Programmers Ahoy!

Robin Gerrow

SIOP 2019 Machine Learning Competition Registration Opens

There are lots of ways data are being used in business processes, but one of the most exciting is machine learning. SIOP will be bringing the challenge of machine learning front and center for this year’s annual conference with an open competition to test the ability to work with unstructured data.

Registration is open now for the free 2nd Annual SIOP Machine Learning Competition and will conclude with a presentation by the winners at  the 2019 Annual Conference in National Harbor, Maryland just outside of Washington, DC, April 4-6, 2019. Participants may enter at any time, and winners will be announced by March 21.

This year’s organizers are all from last year’s winning teams. Isaac Thompson and Nick Koenig are both data scientists at Shaker, and Mengqiao (MQ) Liu is a research scientist at Amazon.

“With diversity and inclusion as our true north, we are excited to open this year’s competition to the entire community,” Thompson said. “Anyone can participate. We set up a website to automate the entire submission process and provide a live public leaderboard. The data was gathered specifically for this competition and after it ends the code and the data will be available online. Anyone in the field interested in natural language processing will be able to download the data, download the code, and have some of the most powerful natural language processing tools available at their fingertips for free. This competition will not only propel the field forward on the best machine learning methods available it will enable anyone to use them for free.”

Artificial Intelligence and Machine Learning together have been named by SIOP as the number one workplace trend for 2019.  Both will allow I-O psychologists to more efficiently analyze and find appropriate uses for the vast amount of data becoming available to them.

“Machine learning invokes a lot of misconception,” Thompson said. “When I think of machine learning I think of optimizing a system—usually a human system in I-O psychology—with the use of data for an outcome of interest. In one of its most simple manifestations, machine learning is nothing more than a linear regression, which we all learned about in school. More advanced types of machine learning can discover non-linear relationships, be optimized for multiple outcomes of interest at once, create their own representations of reality from raw inputs, and even learn to improve themselves. These advanced types are what prompted the SIOP machine learning competition.” 

This year the organizers are taking a different path with the competition in emphasizing the open source approach.

“Science should be transparent, it should be inclusive,” Thompson said. “Everyone should have the opportunity, and if they win, they’ll be out in front of the top researchers, top academics and top consulting firms in our field. They’ll be able to make a name for themselves and that could lead to very lucrative job offers.

“Some of the more advanced machine learning techniques, like deep learning, are so new, that they only started gaining traction around 2015,” he added. “While huge strides have been made, I-O psychology is at an inception point. Because it is so new, more resources are available online in open source communities than traditional classrooms. For example, last year Nick Koenig leveraged his open source programming foundation to self-learn cutting-edge algorithms that applied to the competition and was able, with those new algorithms, to secure last year’s victory. This year’s competition is all about learning something new. We are hosting a website where submissions can be uploaded for with a live leaderboard—instant feedback—which will include team names. Since only team names will be shown, not individuals, there is no risk in participating, only upside. Of course, winners’ actual names will be revealed.”

Registered participants will receive a large data file of five open-ended situational judgment items and five aggregate personality trait scores. The goal of the competition will be to predict respondents' personalities from the open-ended text. After the competition all of the data and the code will be open-sourced on GitHub and winning teams will be invited to present at SIOP.

Anyone is eligible to participate, and registration for the competition is available online.