Artificial Intelligence and Machine Learning in Hiring

May 13, 2019

A common theme over the course of my MBA/MS in Information Systems degree program at Boston University was how artificial intelligence and machine learning can be utilized to increase business efficiency and effectiveness across an organization. For my research project in my final year. I chose to look specifically at how AI/ML are being used in hiring. Using these tools in this sector is of particular importance because it can have an impact on individuals both inside and outside of an organization.

Current environment

According to Deloitte’s 2018 Human Global Capital Trends report, 68 percent of companies either currently are, or plan to use AI for hiring.1 This technology is most commonly used for analyzing pre-hire assessments, text mining digital resumes, identifying promising passive candidates, and minimizing churn. Using AI for these tasks has obvious benefits like saving time and money -- companies no longer need armies of recruiting professionals to sift through thousands of resumes.

In addition to these tangible benefits, companies also see the use of AI as a way to combat unconscious bias in the hiring process.2 Leaving humans out of the equation in the initial screening process would theoretically create a more equitable playing field for applicants. However, corporations large and small -- and especially those in the tech industry -- continue to face criticism for their lack of gender and ethnic diversity despite their visible efforts to fix the problem. This suggests that AI and machine learning tools, as they are used now, are not living up to the promises of vendors and AI evangelists.

Root of the Problem

Anyone who has attempted to build a meaningful predictive model can tell you that your ability to predict is only as good as your inputs. An AI or machine learning model will continue to produce biased results as long as its human creators continue to include inputs based on biased and flawed assumptions. The following are a few examples of how AI can exacerbate a problem that it is attempting to solve:

Recommendations

To reduce bias and develop a more diverse workplace, companies using AI and machine learning tools need to understand that this technology is not a standalone decision-making entity. AI needs human judgment.4 If given the opportunity, I would make the following recommendations to any company using AI to reduce bias and not seeing any results:

Developing tools to reduce bias in hiring is not something that must be done in-house. There are a growing number of startups including Blendoor, Entelo, and GapJumpers with a mission dedicated mission of using AI tools to increase diversity.

Broader Lessons

Working on this project and considering the MS/MBA curriculum as a whole has made me more skeptical of voices who tout the latest technology trend as the broad solution to a business problem. In the case of AI and hiring practices, without the oversight and judgment of humans, using this technology does nothing but worsen a problem it’s trying to solve. This suggests that there really is a need for individuals who can explain technological concepts to a more general audience. It’s necessary for a business owner or a department leader to understand the resources needed -- human and technological -- to achieve a business goal.

Similar to other areas in which AI can be applied, recruiting is another place where certain jobs will disappear while others will emerge. As with all things AI and machine learning-related, there will need to be data scientists who develop and tweak models. But possibly more importantly, there will also be a need for tech-savvy people who understand the nuances of HR, diversity, and bias. It will be these people who quickly identify issues and ensure that AI and machine learning tools do in fact live up to the goals set for them.

12018 Global Human Capital Trends, Deloitte
2How Artificial Intelligence Can Eliminate Bias in Hiring, CIO Mag
3Personality Tests are Failing American Workers, Bloomberg
4Smarter Together: Why Artificial Intelligence Needs Human-Centered Design, Deloitte Insights
5The Machine-Readable Workforce, MIT Technology Review