Third Annual EEODataNet Conference

The third annual EEODataNet conference was a resounding success! The conference met June 16 and 17 at the EEOC headquarters with academic and EEOC researchers, EEOC systemic investigators, as well as private bar lawyers and corporate representatives in attendance. You can see the program here, and the slides of the presentations will be posted to the EEODataNet website soon.

The conference kicked off with a panel on the new pay data initiative. Fidan Kurtulus (UMass) introduced the history of the pay data collection initiative. Victoria Budson (Harvard) outlined a successful voluntary initiative in Boston to compare gender pay gaps across major employers and reduce them. Paul Van Hipple (Texas-Austin) helped the EEOC and the rest of us understand how to analyze pay band data. The session ended with a plea for everyone in the EEODataNet world to respond to an upcoming comment period on pay data collection. We will send you a notice when this comment period opens later this summer.

A panel on “What works?” To foster diversity in organizations was well received by EEOC investigators, despite Frank Dobbin (Harvard) casting doubt on Diversity Training as an effective strategy. Sheryl Skaggs (Texas) pointed out the problem of backlash associated with legal interventions. Alexandra Kalev (Tel Aviv) reported on research showing that active screening for drug use and criminal background checks might actually improve the hiring chances of minorities, who would otherwise have fallen victim to stereotype bias. Don Tomaskovic-Devey (UMass) argued that no one policy works, but policy effectiveness happens when there are external or leader pressures who empower internal constituencies more than they produce backlash from other employees.

In a panel on policing Phil Cohen (Maryland) showed that minority officer employment dropped relative to population baselines as the percent minority in community grew. This result was all the more disturbing when Justin McCrary (Berkeley) reported that police shootings of citizens were lower in police forces that hired more minority officers.

There were two panels on charge data. Lee Badgett (UMass), Reggie Byron (Southwestern), and Sarah von Schrader (Cornell) described their projects on the use of charge date to understand LGBT, race/gender/age, and disability discrimination. This was followed by a panel discussing the design of a standardized charge data set. A working group is being formed to recommend the structure of that data set.

One of the highlights of the conference included two panels discussing the EEO consequences of machine learning algorithms for employee selection. Presentations from IBM, Microsoft, academic computer scientists, and employment lawyers, including EEODataNet member Pauline Kim (Washington University) left the distinct impression that nobody really knows the implications. Vendors claims that their algorithms eliminate cognitive bias need to be weighed against the very real possibility that machine learning reproduces or even amplifies institutional bias in past selections. EEODataNet and EEOC researchers are planning a research design to find out what is under the hood!

Overall, the conference was remarkable in the degree to which the academic and regulatory communities are learning from each other.