This week in Denver, a team of faculty members, students, and researchers are representing Georgia Tech at SuperComputing 17 (SC17), the annual gathering of high-performance computing (HPC) experts from around the world.
Along with networked data transfer, memory-centric architectures, improving energy efficiency, and the path to exascale computing, one of the key topics of discussion this year is diversifying the HPC workforce.
Lorna Rivera is a research scientist at the Georgia Tech Center for Education Integrating Science, Mathematics and Computing (CEISMC) and part of the Georgia Tech contingency at SC17 this week.
Because her work brings together scientific content, teaching, and equity, Rivera has been invited to participate in a number of panels and workshops on expanding diversity in the HPC community.
What are some tips for increasing the number of female/underrepresented minority (URM) applicants for HPC positions?
First, recruitment is key. You have to start early, not when a position is opening soon. Be sure to review all recruitment materials for inclusive language. There are lots of resources for how to do this. Some of my favorites are from NCWIT, as well as from the WISELI group.
Some other tactics that have been successful are providing opt-in training on unconscious bias for search committees, and offering incentives to employees for recommending or recruiting diverse candidates.
Once identified, pursue all qualified candidates (underrepresented or not) equally without making potentially detrimental assumptions about their interest level in your institution/organization because of their background.
A really useful idea is having candidates meet privately with an advocate that has no bearing on the search committee decision and that complies with Equal Employment Opportunity rules. This person should go over the organization’s policies and practices without inquiring about the applicant’s family/marital status etc. and answer any work-life related questions. This exchange must be entirely confidential and inaccessible to the search committee.
What are some challenges unique to reaching underrepresented applicants?
What’s unique about reaching female and URM applicants is that the challenges are often within the search committee, not coming from the applicants. This is largely due to a lack of sensitivity and awareness, as well as bias against various URM groups. It’s important to rethink criteria like “qualified” or “excellent.” These criteria should be objective and defined prior to the search. It’s when things are unclear from the beginning that we lean toward biased interpretations of what is “qualified” and what represents “excellence”, like an Ivy League graduate or someone who resembles the status quo in visual appearance, demographics, etc.
What are some emerging strategies and/or best practices for promoting diversity and growing diverse participation in HPC?
HPC lacks reliable population data due to its interdisciplinary nature. Unlike traditional science disciplines, it’s hard to put a label on and identify someone in HPC. Right now, lots of the focus within the broader community is on defining and measuring the population in order to generate a baseline for future study. While I too am interested in this and have conducted this type of evaluation for groups like SC, my work extends to identify those that are least well served and ask why in order to improve diversity and equity within a particular context.
Is there anything that Georgia Tech is doing that can be replicated within other institutions?
CEISMC Executive Director Lizanne DeStefano and I work together on many HPC evaluation projects. The findings from these evaluations are being utilized across HPC both domestically and internationally. Our work is quite varied from reducing selection bias in competitive HPC student programs to conducting organizational climate studies of large- and small-scale HPC projects to help improve retention of URM staff/faculty.
What advice do you have for HPC leaders working to build diverse teams?
If possible, partner with or hire someone with demonstrated success in the area to help implement strategies for your group/organization. Culture change takes commitment and is very difficult to achieve doing it as a side job. Proposal requests (RFPs) often include a broadening participation component, but many projects are not adequately prepared. The HPC world is lucky to have a resource like XSEDE that has a strong broadening participation program designed to serve the entire community through partnerships and direct programming.
Despite facing the greatest obstacles and receiving the least support of underrepresented groups, women of color are often overlooked when discussing diversification. How can companies fight against the current trend of taking a one-size-fits-all approach to advancing women and women of color?
As a Puerto Rican woman, this really resonates with me. There is definitely no good one-size-fits-all approach, and companies are likely to miss out on the benefits of diversity if they fail to acknowledge that reaching diverse groups requires diverse strategies.
Developing a strategic plan to address key target populations in a stepped approach is one way of tackling this problem. Groups should try to pilot evidence-based programs and activities with each target population in isolation, and then adapt their programs for other groups. All activities should be closely monitored in order to be responsive to any successes or slippage that may occur.
Documenting these outcomes for the community is also incredibly helpful. We are in the middle of doing this now with an international HPC student-training program. The planning committee wanted to increase the number of female participants in the competitive program, but they were not being rated highly. Our investigation found gender-based selection bias and we made recommendations for reducing it, including an overhaul of the application form in 2016. That year no evidence of gender-based selection bias was found.
Workplace diversity efforts are ongoing, but progress remains slow. How can we increase the urgency and speed up the pace of these initiatives?
Organizations like Georgia Tech are setting a good example by modeling and piloting initiatives so that others can adapt based on Tech's lessons learned and hopefully reap substantial benefits. I personally would not be able to attend SC this year if it weren’t for the childcare grant offered by the College of Sciences. Because I’m able to attend, I will be giving more than 10 presentations this week (some closed/proprietary) on evaluation findings in HPC projects – all of which have an equity and diversity component. Hopefully, this work will add a few drops in the bucket to addressing this massive challenge of moving forward on these much-needed programs.