18th Century Math Will Get You Your Next Job

Many of us assume that everything companies invent is brand new, but sometimes the things that make cutting edge-technology run smoothly are borrowed from centuries-old statisticians.

Take, for example, artificial intelligence (AI). The initial concept can be traced, at least in part, to English theologian and mathematician Thomas Bayes – all the way back to 1763.

Bayes’ theorem was the first to establish the mathematical basis for inferring the probability of an event based on empirical observations of related conditions. This theorem is part of what Amazon uses to recommend your next purchase, and what Spotify employs to suggest your next favorite song. In order for IBM’s Watson to understand and answer Jeopardy questions, or for Netflix to recommend what you might like to watch next, AI works in the background to detect relationships and understand where matches should and should not be made.

Bayes’ 18th century AI is surprisingly relevant today, often competitive with far more complex modern techniques. In a nutshell, the theory goes that if you know the probability of A given B, as well as the probability of A and the probability of B on their own, you can calculate the probability of B given A. The key is having prior knowledge of conditions that might be related to an event. As you make new observations, Bayes’ theorem tells you precisely how to update your calculated probability accordingly.

What Bayes Means to Recruiting

The same intricate process that detects relationships between factors is what helps online shoppers compare prices of the same or similar products, government officials normalize public health data, and credit card companies detect fraudulent charges. It is quickly becoming a big deal for companies that want to take a smart approach to recruiting as well.

Recruiting software uses “entity resolution” to systematically search the web for public profiles and aggregate them under the author’s identity. What this means for any potential job seeker is that if you’re posting your code on a platform like GitHub, or sharing that you’ve recently updated your Twitter profile to showcase new skills, new recruiting solutions see these changes and make sure that companies notice them, too.

The resulting database of billions of job seekers may seem like a treasure trove to recruiters, but without Bayesian statistics, that database would just be a messy lake of dirty data and overlapped profiles. Those job seekers would never be able to connect with employers.

There can be significant confusion in the entity resolution process, depending on how many social profiles someone has or how many names or nicknames apply to them. There are countless factors involved in attributing authorship, such as the exact words used to describe a job or school. For example, John might also call himself “Jon,” “Jonathan,” or “Johnny.” He might say he went to “Berkeley,” “UC Berkeley,” “the University of California,” “UC Berkeley Engineering,” or even just “Cal.” Johnny the salesperson can be a “sales engineer,” a “success manager,” or even an “enablement guru.” These titles change, too, depending on which social media platform a person presents themselves on. This is where Bayes’ work comes in.

Bayesian statistics can be applied to determine the likelihood of two social profiles belonging to the same person by precisely quantifying the chances that two observations refer to the same person. For example, two profiles with the names “Jon” and “Jonathan” are much more likely to refer to the same person than two profiles with the names “Jon” and “Linda.” By repeating this analysis many times over many factors and tallying the results, coherent profiles can be built from many data sources. This is what allows companies to find ideal candidates across the web, based on attributes that may be displayed differently across an individual’s myriad social profiles.

So, the next time a platform suggests the very thing you want or need, you’ll recognize those cutting-edge eighteenth century mathematics at work. And the next time you’re describing your work skill set on a public social profile, feel free to say you’re a “social media maven” or a “Ruby on Rails ninja.” Thanks to Bayes, technology is now smart enough to understand what you mean and match you in the job market with prospective employers accordingly.

Cole Goeppinger is vice president, technology, at Entelo.