Artificial Intelligence: Mitigating Identity Fraud In Hiring Practices

As a best practice, identity verification should be the very first step of the interview process

Artificial intelligence (AI) is a term that often makes HR executives cringe – they think of their jobs becoming superfluous in the midst of automation, replacing human decision-making with algorithms. There is the negative association that AI will replace the need for HR altogether.

While true that AI and machine learning are redefining how decisions are made on hiring, development, and termination - AI instead of removing the need for HR is removing the bias from the tough workplace decisions executives need to make. Much of the data provided to a talent acquisition team will certainly be anecdotal in nature – references, cover letters, interview responses and the like.
 

However, the realm of fact-based data – such as identity validation, criminal background or driving history all provides a very cohesive picture of not only identity verification, but behavioural patterns that are directly correlated to organizational risk posed by a candidate. As an example, learning that a candidate has a history of assault in the workplace or risk-taking behaviours that endanger their peers is certainly very valuable data to have prior to hire.

Many HR professionals avoid criminal screening altogether, concerned about the potential for bias or affirmative action dangers. Or the candidate is permitted to provide the data direct to the hiring employer - a concerning path, when considering how easy falsifying documents can be in a world quite familiar with Photoshop or photo-editing tools. As a best practice, identity verification should be the very first step of the interview process – ensuring the candidate you are investing your valuable time with, is indeed who they say they are.

Identity verification is a perfect channel for both facial recognition tools and optical character recognition to be utilized as channels of AI. Facial recognition when used as a vehicle to ensure a match between the individual and the ID they’ve provided is essential to ensuring identity authentication. Facial recognition uses the dimensions of a person’s face to compare to a photo of an individual – often technology will provide a “confidence rating.” Human faces change over time – we wear our hair differently – we age – a confidence rating is a numerical representation of the match between a face and a photo. If you’d like to learn more about facial recognition - https://www.gemalto.com/govt/biometrics/facial-recognition provides some excellent insight.

Optical Character Recognition (OCR) functions to assess the validity of any ID itself. OCR utilizes complex algorithms that are designed to convert images (PDFs, scanned IDs, etc.) into text. A photo of an ID card can be utilized not only to extrapolate data potentially to automate the creation of a candidate profile, but also to determine whether the ID provided is fake. Often fake IDs are provided to HR professionals – simply because it is very difficult to detect a fake ID without any authentication technology available.

OCR can not only check security features, like signatures, holograms, etc. but it can compare the ID to a library of IDs that have been data mapped. Data mapping refers to the positioning of certain pieces of information in exact locations on an ID. For example, the Ontario driver’s license has a minute extra spacing between certain characters. This deliberate extra spacing would be undetectable by the human eye, whereas OCR would immediately detect a fake. Microprinting is another security feature in IDs – it refers to the way in which text is printed.

Text can be printed to produce specific patterns that a human being could not detect, but a machine could easily identify on an ID. Identity verification is essential as the world of work changes to focus on an increasingly contingent workforce. Candidates are often shared across organizations. Unfortunately, there exists a tendency to avoid screening in the contract workforce as organizations don’t have the typical employer duty of care requirements.

However, many contract workers are interacting with the public in unsupervised settings – in situations where the “parent” organization will never meet said candidate. A perfect example is the sharing economy model of ridesharing – the absence of the employer/employee liability does not absolve the organization from a responsibility to public safety. Machine learning is a subset of AI in which technology that identifies patterns can “learn” over time as it is exposed to more and more data. Similar to the way human beings improve our responses and avoid making repeated mistakes – machine learning is an excellent tool to help with adjudication matrices.

An adjudication matrix would be one where an employer has a set list of “rules” that analyzes screening data provided. For example, driver’s abstracts can be compared to a list of qualifying rules that would “pass” or “fail” a driver. Machine learning can be employed in the example of OCR reading a photo of ID incorrectly – i.e. reading an “O” as a “0” – in this example reading a character incorrectly could then result in the technology not submitting the right identity verification data for a background check.

Machine learning would be the solution to correct the mistake in that example. While AI is a fascinating world that continues to grow and evolve, one fact is unchanging – HR professionals and executives are essential to making decisions. AI is simply a vehicle to provide unbiased data.

Sarah Baldeo is Senior Director, Screening, at ISB Canada Connect Sarah Baldeo Visit www.isbglobalservices.com