The Present and Future of Talent Analytics as Seen from #LEC2016

Over the weekend, I had the great fortune to be in Atlanta with SIOP’s Annual Leading Edge Consortium, a practitioner forum for I/O psychologists and related disciplines. This year the theme was Talent Analytics. The speakers covered a lot of ground, sharing their work on topics such as predictive analytics, big data, and data visualization ("data viz") among others.

As is the sign of any good conference, I flew home with a lot of ideas swirling around. After 48 hours to let it all soak in, here are a few of the highlights that will stick with me.

Data science will revolutionize the field of Talent Analytics over the next decade. And actually, it already has. It is both exciting -- and depressing -- to learn from Dan Putka of HumRRO how the traditional methods we learned in I/O for developing predictive models are but a tiny fraction of the methods used by data scientists today. According to Dr. Putka’s research using biodata to predict job performance, the modern data science methods out-predict the traditional ones (e.g., OLS regression) by the biggest margin as the number of predictors is increased (e.g., analyzing items as opposed to scales). In other words, data science wins with complex models. This means that if I/Os want compete inside the ‘quant’ arena in the long haul, we will not only need to learn more sophisticated modeling techniques, but we will also need to experiment outside of our training and assumptions about traditional scale construction and psychometric theory.

The sophistication of machine learning will quickly surpass our ability to understand it. In fact, it already is. Ben Taylor from HireVue described their work using deep machine learning techniques to optimize the predictive validity of job interviews. Their software encodes millions of data points from text, audio, and video and translates it into computational models that are so complex, they press the boundaries of human capacity to unpack and understand the algorithms. Sound like Battlestar Galactica? See how Ben’s team measured the emotional output of Donald and Hillary in the recent presidential debates. *On a side note, I also learned that Ben is a cool guy and enjoys participating in the existential crisis discussions his work helps to create :) I’m 98% sure he is not one of the Cylons.

Data translation is a science too (and a skill set). If data scientists from applied statistics and programming will eventually replace I/Os as the main “quants” in HR, the new positioning to consider might be the space between data and action. For this role, data visualization and storytelling skills will be in high demand. Evan Sinar from DDI shared several best practices and watch-outs when it comes to data visualization and dash-boarding, as well as several practical tools and resources you should check out and try out for free (e.g., See: http://raw.densitydesign.org/ and www.tiny.cc/dvresources).

Sociometric data will greatly enhance our understanding of team dynamics and emergent phenomena. Steve Kozlowski from Michigan State University is partnering with NASA and several other groups to develop and study sociometric ‘badge’ technology which captures in real-time a full range of data to describe complex human interactions within teams. Eventually the technology will have the capability to translate the data back to the teammates, also in real-time, to support new forms of in-process feedback and self-regulation. In addition to unlocking new insights for the science of performance in teams, Dr. Kozlowski hopes these sociometric innovations will support the first manned mission to Mars. So all in all, no big deal :)

A few thoughts in closing

I thoroughly enjoyed the full spectrum of Talent Analytics on display at #LEC2016. I would describe it as reaching all the way from the here-and-now to science fiction. And though it would be healthy to exercise skepticism about the more far reaching ideas that were presented, we could easily fall behind and struggle to catch back up. Nathan Mondragon from HireVue summarized it well when recounting his experience 20 years ago in front of a room of I/O psychologists who were balking at the notion of online assessment. Here we are 20 years later, and it appears to me that at least three or four new trains are now pulling out of the station... with or without us.

The other reflection I'm left with has to do with how all of this will link up with action. In other words, the bottom line for us data geeks -- and particularly for folks like me who work in the OD-Change Management-Coaching space -- is whether we can translate new sophistication on the data side into new interventions that are more effective than the old ones. A few of the presentations approached this topic, but still did not make it all the way to action... actionable insights, yes... the action itself, no. It will be interesting over the next decade to see whether I/O reorients in this direction or further embraces its data geek. Maybe next year's LEC theme is an indication: Coaching. Look forward to seeing you there.