Thursday 28 September 2017

Selling Ice to an Eskimo


There's an interesting new company called Pymetrics that aims to bypass the conventional hiring process and try and find character traits that high skilled employees display in order to help employers narrow down potential candidates. The idea is that you can play a series of mini games that don't have win / loss metrics, but the way that you play those games will show what kind of traits you have. The company will have the stats of their current employees' results, and the company will be presented with potential candidates who have traits that match their current high-performing employees.

They claim that the process should reduce personal biases based on things like race, pedigree (i.e. which university you studied at, if you have a degree at all). The algorithm isn't just a perfect match based on similarities, but they say there's some machine learning behind it all, so as I discussed in earlier posts on AI, sometimes it's hard to work out why one candidate was preferred over another.

I thought I'd give it a shot. I couldn't play the games on my phone without downloading their app, so I had to wait until I got home - because not only is playing games on your work computer frowned upon at the office, it's probably even worse if you're playing a game designed to help you find a job!

After each game you play, you get a result screen telling you what trait you displayed while playing that game.






I won't go into all the games, in case you want to play it for yourself - since knowledge of how the game measures things may bias your result. I found my results interesting though. I have always considered myself incredibly risk averse. Yet according to this game, I have a risk preference for high-risks. It's probably because of that that my recommended jobs were sales, consulting and hedge funds.


My top traits seem to be preference for ambiguous risks, staying focused on one strategy, and putting in effort for low reward.

I spoke about this with Michael, who was the one who told me about this test. He did it as well, and I find it fascinating how different our results were. He got the results I wish I had gotten. :(

I feel like we are quite similar in a lot of ways, we both tend to have the same moral beliefs, and similar interests, but I think we are a case of convergent evolution - we both have quite different backgrounds, but have converged on similar values. We talked about our approaches to the various games, and one thing that struck me was that while I was playing them, I had a definition of "winning". There were some tasks where the "win condition" seemed obvious, and so I set about trying to achieve that. When looking at the results I got, I looked at some of the traits and figured the reason I got that trait was because I didn't do very well on a particular task.

Hearing Michael explain his thought process when he was doing those same tasks made me realise that my win conditions weren't the only ones. So it's likely that it doesn't just measure traits directly, but perhaps it tries to guess at the motivation for why someone may play the way that they do.

At first, I wanted to completely write off the results. I mean, come on, I'm one of the most risk averse people that I know. However, the more I think about it, the more I'm starting to wonder how much merit there is in the results. I'm not the world's greatest salesperson, and I don't think any amount of training will ever get me there, but I am pretty good at customer service, having worked in retail and hospitality for a long time. I always thought that I would be a good diplomat, since I'm a natural peacekeeper.

Plus, if I've managed to convince you to give it a shot, then maybe I'm not the worst sales person. Note: I didn't have to tell them my age, gender or ethnicity (and if they live by their words, those things shouldn't matter anyway). I will admit that I did lie and say that I'm unemployed, as I wasn't comfortable telling them where I worked, so I won't judge you if you choose to do that, too. Though I do feel a bit bad, as I'm worried I may bias their AI.

No comments: