The AI Assessment

Simon Reichwald, Chief Progression Officer at Connectr considers the impact of AI on talent assessment, forecasting challenges ahead.

For many years now employers have been using tech to drive efficiencies and address bias in their recruitment processes be that through their ATS, online tests, video interviews, the list goes on; all with the aim of both hiring the best talent and having a fair process. AI is impacting more and more areas of our lives, and none more so that when it comes to assessment.

When it comes to using tech based assessment tools to bring efficiencies, especially for volume recruiters, we know for many candidates they have not enjoyed the feeling of being ‘processed’. But have new AI tools, like Chat-GPT, turned the tables on employers with the unimaginable speed at which it can operate and its capability?

Never mind the recruiters, it is the candidates who can use such tools to more efficiently deal with all the assessments employers ask them complete, and to complete them in an impressive and ‘winning’ way.

For some sense of that acceleration for GPT-4, when given large numbers of degree-level exams to complete (such as physics, English language, law, computer science and others), GPT4 scored in the 90^th+ percentile.

In recruitment it is far more capable than most humans. With verbal reasoning texts it scores 99th percentile and numerical reasoning 90^th percentile; with real world assessments such as Situational Judgement tests, GPT-3 in Dec 2022 scored 63/100, and GPT-4 in April 2023 scored 96/100.

GPT-4 can ace application forms, coding tests, one way (recorded) video interviews, the list goes on. And GPT-4 will soon be available to the public and when used with browser plug-ins, candidates can be guided through online assessments with no cut and pasting.

We know from surveys that job hunters attitude to GPT varies from it’s an amazing tool to use, to: if I don’t use it I will be at a disadvantage.

What can employers do in response to this?

The truth is AI is moving so fast, solutions are not yet plentiful. Some early responses cover:

• There are GPT generated text detection algorithms BUT ‘false positives’ have been found, and candidates can evade being found out if they edit the text provided by GPT.
• ‘Proctoring’ i.e. observing candidates undertaking all tests BUT that is both highly invasive (and will lead to greater levels of anxiety and therefore impact on performance); plus it brings risks of higher candidate withdraw rates, as well as being resource heavy.
• Remove all online testing BUT how practical is that?!

What else is currently being done?

There are new solutions coming onto the market, such as AI-powered assessments which focus on how people make their decisions in real time. These new solutions, like strength based assessment tools, seek to get a clear picture of who people really are.

Is another option for lower volume hiring to revert back to two-way video interviews and telephone interviews? Or assessment centres that ‘re-test’ candidates’ strengths and behaviours in real life.

Whatever happens our attitude to AI will be key. And as this technology progresses we cannot lose sight of what we seeking to do when hiring – one aim is wanting our employees to find and use tools which enable them to be more efficient – so them using GPT is an example of this.

As employers our assessment needs to focus on finding ways for us to assess their behaviours such as their decision making abilities, how they interact with people, and give them opportunities to showcase them being used.

So as with so much in innovation, AI brings good and bad. The genie is out of the bottle and employers need to seek ways to embrace this new tech to benefit both them and their people.