The MICRO Diversity Survey

January 23, 2018

The MICRO steering committee recently published a survey on diversity at the conference. This survey is the latest development in a renewed conversation among computer architects that began with a September 2017 blog post by Natalie Enright Jerger and Kim Hazelwood that highlighted unsettling statistics on gender diversity in the community. If you count yourself as part of this community, please take the survey.

Here are some of my responses to this survey.

Please comment on your thoughts and reactions to the blog post.

I knew already from first-hand experience that architecture has worse diversity problems than some other areas in CS, but the level of gender disparity depicted in the blog post still surprised me. Here are some especially shocking statistics: that only ~10% of authors in major conferences are women; that there was no female MICRO PC chair from 1991 (when I was four years old) to 2017; that no keynote speakers at ISCA have been women in the last five years.

In other areas of my academic life, such as the internal department culture at my university, I’m accustomed to public, concerted efforts to make sure that women and underrepresented minorities are well represented in important positions. This is hard work and people—even well-meaning people—often fail. The data from this blog post suggests that the architecture community is failing more often than it’s succeeding. It convinced me that the (again, usually well-meaning) leadership hadn’t made the issue a priority. Clearly, that needs to change.

Please share your thoughts in general on gender diversity and inclusion in computer architecture.

I wrote above that the main conclusion stemming from this blog post is the reminder that well-intentioned people in our community were not trying hard enough to prioritize diversity above other issues. That is certainly the main problem. It’s the same problem that made me and others bristle at the all-men “Legends of MICRO” panel last year and other embarrassing reminders of our collective failure to prioritize a serious problem.

But a secondary problem that shocked me during this period was the set of online disagreements from people in our community who seemed not to be in the same category of well-intentioned but negligent leaders. Instead, these disagreements suggested that some people don’t believe there is a problem. The most prominent example was a MICRO steering committee member who posited in a series of comments that the data was not worth being concerned about.

Perhaps because of my nature, I still want to believe that he and others who expressed similar views want to be on the right side of this issue! But comments like the one above revealed more than just not trying hard enough—they betray that people in our community are behind the times on the national conversation about representation and diversity, in CS and in society as a whole. They demonstrate that we have a long way to go in convincing people in our community even of the basic tenets of how to make progress on diversity and inclusion.

What topics do you think should be included in such a workshop? [The question is about a hypothetical co-located workshop on diversity topics. —ed.]

I’m not sure whether this is a good idea because I’m not sure whether it would be productive. At this point, I feel like the right next steps are fairly clear—it’s hard for me to come up with workshop elements (talks, panels, debates, etc.) that would help more than action at the executive level. Or, to use an aphorism: perhaps we’ve done enough talking, and now we just need action.

One format that I can imagine being useful would be to invite outside experts on implicit bias and structural inequity to run trainings for our community. These would not be specific to computer science, but given the level of expertise on display in our community, it is probably a better place to start than having computer architects give talks to each other.

What should be the goal of such a mentorship program? [The question is about a hypothetical program where senior members of the community volunteer to mentor newer researchers. —ed.]

We should basically steal the programming languages community’s model! The Programming Languages Mentoring Workshop (PLMW) is an NSF-funded, incredibly successful series of workshops that takes place at every PL conference. It has a diversity mission, but it also functions as a venue for mentoring even students who are not underrepresented. Crucially, it provides funding to attend conferences.

Do you have any other ideas that you think would be helpful for enabling more diversity and participation of women and underrepresented minorities at MICRO? What do you think, as MICRO, we should do to improve diversity?

This survey didn’t touch on the related issue of the opacity of MICRO organization. I’m relatively new to the community, which means I don’t know the (apparently long and complicated) history here. But I’ve always been mystified by the different way that MICRO is organized. Other conferences I attend—in architecture and programming languages—have clear rules for how elected, apparently accountable organizations oversee the SC and the conferences’ processes. For example, the PL community has started publishing long, detailed policies about how to run each conference—here’s the one for PLDI.

MICRO, in contrast, always seems to be run by the same people. At every business meeting, the same set of people seem to have the same set of arguments. It gives me—again, as a very junior professor—the impression that an “old boys’ club” runs the conference, which in turn makes it hard to see how to participate meaningfully in the way the conference works.

I bring this up not to complain about my own vague sense of exclusion (who cares), but to make the point that this kind of management structure works against efforts on diversity and inclusion. Because our community is currently so imbalanced, it’s obvious that the only way to improve it will be to include new people—mostly young researchers, i.e., PhD students. Any structure that appears to work mainly on seniority and connections will systematically exclude new people—and, thereby, disproportionately affect women and underrepresented minorities.

By making the election and leadership policies for MICRO more clearly participatory, we stand a chance at making our community more inviting for the categories of people we’re trying to attract. Sticking with the status quo, however, will continue to undermine well-intentioned efforts on this front.