As far as I know, there is no software that can overcome the unseen and unquestioned impact of stereotyping
It turns out that I might be wrong as it relates to diversity issues but not for the reasons you might initially imagine. The last few weeks, we have been looking at the impact that machines might have on many different business processes.
In his talk at Dreamforce, prof Andy McAfee noted that Google had switched from asking candidates mind bending puzzle questions to operating hiring methods that focus on behavioral characteristics. In layman's terms, these boil down to two things: does this person have the right skill set for the job profile to hand and, more broadly, will they fit into the business culture.
Behavioral science to the rescue
When viewed in those terms it is possible to factor out stereotypical issues like age, gender and race. Here's how it works with Infor customers who use Talent Science:
In conjunction with behavioral science consultants, Infor assesses the target role's incumbent population using a single online assessment to identify key attributes for success, and then apply that ideal set of characteristics to potential job candidates.
The system reduces diversity problems as a by-product for candidate selection because the assessment is race, gender and age blind. The only variables that matter are those that match candidates to characteristics. Impressed? I am.
It turns out that the people behind Talent Science have been collecting data sets for 15 years, This gives them a very large pool of data against which to setup appropriate skills parameters and then measure results.
According to Jill Strange, Director, HCM Behavioral Science at Infor specializing in this field, hiring staff can infer gender from candidate names but results achieved suggest that gender bias is significantly reduced. More interestingly, Strange reports that Infor is seeing encouraging results for hiring in of African-Americans and Hispanics with race diversity improvement of anywhere between 15 and 50 percent. Strange told me that a key part of understanding value comes in measuring both the before and after positions to assess outcome benefit.
I was curious about the number of data sets Infor has developed and the kinds of role that it can profile for. Strange says it as more than 2,000 job profiles but each of these can be tailored. Apparently, contractors for NASA have used the system to recruit rocket scientists, while in retail, Infor can help stores hire for open positions down to the department level.
I queried how well this works because my retail experience varies from store to store within the same retail chain. Strange said that location is a distinct variable. She cited the strip mall and neighborhood malls as examples where the customer footfall profiles are often very different and that in turn impacts the type of hire for which the retailer optimizes.
I asked about the challenges. Predictive analysis requires large amounts of data and continuous refinement. Strange confirmed something I have long worried about - the lack of baseline data from job incumbents to understand their characteristics plus the performance performance data that goes alongside. Most often, company data is incomplete and/or unreliable. There can be many reasons why data sets are incomplete or dirty. The most common reason is that companies do not know the data needed for pre-assessment purposes.
That data gap should not surprise because even given the state of the art, the ability to understand and find value in the impact of behavioral analysis is relatively new to HR professionals. The fact that technology makes it much easier to marshall data is of little consequence.
I started out thinking that I was looking for a solution to the diversity problem. It turns out that what we should be looking at is the matching of hiring needs, skills and performance. It sounds simple at a high level and Infor points to plenty of success, some of which is little short of spectacular. However, as with so many things technology related, adoption is impacted by the inertia and resistance of people confronted with new ways to improve performance.
It was clear from the conversation that these engagements are as much consultant led as they are technology driven. A sensitive approach is required and so it doesn't surprise to find that many of the people Infor uses on projects are industrial behavioral PhD types. These are the real data scientists who have to traverse not only the statistical and quant evidence but also the fears, hopes and dreams of managers tasked with performance improvement in often highly complex environments.
As we closed out a fascinating and useful discussion, I left one question hanging in the air: predictive analytics in hiring is a rich seam through which to gain valuable insights and develop hiring strategies. But I wonder the extent to which we should be careful not to dampen or destroy the individuality and uniqueness that make us human? I guess that's where man and machine truly meet.
Disclosure: Infor is a premier partner at time of writing