Here's a question - is AI helping or hindering the recruitment and assessment process? LinkedIn is full of tales about how hard it is to get hired because of automated processes. Will AI advances make this situation any better?
To explore this, I spoke with Elaine Pulakos, CEO of workforce assessment provider PDRI by Pearson. PDRI, founded in 1975 and bought by Pearson in 2022, pitches that it uses evidence-based research to provide talent management for both the public and private sectors, covering hiring, assessment, and employee management in organizations.
Pulakos believes there is huge potential for AI, with applications ranging from helping to build test assessments, thru screening resumes, to monitoring job performance. But crucially she is cautious about the role of AI, arguing that it has to be used in conjunction with lots of evidence-based research, and constantly benchmarked against human intelligence alongside the artificial variety:
We are finding that AI is very useful when it's properly employed and tested, and you do the right research and training. It develops items much more efficiently, and quickly. The reality is it's really good when you have that human oversight and training.
That said, she adds:
We're a science-based company, and it takes a lot of time and effort to do the research. I think, in a lot of cases, people are just using AI without that. I believe we should be leveraging the technology, but there is that possibility that it could not be giving you the right answer. Typically you have real impacts on people's lives. I think the responsible companies are mindful of that.
We want to be careful that the measures that we're using to select people are going to work and that they're going to give companies the best people to do the jobs. It is not easy research and takes time. But unless you do that research, you can have a lot of assessments out there where there's no proven relationship between the assessment and how people do on the job."
For its part, PRDI by Pearson uses algorithms and gen AI in recruitment to evaluate responses, although Pulakos stresses that the models do need to be trained properly, as well as having humans in the loop:
We have expert human raters evaluate things, and then we train the AI on those things. We test what the AI is doing against experts to make sure that it's giving us what we think it should. AI can do tremendous things, and offers very powerful tools, but you really need to monitor and evaluate, and make sure that the AI is trained on solid data, and that requires a lot more work than just letting it run amok.
She adds:
If we could be guaranteed that the AI was giving us more accurate and correct evaluations of people in pre-hire and post-hire, it would be a tremendously powerful tool, but the reality is, we're not there yet, and we can't be assured, and it has consequences.
Pulakos cites bias as a good example of how things go wrong. In the pre-hire space, for example where AI algorithms are applied to resumes, Pulakos has seen research showing that they can be fraught with bias, and actually replicate human biases:
You might have certain groups of people who are disadvantaged, or even wonky things, like the person's name is Fred, and somehow Fred's a good name, but Billy isn't. Some of it is mass bias, and some of it's just one-off, wonky stuff.
On the employer side, we worry about how AI was trained. There are biases that are built into the AI that then disadvantage certain types of people. Seeing evidence of that, we get what appears to be legitimate assessments of people, but bias has been built into the algorithms, and come to conclusions about people's skills, capabilities, and their job performance based on that.
Then there is deliberate AI 'cheating'. PDRI by Pearson has done extensive research into whether it could detect if a job applicant was using AI to respond to interview questions. The company posited that if an interviewer probed deeply, following up on specific questions AI wouldn't be able to provide convincing answers. In fact, it did better than expected, admits Pulakos:
It was pretty credible, but we were able, at least today, to detect when people were using AI. The language they used wasn't natural, and there's a bit of a time lapse between when AI hears the question and produces a response. But maybe it's just a matter of time before the tools get so sophisticated we won't be able to tell.
The company is currently developing assessments that incorporate AI, which Pulakos believes will get around cheating, and has developed certain types of assessments that are hard to cheat even with AI. But she thinks that skills-based hiring needs to adapt to the rapidly changing technology landscape, where people need to regularly learn new skills:
When you're hiring, it really doesn't always make sense to hire people based on the skills they possess today, because they're not going to last long. People are going to have to learn new things rapidly. They're going to have to adapt to changing situations, and tolerate stress well. There are personal characteristics and professional skills like communication skills, agility, and interpersonal skills that people need to bring to any job in order to be successful in their careers.
So when we're hiring, we need to be focused on these enduring skills that are going to help companies identify who's going to be the best worker, no matter what happens, no matter what they need to learn, they're going to be agile, smart enough to learn it and be able to cope with it just with their personalities.
To that end, PDRI has developed workstyles assessments to measure these skills which uses a sophisticated algorithm on the back end to score them in real time. Pulakos said that AI helps to find out about hard skills, like certified credentials by scraping people's online activity. But when it comes to soft skills, it is much harder.
She concludes:
When it comes to soft skills that I believe are going to be most important going forward, you just can't do it without some sort of formal assessment of how well you can handle changing situations, and how truly interpersonally skilled you are. That's why we think there's still a place, even in the age of AI, for looking at soft skills in a rigorous kind of way, and AI helps us develop those assessments. To an extent we use it to score the assessments, but without some sort of really good scientific assessment it isn't there yet.
So, is AI helping or hindering? Well ,whilst AI will have a significant impact on the recruitment and retention conundrum currently facing organizations, right now it needs to be handled carefully. Throwing untested AI at recruitment won't get you the best candidates, and using AI to cheat on job applications is likely to be spotted. In other words, proceed with caution.
