Using semantic search to transform novice recruiters into experts
Remember The Matrix? Yeah, that movie about a post-apocalyptic future in which machines enslaved humanity. That summer, teenagers, children, and adults dreamt of dodging bullets and performing acrobatic jump-kicks in slow motion. But not me. The scene that really stuck with me happens before the punching starts.
When our inexperienced hero —with the help of a floppy disk, a cable and a handy plug in the back of his head— goes from not knowing the first thing about jujitsu, to being an expert fighter in two seconds. That’s when I got hooked.
What the movie presented was a tantalizing idea. Is it possible to use technology to transform a novice into an expert just like that?
Turns out it is. For example, it takes recruiters years on the job to learn the terms, the position titles, the skills and the ins-and-outs of an industry. And yet we’ve solved that issue by augmenting the capabilities of new recruiters with clever software that integrates seamlessly into their daily work life.
How do we do it? The first challenge is to write-down this knowledge quickly, precisely and efficiently. You might think it’s impossible for a human programmer, or even an entire army of programmers, to specify what matters in different industries, for different roles, in different countries and companies. And you’d be right. That’s why we use big data. Instead of trying to manually specify what the relevant variables are, we use data analysis to create ontologies that help us estimate how relevant things are in relation to each other.
The second challenge is transferring that knowledge to new recruiters, ideally without drilling a hole into the back of their heads! Imagine two recruiters trying to accomplish the same task, one with years of experience, the other with no prior knowledge of the industry. Their search queries would probably yield completely different results.
Here’s where the magic happens! We might not be able to upload your experts’ knowledge into the new recruiter’s brain, but we can make their search results look as if we had.
Semantic search is designed to understand what users intend to find and then look for that, rather than just scraping for exact word matches. When a user searches for candidates with experience in “machine learning”, the system also includes people with experience in “artificial neural networks”. While our inexperienced recruiter might not be familiar with the industry and the connection between the two concepts, our system is.
However, this technology is only as good as the data it was built upon. This is where Avature Semantic Search differs from other solutions in the market. Unlike most vendors that simply adopt an off-the-shelf tool —designed to differentiate between mundane everyday things, like a motorcycle and a bicycle— Avature took the hard route and built an in-house solution with the explicit goal of maximizing the chance of finding promising candidates in your database. It was built using recruiting data to achieve recruiting goals.
In some ways, this approach is actually better than what the Matrix had to offer. The movie’s version of knowledge, represented by an immutable floppy disk, is fixed in time. Our conception of knowledge is dynamic thanks to our continuously evolving algorithms. Every enterprise is different, so we often work with our clients to build custom ontologies that fit their needs. And the stronger our software becomes, the more it helps your recruiters improve.
Another issue with the Matrix-style knowledge transfer is that you don’t know for sure what’s being uploaded into your brain. What if it’s biased? What if it’s wrong? Because it’s a black box, we don’t know. This has real-world implications. Amazon had to scrap its recruiting AI when it discovered that it was biased against women. For five years, their scoring algorithm penalized resumes containing phrases such as “women’s chess club.” Imagine all the talent Amazon lost because they didn’t know what the system was considering.
To avoid this, all of our artificial intelligence initiatives employ a white-box approach. Instead of keeping users in the dark, the black-box approach that is now standard across most industries, we give our customers visibility and control over what the system is considering. You can see what are the additional terms, you can remove some or all of them. Using Avature, slips in the system can’t go unnoticed, transparency is one of our shields against biases and errors.
This is a fascinating topic that deserves its own article, but the bottom line is this: At Avature we believe in technology that empowers HR leaders, hiring managers and recruiters, not in systems that attempt to make decisions for them. Avature Semantic Search uses ontologies custom-built for HR to do the same thing we always strive for: To put the power in our user’s hands.
