Technology Driven Diversity
Our DNA
Job.com and its founders have been committed to diversity in the workplace for 20 years. Our CEO Paul Sloyan had the first diversity accredited staffing agency in the UK in 2003. Our Chief Visionary Officer has been at the forefront of publishing various written works regarding technology issues throughout the hiring process that actively stimulate diversity issues. As a business and as founders, Paul Sloyan and Arran Stewart are committed to serving the recruitment and staffing industry with a greater level of diversity and equal opportunity. Our commitment runs into senior management with our President of Staffing, Steve O’Brien, a former IBM RPO head of staffing and who is listed within IBM Watson Patents, cascades diversity throughout his practices and currently sits as an advisor to the recently funded job board hireblack.com. Diversity is in the DNA of Job.com.
Our execution towards this begins with the way we build our technology to help drive towards diversity. Diversity in organizations starts with the first click of a matched job.
Background
Over 60 million people change jobs annually in the USA with advertised jobs receiving an average of 250 applications per role. The average recruiter will spend 7.4 seconds reviewing a resume, meaning that today, with digital and billions of candidate applications to review, the rise of artificial intelligence matching software has become essential in order to cope with the top of the funnel talent pipeline.
AI matching software uses supervised machine learning algorithms in order to continuously improve, grow taxonomies, and enhance its level of learning to find the most suitable candidates for the job. However, the process of supervised learning up until today has been deeply flawed, especially in industries such as technology.
The Problem - Two Layers
First Layer - AI Unconscious Bias
AI uses an existing pool of resume talent and selection outcomes to inform the subsequent matching and ranking of submitted applications. However, what happens when the previous set of successful resumes were all white males?
Scenario
These ‘white males’ may have also gone to a select few universities and may also come from a similar geographic area. This collective group of similar individuals will inherently have a similar writing style and will talk (for example) in masculine language.
The issue of unconscious bias was famously uncovered by Amazon over a four year period as they attempted to create their own recruitment matching software. It uncovered a fundamental flaw within the learning of their system, which was they had created AI that was biased against women. It was this information that provided the evidence that unconscious bias would stretch beyond just "masculine and feminine" in fact, impacting race, culture, and geographic biases too. Essentially, AI would go into a ‘rabbit hole’ and would only ever reinforce bias and a lack of diversity.
Layer 2 - Remove Discrimination Triggers
A research paper written in 2016 called Whitened Résumés: Race and Self-Presentation in the Labor Market (link to read PDF at the end of document), uncovered alarming statistics that have been illustrated below:
It uncovered that by simply taking the same resume and attaching a white name to it, rather than an African American, or Asian name for example, you would increase the chance of a call back from a recruiter by 150%. It’s alarming that in today’s society this level of inherent discrimination still exists, but sadly it does and technology is needed to overcome such an issue.
On a mass scale, if you think that the process of simply anonymizing every candidate that comes through the hiring process feels like the natural progression towards overcoming this issue, to a degree, you would be right.
But, opening up the hiring process to completely anonymous hiring also leaves room for abuse, as we have seen in testing trials using anonymized data.
Graphic by Blair Storie-Johnson (Source: “Whitened Résumés: Race and Self-Presentation in the Labor Market”)
Solutions
Un-training Bias within AI
The route to overcoming this is to first accept that bias within AI is inevitable. Firstly, if you statistically have 80% of the labor force creating the system being white male, putting parameters in place would only ever enforce their own unconscious biases. Secondly and most importantly, it comes down to the data set used to train the AI, which must be diverse itself.
In 2015, the staff breakdown of Google was 80% male, of which 60% of those males were white, with merely 1.9% being Black. Black female technology workers were almost nonexistent in the company. Using their dataset makeup of the most successful candidate for Google would always look for the white male if you took the entire working population of Google as the training set for supervised AI matching.
Job.com created a process that allows us to completely anonymize a candidate from a resume, leaving only the necessary matching and shortlisting data needed to determine whether a recruiter would like to ask more skills or experience related questions and request a 1st stage interview with a candidate.
However, if you took a dataset of resumes, all of which had one thing in common - that they were successfully hired and worked for Google, but used equal amounts of Male/Female - White/Black/Asian/Latino and used this as the training set for AI’s machine learning, you would, in theory, create a much more diverse viewing AI based on commonly used language by each subset.
The outcome we believe will result in the penetration of more diverse amounts of resumes reaching the recruiters and hiring managers of companies just like Google, to help them shortlist more effectively, with less impact towards their ability to recruit diverse hires. This segues into the second layer of diversity issues within the hiring process, to which technology can simply overcome discrimination by recruiters.
Summary
Job.com continues to develop candidate and jobseeker career history verification using a hybrid cloud blockchain technology stack. This will allow our system to verify the validity of a candidate’s profile and history, then at the point of verification, we'll provide an anonymized application to the employer.
Job.com is in the pursuit of a more diverse and equal workforce, the process and work required for this in technology is a continuous process. We continue to focus our efforts on developing technology that actively promotes diversity and overcomes the two major issues highlighted in this document.
References
https://zety.com/blog/hr-statistics
https://becominghuman.ai/amazons-sexist-ai-recruiting-tool-how-did-it-go-so-wrong-e3d14816d98e
http://www-2.rotman.utoronto.ca/facbios/file/Whitening%20MS%20R2%20Accepted.pdf
Notice to Recipients
By accepting this confidential material, you acknowledge, understand, and accept the following:
This material has been prepared by Job.com (“Job.com” or the “Company”) and it may not be used, published, or redistributed without the prior written consent of Job.com. The material is subject to change without notice. This document is for discussion and illustrative purposes only. It is not and should not be regarded as “investment advice” or as a “recommendation” regarding a course of action, including without limitation as those terms are used in any applicable law or regulation. No representation or warranty, express or implied, is made by Job.com as to the accuracy or completeness of such information. You may not rely on the material contained herein or the accuracy or completeness of any other written or oral communication transmitted or made available. Accordingly, no representation or warranty, express or implied, is made as to, and no reliance should be placed on, the fairness, accuracy, completeness, or correctness of the information and opinions contained in this document.