It is increasingly common to hear HR Tech and Assessment companies talk about their AI features. By AI, of course, we mean Artificial Intelligence. In this article we intend to demystify Recruitment AI, define the common pitfalls, reinforce why human oversight is essential, discuss elements used by Harrison Assessments and reinforce how the proper use of AI enhances recruiter decision making.
According to Techopedia, "Artificial Intelligence (also known as machine intelligence) is a branch of computer science that aims to imbue software with the ability to analyze its environment using either predetermined rules and search algorithms, or pattern recognizing machine learning models, and then make decisions based on those analyses. In this way, AI attempts to mimic biological intelligence to allow the software application or system to act with varying degrees of autonomy, thereby reducing manual human intervention for a wide range of functions."
There is a common misconception that AI somehow has independent intelligence. Actually, AI is a complex program, defined and programmed by humans. It can only do what humans program it to do even though it can provide valuable and sometimes unexpected insights. Its "decisions" are based on the information sources we provide as well as the algorithms we have given to interpret. That being said, it is an exciting field of development.
One of the big concerns when considering AI for recruitment is transparency. If the computer provides hire or don't hire results without context, we have no way to legally justify our decision or even know the basis for the decision. That's why AI or Machine Learning needs to be developed in
There was a well-publicized story in 2018 about Amazon developing a recruitment AI that independently decided that being male was a predictive factor. Amazon's AI analyzed thousands of candidates and final hire resumes from the past 10 years for technical developer roles. The technical roles had predominantly been filled by males in the past, and therefore the AI decided this was a key factor. The program failed because it did not identify the factors that differentiated the high performers from the low performers. Furthermore, the factors it identified would have made any hiring inaccurate and illegal. To make matters worse, the so-called Artificial Intelligence was not taught to identify and present the reasons for its hiring decisions. It was effectively a black box completely unknown to the recruiters. This is a dangerous place for recruiters to be in. It is crucial to have human oversight of the data being produced.
When we think about machines making hiring decisions, there is a multitude of emotional responses. Some are excited by the prospect of increasing predictive accuracy and reducing bias. For others it brings a dystopian fear that machines will take over our jobs. Properly designed machine learning can automate parts of the recruitment process but cannot replace hiring experts.
Each job is different and thus requires an intimate knowledge of the requirements to create effective hiring criteria. Machine learning can assist by providing researched data related to similar jobs; recruiters are needed to analyze this data. Analysis enables them to quickly and efficiently select top candidates, perform targeted interviewing to assess behavioral fit for the job, verify answers, and make the final decision.
In short, AI for recruitment is really a cooperation between the machine and recruiter that ultimately increases the value of the recruiter.
We believe that all employment is based on mutually beneficial relationships, beginning with the candidate experience. Our technology shortens the assessment process and increases efficiency, while also reducing human bias and the cost and time to hire. We can quickly identify, interview, attract, and hire the best candidates with Predictive Analytics.
Harrison's recruitment technology is designed to "screen in" candidates (as opposed to the more orthodox "screen out"). This underlying concept is crucial to diversity and inclusion efforts as it focuses only on the skills and behaviors that increase the likelihood of candidate success in that specific job.
Our Job Success Formulas (JSFs) are weighted Algorithms. Eligibility Requirements (experience, skills, qualifications), Suitability Requirements (behavioral fit for the job), and if applicable, Cognitive Requirements (problem solving ability sufficient for the job) are calculated and easily accessed by the recruiter for further exploration and verification.
To further customize a Job Success Formula to your organization's actual performance data and distinct job requirements, our Benchmarking solution uses Machine intelligence to analyze more than 175 factors, identify the factors that relate to success for the specific job, and formulate those factors into a predictive model.
This combination of automation, machine intelligence, rich behavioral data and human insight is powerful and is applied to facilitating great conversations between employers and potential employees about the areas that matter most to them. What we don't do is allow the program to make autonomous decisions without human oversight into exactly the parameters that are being used to assess the candidate.
General personality assessments that do not compare an individual to the specific requirements of a job cannot be legally defended, nor can they be accurate. Similarly, an algorithm that scrapes the internet for a candidate's personal data is not in compliance with hiring standards or legal requirements.
It is essential that psychometric assessments used in a recruitment context are fit for purpose. For an assessment to be accurate and legal it must focus on job specific requirements and use a different algorithm for each job. Which factors contribute to success and which factors derail performance? Remember computers can only do what you program them to do. The real danger in the use of recruitment AI technologies is poor design by humans.
Our assessments comply with EEOC regulations as well as ISO 10667 for Job Specific Assessment and our Benchmarking Analytics option provides job specific research and scientific validation for your specific custom criteria in a way that is legally defensible.
Though it sometimes feels like it, AI is not a magical solution. It is a complex program that helps us to analyze a wide range of complex factors and create an ideal mathematical formula for measuring those factors. It allows us to sort large amounts of data in an efficient, systematic and unbiased way.
AI in Recruitment should enhance human decision making, not replace it. With that in mind, reflect on this paradox: "We use AI or machine learning to make us more human".
Follow Us
