Putting AI to the Test: an interview with our CTO

Our latest white paper ‘Ethics, Equality and Empowerment’ takes a look at the opportunities and challenges posed by AI in assessment. Is AI living up to its promises? Can AI make fair decisions about people’s careers? How can HR and recruiters make sure they’re getting what the need from their assessment tech?

If you’re considering these questions or looking to review your assessment offering, the advice from Sova’s CTO, Jarret Hardie, will help you make an informed decision. Here, we asked Jarret for his thoughts on AI in assessment from a technology point of view.

Is AI living up to its promises?

In our industry, probably not yet. But it’s starting to reach the tipping point where it will deliver on its promises. A lot of what we call AI is machine learning and it’s the ‘learning’ that’s lagging behind rather than the technology itself. In every industry, it’s the time taken to test and train models that dictates the pace of development.

AI in breast cancer screening has been researched, developed, tested, applied and verified and is showing life-saving results. This AI is only answering one question – ‘are cancer cells present?’ – and so this has been a very different development process to what our psychologists are working on at Sova.

How can we get the right mix of technology and psychology?

When we’re working on product development at Sova I am the only technologist working alongside three psychologists so that provides an idea of what we believe the right balance is. We have the computing power and algorithms and we’re finding ways to use it to do what our psychologists are already doing, but at scale.

There are instances where I believe the technology has run faster than the psychology. When this happens, you create risk. For example, I believe there is further work to be done on using AI for natural language processing and facial recognition. In its current form, the understanding of facial expressions throws up huge challenges based on race that could mean people are consistently misrepresented. The psychology and technology need to work in tandem.

How can organisations make sure they apply AI tools ethically?

We’re conscious that our algorithms are influencing people’s careers and livelihoods – it’s a much more important algorithm than your Facebook feed! But algorithms are blunt instruments. They’re powerful but they don’t have ethics so it’s up to our psychologists and data scientists to train algorithms to make unbiased decisions.

We do this in two ways:

• Structured learning where we tell a model what to look for and a human carefully controls the inputs and reviews the outputs and does some statistical validation.
• And unstructured learning where we throw an algorithm at a broad piece of data and say ‘what do you make of this’?

By comparing these outputs and validating the results we in effect instil ‘ethics’ in our models.

How can HR and recruitment buyers differentiate between providers?

1. Don’t look at what the tech promises, look at who is building it. If the focus is too much on cutting-edge ‘whizz-bang’ technology that makes big promises, then question whether enough attention is being paid to accuracy, bias and ethics.
2. One size doesn’t fit all when it comes to assessment. The solution should be psychometrics delivered digitally, not the other way around. Make sure there is a degree of flexibility built in so you can model your assessment outcomes around what your business needs.
3. Hiring the best candidates is still the goal. The technology you choose should be built around the people you need now and in the future. Buyers don’t need to understand the algorithms, but they do need to have faith that it will deliver on its promises.

A final word on what the future holds for Sova

We’re extending our core assessment platform into more of the employee lifecycle; so, from screening questions, into hiring, and onwards into performance reviews, promotions and development. In the near future we’ll be able to offer a seamless digital process for the entire employee journey.

Of course, as well as testing technology that will soon be launched, we’re also researching and testing new concepts that take us further into the future.

To find out more about how AI is changing the world of assessment, download the second in our series of papers on the topic here.