Unbiased Tech

COVER FEATURE – Simon Kent discusses the strengths, weaknesses and challenges tech brings to diverse recruitment.

Forgive me for mixing metaphors and mythology, but if diversity is the Holy Grail of recruitment, is technology the Silver Bullet? Certainly the ever increasing presence of AI and machine learning in the sector would suggest so. Here is something which will purely crunch the numbers, analyse outputs and draw conclusions based on figures rather than concepts or preconceptions. But when an employer is seeking to maximise the value of diversity we still have to ask is this enough?

The reason for that question can be seen in the responses of recruiters to the idea that technology equals diversity. “When used effectively, technology can be a great way to help eliminate bias in recruitment,” says James Calder, CEO at Distinct Recruitment. “Technology doesn’t have emotions and mitigates the likelihood of decision making occurring that’s based around feelings. This means that any choices made will not be implicated by any subconscious, or even through conscious bias. Using programmes that sift through candidate data – ignoring gender, race and age as examples – recruiters can often identify candidates who meet the required criteria while reducing the bias inherent in humans.”

Ross Crook, vice president at Cielo is also enthusiastic about the impact of technology in the diversity arena: “Organisations that are growing quickly often haven’t had the luxury of time and can fall into bad habits when it comes to hiring – allowing unconscious biases to flourish,” he says. “Technology can therefore be a big advantage for the talent acquisition function – having both time saving capabilities and the ability to screen candidates and select the best interviewees without prejudice, most systems will focus on the main candidate criteria, such as experience, qualifications, and skills, which reduces discrimination; they’re programmed to ensure that they don’t run the risk of having any unconscious bias that humans occasionally do.”

For Cielo, technology has helped to assess assumed future performance indicators and verify if they are valid or not. Ross gives the example of a financial services client where Cielo found that where people studied had less relevance to their future success then the subject they studied. Excluding this information effectively removed a social mobility barrier. “Vitally, a human touch is always needed to provide vital candidate experience support,” asserts Crook. “Technology and humans need to work together to provide a fluid experience.”

However, people and technology working together doesn’t always result in an unbiased output. Technology is great at following set rules and following those rules over and over again without question or error. Learning technology can get to grips with what is viewed as ‘success’ and gradually adapt to deliver that ‘success’ more frequently. But that does not necessarily result in diversity. 

Bias present

Jozsef Blasko, senior HR director for the Coca-Cola Company notes that technology which relies on following algorithms to make selections is only going to be as non-discriminatory as its design and implementation. Suki Sandhu, founder and CEO of Audeliss agrees: “There is a danger in treating AI as a quick-fix tool for workplace diversity,” he says. “After all, algorithms aren’t neutral by nature just because their consciousness is built on figures.

“While over a quarter of Brits believe that artificial intelligence can bring a fairer hiring process, the very foundation that this technology is built upon demands input from engineers; it learns from the pre-existing, real-world data that it is fed and develops its own behaviours in accordance with the training it has received,” he adds. “In turn, heavy reliance on machine learning software in the hiring process comes with a number of risks. Primarily, unreasoned unconscious bias.”

James Calder cites the recent experience of Amazon who ultimately deserted their AI solution as it appeared to be creating selection problems of its own: “It is said that the tool was penalising the CVs of female candidates that contained the term ‘women’ or ‘women’s’ and is believed to be as a result of the system being trained on data submitted by people over a 10-year period, most of which came from males,” he says. “It is important to recognise that whilst algorithms are not told to be biased, they can become bias through the data they use.”

The old phrase ‘rubbish in rubbish out’ springs to mind, but this is more complex. Without careful analysis of what the technology is actually doing, how it is doing it and the data it is using to carrying out those actions the ultimate output can still be biased. More dangerously, if left unchecked, the scale and speed by which technology can operate means a company can find they have a significant problem with their recruitment process very quickly. The idea that technology will protect an organisation from charges of unbiased selection cannot be taken for granted. The use and management of technology needs to be as closely considered at all stages of the recruitment cycle as the make up of an application forms, the type, format and placement of a job advertisement.

A deeper dive

However, the argument around the use of technology to enhance diversity opens a more philosophical rather than practical discussion challenging recruiters to consider, fundamentally what it is they should be measuring or analysing among their candidates. Compare and contrast the opinions of two leaders from companies who are using e-platforms for their companies and recruitment. Hoxby Associates is effectively an online collective of freelance workers connected and managed by a central organisation in order to provide project work to clients. Alex Hirst, joint CEO, describes their online application process as one which first and foremost ensures anyone who wishes to join Hoxby absolutely and completely identifies with and fits into their way of working – or ‘the movement’.

“We don’t ask for any kind of information that might be perceived as biased,” says Hirst. “We don’t ask for their age, gender, ethnicity and so on – just their motivations for working. The first filtration is done by human judgement and that is to do with their connection to the movement. There has to be a clear personal connection.”

Morten Petersen is CEO and co-founder of the worksome platform which also delivers freelancers to clients, to a different model. “When it comes to hiring, the ‘personal fit’ is very much overestimated,” he says. “It’s more important that the hiring process starts and ends with finding the right candidate who knows how to do the job, who can solve the problem and complete the tasks in hand, rather than relying on an existing network of professionals who may ‘fit’ but may not have the competencies to fulfil the role.

“Once you have a shortlist of candidates that can clearly do the job, then you can turn your attention to finding the right personal fit.”

While these two statements may seem contradictory, they are the result of each organisation’s view, the culture and work model those organisations are seeking to build. Both eliminate bias but foreground different elements in their first sift but then Coca-Cola’s Blasko throws the whole debate open once more by questioning what exactly is meant by diversity for employers in the first place.

Blasko explains he perceives two different types of diversity – a diversity of identity covering sex, ethnicity, race and more which are for the most part demographically related. The other type of diversity is cognitive diversity. “If you have a group of highly numeric people and you’re recruiting into that group, do you want another highly numeric person in that team or are you looking for someone different in order to make that team more productive?” he asks. “I would be keen to see whether AI can work on this.”

Furthermore, if technology and employers want to and do achieve cognitive diversity – which arguably is the ultimate aim of diversity in the first place (otherwise is it just a tick box exercise?) – will that result in a diverse organisation in other respects? Does a fully cognitive diverse organisation also deliver a diverse organisation in terms of race, sex, and demograhpics?

Hoxby Associates has no idea whether its 450 members are diverse at all – because they don’t collect that kind of information. Hirst speculates that they may not be considered very diverse simply through the way the business has grown (organically and word of mouth) since its inception. 

Certainly technology can make unbiased decisions for recruiters – if it is designed and set-up in an unbiased way in the first place. But what this discussion also highlights is the need for organisations to determine what ‘diverse’ looks like for them in the first place. Certainly the trend towards measuring pay gaps and so on will power this process forward, but care and consideration needs to be taken at every stage if the real value of diversity is to be achieved.