By Kristin Cifolelli, courtesy of SBAM Approved Partner ASE
Can a computer algorithm be a better predictor of a successful hire than a manager’s decision based on work history and a job interview? That’s what many companies and various software vendors are betting on. They are using complicated algorithms to sift through thousands of resumes to filter and recommend the most qualified candidates. They hope that they can cut the amount of time involved in traditional recruiting and make better-matched hires that avoid costly turnover.
Using algorithms is more than just using key word searches to eliminate candidates. Bright.com is an online job matching tool that was launched in 2011. Very much like computerized dating sites, it uses a computer algorithm to match potential candidates with job openings and hiring managers. It analyzes many variables including education, prior employment, and skills. It also uses that information to infer whether a candidate would be successful in other areas not listed. For example, if a job candidate is in public relations, it will be assumed that they will also be a good speech writer, even if those skills aren’t listed. According to Bright.com CEO Steve Goodman “we take your resume and build a bigger resume around it.”
The algorithms take the matching process a step further by analyzing an employer’s hiring decisions to identify patterns. If an employer tends to recruit employees from certain schools or avoids candidates from certain companies, the program will detect theses patterns and adjust candidate recommendations accordingly. According to Bright.com’s Chief Scientist David Hardtke, “Over time, the algorithm learns” in the same way that Google learns what their users’ interests are by documenting their search history. Once all the criteria and hiring patterns are analyzed, the algorithm assigns each candidate a score and then reports the results to both candidates and employers.
Other data mining companies are looking at developing algorithms to analyze blog posts and tweets—i.e., the candidate’s so-called “cyber imprint”—in order to predict workplace success. Some programs are incorporating the results of personality tests. These programs are powerful tools that allow for evaluation of more candidates, amass more data, and peer more deeply into an applicant’s personal life and interests. Xerox is using software with an algorithm to make recommendations for all of its call center jobs. Their program determined that prior experience didn’t matter when it came to predicting a good hire, but that a certain personality type did (creative personalities were more successful; inquisitive types were not). After a six-month trial, they have cut attrition by 20%. Other companies have used algorithms along with the results from personality profiles to help determine which candidates have a higher likelihood of getting injured on the job, stealing, or abusing workers’ compensation.
LinkedIn has developed its own algorithm that it is offering to customers who purchase its profile searching system called “Recruiter.” It analyzes the profiles that recruiters show an interest in and then creates a list of prospects that have similar careers, skills and education. If a recruiter starts a dialog with a candidate via email, then that profile and others like it get a higher priority than the other “qualified” profiles. Then it narrows the subsequent search to profiles that more closely resemble the one(s) contacted by the recruiter. Over time the algorithm will adjust its profiling pattern in order to provide better matches.
Some of the benefits that may result from using an algorithm are that it provides consistency in analysis and can significantly reduce the impact of a hiring manager’s bias. But there are potential legal risks if the algorithms are not programmed correctly. Analyzing large amounts of data may increase the statistical relationships that could inadvertently screen out protected groups such as older or minority candidates. If someone challenges a company’s screening process. it will need to demonstrate that the criteria used to build the algorithm are relevant to the job. The law firm Jackson Lewis, LLP has reported that it is getting an increasing number of requests by organizations asking for an evaluation of various software tools to ensure there are no unintended discriminatory recruiting practices and that they are complying with EEO laws. Another significant potential downside is that these filtering systems can also screen out some qualified candidates.
Ultimately, despite the savings in time and money, and other benefits these programs can demonstrate, they should supplement, not replace, traditional hiring decisions that HR professionals and hiring managers make. Instinct and intuition still are important parts of the recruiting process. Despite all the best computer models, human behavior is not predictable. Some of the best candidates may have non-traditional profiles and be overlooked, to the detriment of both them and the employer. Some of the best talent can be found in the most unusual places.