"All-Natural" Recruiting: Going Beyond the "Artificial Intelligence" Snake Oil

HR technology companies are capitalizing on the buzz around artificial intelligence (AI) in the consumer world to sell the future of AI for recruiting. That’s a future that we’re excited about, as we think RPO holds an exciting opportunity to lead the way to that future, but there’s also a lot of snake oil being peddled in the market. It is unfortunately easy to do, because there is a lack of understanding of what artificial intelligence is, what its related technologies are and what its possible uses are in recruiting. As TechCrunch recently coined, AI is becoming like “all-natural” – a phrase that means nothing because it lacks a definition.

Lacking a definition of AI and its related technologies, we’re susceptible to confusion. Many of the tools that were dubbed “predictive analytics” a couple of years ago have simply substituted those passé buzzwords for the newest – artificial intelligence. As talent leaders, we want to trust that a vendor’s software does something that uses an AI computing principle, but how do we know it will deliver on the promise of intelligence that the phrase implies? There’s a big difference between, for example, a logic-based chatbot and one that is backed by machine-learning and a neural network. The intentional confusion created by vendors hoping to jump on the AI bandwagon has a chance to set back the industry in its adoption of tools that are truly sophisticated.

I find it frustrating when authors write articles about what can’t or isn’t being done without offering solutions. So, here is my shot at some definitions that will further our common understanding of AI. Below is what these terms mean technically, along with some examples of what they might mean in the context of recruiting. If you’ve got more to add or want to quibble on a definition, drop a comment and let’s explore together.

TERM DEFINITION RECRUITING EXAMPLES
Artificial Intelligence An umbrella term referring to a broad set of technologies that make software able to seem humanlike to an outside observer. Any number of the examples below might be considered as using AI in recruitment.
Machine Learning A subset of AI that recognizes trends from data. Common Machine Learning techniques include neural networks, support vector machines, Bayesian belief networks, decision trees/forests, k-nearest neighbors, self-organizing maps, regression techniques, Markov models, instance-based learning and case-based reasoning. Machine learning is the most commonly used AI technique in recruitment, particularly applied to the problem of matching candidates to jobs. In these scenarios, the criteria upon which the machine makes a match are altered based on user input – a recruiter sorting candidates with left and right swipes, for example – informs the machine as to what candidates fit the job.
Neural Networks A type of machine learning loosely based on how neurons work in the brain. A subset of Machine Learning, neural networks will have similar applications as the larger set of Machine Learning, such as candidate matching based on multiple factors (personality derived from text, social media behavior, recruiter inputs against previous successes of candidates who were similar).
Deep Learning A complex or “deep” neural network with many layers. Deep learning in recruiting may manifest itself in a deep neural network or combination of those networks that interact with each other.
Cognitive Computing Focused on reasoning and understanding at a higher level, often in a manner that is analogous to the human condition. Deals with symbolic or conceptual information with the aim of making high-level decisions in complex situations. Cognitive recruiting would likely entail multiple AI-related processes to simulate the actions of a human recruiter in a wide range of situations.
Machine Intelligence A synonym for artificial intelligence.  
Natural Language Processing NLP is a branch of artificial intelligence that deals with analyzing, understanding and generating the languages that humans use naturally in order to interface with computers. NLP helps hiring managers speak or write their requirements for a position to aid in matching the requirements of a job to candidates for the job.
Artificial General Intelligence For purists, true AI needs to pass a Turing test, which has only been accomplished by some of the world’s largest supercomputers. A Turing test requires that a human judge be unable to distinguish between the answers of a machine and a human in a blind test. General AI solves from a broader set of problems than narrower AI. A future state where a computer is indistinguishable from a human recruiter to a hiring manager, candidate, or other human interactor.
Interfaces AI often uses non-traditional interfaces to simulate how a person would interact with another person rather than a traditional computer. Chatbots, voice interface (Alexa, Siri, Google Home, etc.). Interfaces that simulate human activity can be AI-based or logic-based.

 

Post contributed by Cielo’s Vice President of Technology, Adam Godson. Follow Adam on Twitter @AdamGodson or connect with him on LinkedIn.

Comments

David Gossett 1/24/2017 1:19:01 AM

1/24/2017 1:19:04 AM

For machine learning to be pervasive, it must first be pragmatic. To be pragmatic, ML has to respect the humans it serves. ML is coming no matter what, but if we want to be first in line to really make a difference, ML has to team up with current business processes. This is a really hard lesson for a technologist to swallow. We like to break the china, live inside the code. But this time feels different. It's time for IT to have a little humility, in my opinion.