Machine learning is streamlining many industries, from e-commerce to healthcare to finance and social media. The hiring industry is no different, and while technological leaps are hard, progress is already in place to make hiring more efficient and cost effective. Our own VP of Data Science and Analytics, Tammy Wang, penned a post on machine learning in hiring – read on to find out what the future looks like and what one thing she says absolutely cannot be replaced by machines.
(Medium) How Machine Learning Will Shape the Future of Hiring
We are living in an amazing era when cars come to you whenever and wherever you need at the cost of a cup of coffee; when groceries and meals are delivered to the front door. Yet, when it comes to looking for a job, a thing that consumes a big portion of time in our lives and composes part of our identities, we are still experiencing the long, slow, confusing process that has been around for decades. Traditional hiring methods are outdated, ineffective, lack transparency and impact the efficiencies in hiring that ultimately affect the progress of everyone’s professional lives.
With the advancement of computing technology and decades of data accumulation, it’s now time for the hiring industry to follow the footsteps of many businesses and industries where technology and science have already had a dramatic impact — like e-commerce, healthcare, finance and social media.
In hiring the most difficult problem to solve is to predict the “fit” between the company looking to make a hire and the potential candidate. In other words, to create a perfect (or at minimum very strong) match. The challenge then becomes how to train learning algorithms in this exceedingly complex space, and while it might seem impossible to pinpoint the right person to a specific job, it’s actually a great efficiency gain to even just narrow down the candidate discovery range.
Why Hiring is So Complicated
Existing machine learning platforms have gotten incredibly adept at making highly targeted product recommendations for people as there are well-developed applications to capture the relatively static features of products, such as genres of books or movies, color, and shape of products. There is also a large volume of past purchasing or browsing behavior to build each individual’s taste and preferences. In hiring, that same type of matching algorithm is much more complex when the goal is to match a human to a group of humans, both of which have an infinite depth of preferences and hidden decision factors.
A job candidate is a complicated person and hard to describe in the “feature” space like a physical product, while the hiring end is even more complex because it not only factors in the human element but also group dynamics and each candidate has to “fit in” both technically and culturally.
The second complexity is the dynamic nature of the candidates and jobs. Both ends of the matching equation are constantly evolving as people develop skills and experiences, and companies or businesses evolve as markets change.
It’s also worth noting that people don’t hop around that frequently in the full-time job space (compared to the data that’s readily available around purchasing activities, for example), further limiting the amount of data available to advance machine learning. Job hunting or hiring domain data is not only sparse but also highly fragmented. People often search for jobs in one spot, submit resumes from another channel, and the interaction data are largely kept in the hiring companies, not available for a platform to process the end-to-end funnel to build good learning algorithms.
Hiring Now and in the Future
In the hiring space, technological leaps are hard, but signs of progress are already in place. Much of this progress is due to the ability of learning algorithms to spot patterns in data. Systems can scan through a big collection of candidates and use predictive signals to identify those who are relevant. For example, it had been observed that people’s career movements are very rhythmic: as if there are intrinsic leitmotifs in each of us or reiterative behaviors that can be identified. Some people tend to hop around, while some like to stick to the same place. These patterns can inform recruiters around the timing of the candidates, and only propose opportunities to those who are ready. This modeling approach yields better candidate experience and recruiting efficiency. Moreover, advancements in natural language processing technology provide a powerful collection of tools to help companies map candidate experience in skills, specialties and leadership domains so the right candidates can be identified with confidence and precision, reducing both costs and resources in the hiring process.
Since the subject of the recruiting market is human, machine learning has the potential to generate negative impact: it could repeat, or even amplify, the human bias in the process, such as gender or age discrimination. These concerns are valid because machines learn from behavior data and try to imitate the practice, however, we remain optimistic that with data, machine learning is also good at parsing out conditional probability, like bias. With well-structured learning algorithms, not only can the biases can be detected, the process can also be adjusted to an unbiased and more accurate decision support tool.
If we look toward even just five years from now, we’ll experience big improvements when companies and potential candidates reach out to each other mediated by AI or Augmented Intelligence. We anticipate a “human in the loop” recruiting technology will shorten the time to close a position. We also see the interaction between those hiring and job candidates will be supplemented by natural language bots that can service the candidates or companies 24/7, and in the most insightful ways. For example, machine learning can help candidates parse out differences among large ranges of potential jobs and project the outcome of a particular choice. This type of transparency, especially around information organized to help people, will help the candidate make more rational decisions and thus build happier work environments.
If we look ahead 10 years, jobs and the job market might have a very different meaning to people. Job candidates and companies seeking to hire will be able to relate to each other in a smooth and intuitive way. People will be able to find the “best” job to accommodate personal lifestyles, passions, and motivations.
Machine learning and artificial intelligence hold the keys to streamlining many industries, not just hiring, by automatically mining massive data sets, making predictive inferences of behavior, and predictions around relationships. And while all of this is truly life changing, there is one key thread that ties it all together that absolutely cannot be pulled from the cloth — the human element. Especially in recruiting, no matter what stage of development these technologies reach the human element will remain essential. Humans will always be responsible for understanding emotion, applying intuition and common sense, and making ethical judgments — qualities that machines will never have. Conversely, machines are capable of computations at a speed and scale the human brain will never be able to perform, which is why the idea of man plus machine is truly how the magic happens, as each side brings something to the table the other will never be able to master. When man and machine intelligence come together, that’s when machine learning has the power to shape the future of how we hire and how we work.
