Eliminate bias in AI and drive soft skills development

Amazon abandoned its attempt to use artificial intelligence (AI) for recruitment after a number of years of struggling to make its recruitment engine gender neutral. Despite its best efforts, and brightest brains, its solution continued to discriminate against female applicants. As diversity and inclusion have become increasingly important issues in the workplace, algorithmic recruitment and performance management systems were supposed to be the answer. By removing human bias and workplace politics, the idea was that data-based talent development would be fair and equitable. However, the shortfalls in many of the AI systems developed so far have led to a growing movement for ‘AI for good’ – so, how can we eliminate bias in AI?

The limitations of AI are related to the parameters of its data. As with all computer systems, AI is only as good as the data on which it is based – we cannot eliminate bias in AI if we are not aware of our own bias. So, decisions based on data soon run into trouble if they don’t consider all the parameters. Even intelligent recruitment and performance development systems tend to focus on hard evidence and hard skills, such as technical qualifications and certifications. More often than not, they do not take account of soft (or behavioral) skills, partly because these are perceived as hard to measure. They also do not take account of social context, which relates to an individual’s ability to network and communicate well.

HR expert Josh Bersin has recently described behavioral skills as ‘power skills’. In his view, these are the skills that give individuals real power at work. In fact, he points out, “Hard Skills are soft (they change all the time, are constantly being obsoleted, and are relatively easy to learn), and Soft Skills are hard (they are difficult to build, critical, and take extreme effort to obtain).”

If organizations are to achieve their corporate and social responsibility goals for inclusion and diversity it is increasingly important for them to look at the skills that truly matter across the board, eliminating all considerations relating to the race, religion, or gender of a person, instead prioritizing key soft skills – to eliminate bias in AI we must first begin by looking at ourselves.

Often companies will circumnavigate tender processes, in order to choose a supplier who may not be the cheapest but with whom they have a good relationship and work with more effectively. This is the same with employees. Someone may have all the right technical skills, and even good soft skills such as negotiation and leadership skills, but if they are lacking communications skills and foreign language proficiency they will not be able to work effectively across an international business.

Talent development and recruitment solutions that are largely based on data do not offer a level playing field. Too often, soft skills are not identified in a structured way within organizations. For example, few companies have a clear idea of the level of language and communication proficiency for key languages such as English in their workforce.

Companies can put AI to work to level the playing field in three ways:

According to Bersin: “One of the hottest topics in business today is upskilling, reskilling, and redefining jobs for the future of work. It’s so rampant that 34% of CEOs now rate it one of their top three threats to growth.” AI can help if used properly – backed by human decision-making – by placing value on real performance and an employee’s full range of skills, rather than on opinion.

Intelligent technology can then help provide the right training to the right people at the right time, extending upskilling opportunities to a wider population. It is common to see training provided to employees and departments who shout the loudest, rather than those whose need is greatest. AI can help ‘budget match’ training, so that money is spent on training people who need it the most, delivering the greatest business benefit. And when it comes to training management and saving time and resources on training management, AI and automation are leading the way. When AI is used for good it provides personalized and adaptive learning, allowing employees to build and maintain difficult to access skills such as foreign language acquisition.

It may be used to find out why a certain group of employees is not responding well to development and training, for example.

Artificial intelligence can be very valuable in supporting talent development and even employer branding, as long as care is taken to make sure it does not inherit biases contained both within data and within the selection of data parameters. At the same time, while the use of data relating to individual employees to create personalized training is highly effective, it comes with ethical considerations around protecting that data and using it for the best interests of the data subject. Data should not become the product when AI is used properly.

Machines can identify certain attributes and drive actions based on a set of rules. But human input and a sound training management system  is necessary to check whether what you are doing based on an algorithm is the correct decision and the right thing to do. Just as the best type of learning delivery involves an effective combination of digital learning and human coaching and mentoring, intelligent technology works best in combination with even more intelligent humans to deliver the skills and organizations will require to succeed in the future.