Diversity and inclusivity is important to SC16. The key components in our effort to diversify the HPC community are providing opportunities to underrepresented groups as well as educating the community on the benefits of improving diversity. The SC conference series gives us a great platform to spread this message.
This year, we are pleased to offer the following:
- Breakdown of the conference committee demographics and also the demographics of attendees (see Graphs section below).
- SC16 Code of Conduct
- Child Care: Providing childcare and a parents’ room for an experience that better accommodates the needs of parents.
- First Time Attendees: Hosting an Introduction to SC16 for First Time Attendees session on Monday evening, November 14, 2016, to help those new to SC have an opportunity to meet new people and better understand what the SC16 event has to offer.
- The Diversity team is evaluating the conference and looking for opportunities to improve the attendee experience particularly for members of underrepresented groups
Did you know?
- In 2015, 14% of attendees were women – our goal is increase this to 20% by 2020.
- Supercomputing is publishing our demographics. In addition, other leading HPC organizations around the world are also doing workforce demographics. By working together to publicly document today’s workforce, we can begin to understand what is needed to properly diversify the workforce and create a more inclusive working environment. These contributions are a positive step in creating a more diverse group of attendees, presenters, authors, and exhibitors in the years to come.
Still, there are other contributions that can be made to increase these efforts, and they start with you. Start a conversation within your organization about how you can become part of the effort, publish your data, and share your results. We cannot get to where we are going if we do not know from where we are starting.
Summary for Entire Committee
Breakout by Committee Type
Summary of Entire Committee by Geo
Data as of 10/28/2016; includes 654 unique committee members.
SC15 attendees by gender and class of registration:
Because we have incomplete data on the gender of participants, we are predicting the gender of attendees based upon first names. This prediction relies upon the condition that many names are associated more frequently with one gender than another (notably this is not generally true with Chinese and Korean names when rendered using a Western alphabet). While it is certainly possible to come up with Western names that do not have a strong association with a single gender, over a large population of names we expect to be able to make at least general observations about gender proportions based on name.
Because SC registration in 2015 included the optional question, "Do you identify as male or female" with the three responses "male", "female", and "do not wish to provide an answer", we are able to do some evaluation of the reliability of our approach for predicting genders based only on first name.
- 5283 Attendees chose to specify a gender when registering (of 12857 total attendees), or 41.1%; 248 of those who answered the question (2% of all registrants) indicated they did not wish to provide an answer.
- 58.9% of attendees chose to skip the question entirely.
- Of the attendees who answered and choose to specify a gender, 13% self identified as female and 87% self identified as male
- Our methodology correctly predicted 83% of the females, misclassifying 13% of them as male and unable to classify the remaining 4%
- Our methodology correctly predicted 94% of the males, misclassifying 2% of them as female and unable to classify the remaining 4%
- In interpreting prediction misclassifications it is important to remember that the comparison data — self reported gender — may itself contain errors. Although the error rate in this data is expected to be low, the actual extent of misreporting here (whether deliberate or accidental), or whether it even exists, is impossible to determine from the data provided.
The accuracy rates relative to self-reported gender indicate that we may be under-reporting the proportion of females by about 2% overall (a correction for this would predict the total percentage of female attendees is closer to 18%).