Games + AI = incredible hires

Recruitment is a constant pain point for many businesses.

From startups to SME to MNC, the importance of hiring good people cannot be emphasized enough.

Applicant tracking system (ATS) was supposed to help. Instead, typical systems deploy non-semantic matching and basically surface the resume with the best SEO.

LinkedIn helped somewhat by providing the world’s biggest database for companies and agency recruiters to mine.

But with more than 106 million users, it also became a keyword optimisation game. If you are not on first 3 – 5 pages, you are probably invisible.

Even if you get seen, how does one assess the suitability of a candidate newly minted LinkedIn profile?


The flaw of profiling tools

That led to a wave of profiling tools in the market professing to help companies identify the right talents for the roles they are hiring for.

Beyond what a blood test can tell you, these tools are supposed to tell you the personality, behaviours, characteristics, vocational fit and many more.

From DISC to MBTI, from Harrison to Emergenetics, from Performance Indicators to RIASEC, there are probably more than 50 just in a small market like Singapore alone.

But still, how do you ensure the results of these tests would equate to a good hire? On whose benchmark are you depending on to arrive at the decision?


Neuroscience games and bias-free AI

New York-based pymetrics tackle this problem with a different approach.

Founded by two Harvard/MIT PhDs, it applies neuroscience games and data science to predict job performance based on complex neuroscience data.

As a candidate, you begin by creating an account on their website or through their phone app.

There are 12 required games that you have to play and each of them involved some form of reflexes, quick thinking and decision making. Very much like how you would operate when you are playing an actual video game.

The traits that you exemplify from the results of the games would be used to determine your suitability for a list of jobs.

And how would that match come about?

At the company level, their all-star employees would play the same set of games. pymetrics determines which traits equate to high performance for specific roles in the company.

This would form the basis of each company’s unique benchmark.

Because the candidates’ name, gender, race, age or resume aren’t factored in, it also allows companies to eliminate any form of discrimination during their recruiting exercise.


Personal thoughts

pymetrics is the 2nd company I came to know about using games to aid in recruiting.

Scoutible was the 1st one. And beyond just brain teaser-ish kind of games, you actually participate in a role-playing game.

And along with your quest, you would be faced with issues and obstacles. The way you react to them would be telling to the system on the kind of person you are, and how that might fit you into a potential job role.

Scoutible gameplay

Both are relatively new and time would be needed to validate how accurate their games would be in helping companies make the right hire.

But in the face of the current landscape of mainstays (ATS, LinkedIn, job aggregator, agency recruiters), introducing a gaming element into the recruiting process is an interesting concept.

The process alone could generate a lot more applicants and help promote the employer branding of the company.

With a recent US$8million funding (US$17 million in total), pymetrics will be expanding into the city of London and Singapore as it seeks to sign more international clients to add to their stable of Accenture and Unilever.

Please note: I reserve the right to delete comments that are offensive or off-topic.

  • Adrian Koh

    This is over simplification of the assessment and testing process. Internal validation is more important. Gameplay is just the new flavour of the month. Knack has been around for a while. The first question I would ask is if the game play affords an advantage to those who are more interested to such media. Finally, with all selection process, validation occurs when the hired individual actually performs and there are many factors involved (such as teamwork). Does this approach have the research to back it up in this respect?

    • Cliffton

      Hi Adrian Koh, I’m with pymetrics based in Singapore and thank you for your comments. Couple of things here –

      First, the games which we use are adapted from the neuroscience community to measure the behavioural traits these games are meant to measure even in a research context. Our games are deliberately designed to look more academic so that it doesn’t provide advantages to people who are frequent video gamers vs. those who aren’t.

      Second, I agree with you that validation is important, which is why our games are designed to ensure we are capturing the traits accurately, and we do not allow candidates to provide self-reported answers to skew the results in their favour. Validation for any assessment requires tracking over time, and our clients have seen success from their candidates who come through this process.
      I hope this helps to address some of your questions at a high level, and I’ll share an article here to give you a clearer idea of what we do (

      • dave

        As a client of Scoutible (Vs an academic which you seem to be) I can give you my perspective. First, out science team evaluated the things you shared and others. They were very confident that the system was created recognizing and calibrating for skills in gaming. Also, as they had also looked at the other you referenced they felt Scoutible as a true gaming system got people immersed in a way that gave very true answers without the potential for manipulation the saw in other. To your last question it took us a little time to extract results but now that it has been deployed I can say the results stronger than what Scoutible suggested and almost unbelievable to our leadership.