of unstructured data (resumes, profiles, other documents) using customizable taxonomies containing jargon that’s relevant to your business, and allowing users to filter the results and control the ranking (“set the bar”), results in highly accurate, verifiable results that are more precise than Semantic and AI.

    Check out what Guha (responsible for Google Custom Search) says about Semantic search not being applicable to “navigational search,” where the user is using the search engine to navigate to a particular intended document https://en.wikipedia.org/wiki/Semantic_search

    • Matt Charney

      Thanks John. Appreciate your perspective and insights!

    • Talent Analytics

      Hear Hear John! Semantics is definitely NOT machine learning 🙁 There are some companies out there doing it the right way, the data science way… with real analytics and predictive menthods!

  • Janet Dwyer

    #non-Semantic text analytics (language-independent yes)


Just add your e-mail!