Skip to content
Switch branches/tags

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time

NeurIPS 2018 Demographics and Inclusion Survey

This repository hosts the NeurIPS 2018 Demographics and Inclusion Survey Summary of Responses.

About this document

The most recent version of this document is from 6 November 2018, based on responses to the survey through 13 October 2018. The survey was designed, analyzed and written up by Hal Daumé III and Katherine Heller with help from William Agnew, Samy Bengio, Timnit Gebru, Shakir Mohamed, Alice Oh, and Hanna Wallach. More information on NeurIPS 2018 diversity and inclusion intitiatives can be found here.

Executive summary

We report the results of a survey conducted from August–October 2018 on demographics & inclusion in the NeurIPS community. At analysis, 2375 people participated; the range of responses is vast. Here, we attempt to capture the key themes, with pointers to where more information can be found. Such a summary runs the risk of ignoring concerns of some members; we encourage all interested to read the full report. The below concerns are listed arbitrarily; there is no implied priority. At the NeurIPS 2018 conference, during the lunch period on Tuesday, there will be a moderated and guided townhall; one goal is to develop action items to improve the level of respect and inclusion at the conference. Thank you to all participants.

Representation, Respect and Awareness of Others: The environment at the conference is one in which many have experienced harassment, bullying, microaggressions, or lack of respect as a result of: their gender (2.3.a, 2.3.b, 2.7.l); their sexual orientation or gender identity (2.3.b); their race, ethnicity or religion (2.4.b, 2.7.h); their native language (2.4.a, 2.4.b); their political views (in particular, an assumption that everyone is left-leaning: 2.7.i); or their disability status (2.5). Experiences of sexual harassment in particular, and a lack of acknowledgment of these problems by the organizers (and therefore a perception that the conference condones such behavior), is a recurring theme (2.3.a, 2.7.m). These problems are exacerbated by the underrepresentation of women (2.3.c), queer (2.3.a, 2.3.c, 2.7.k), and minority (2.4.a) attendees; this occurs both in attendance and in organizational structure (2.4.a). Some specific issues that were raised with the goal of increasing awareness of needs of others: generally improved respectfulness for all attendees (2.7.k), suggestions for how to make talks more accessible to colorblind attendees (2.5.b), or attendees from non-English-speaking countries (2.4.d).

Community Openness: The conference can have a closed and elitist feeling to newcomers, especially those from labs or jobs perceived as non-standard within NeurIPS (2.2.b), and those who work in different areas of science or work on non-trendy topics (2.2.e, 2.7.e). There is a perceived bias toward researchers and labs in the U.S. in the conference and in reviewing (2.4.c), particularly when authors are non-native English speakers (2.4.a). The lack of openness makes it difficult for newcomers and junior researchers to network (2.7.a, 2.7.g), a problem which is exacerbated by non-inclusive corporate events (2.7.b), and by a lack of transparency regarding how the conference runs and is organized (2.7.m).

Conference Logistics: Specific logistical choices around how the conference is run have led to difficulties in participating fully for many. Loud, crowded poster sessions create challenges for those who have difficulty standing for long periods, and those with claustrophobia or hearing loss (2.5.a). Additional issues with evening corporate parties include substantial difficulties for newcomers, those from outside the US, those who do not consume alcohol, in addition to harassment that has occurred there (2.2.c, 2.7.b). Catered food arose as an issue for people, with health reasons (2.5.c), and religious or personal reasons (2.7.f) for maintaining a particular diet. Other religious issues include difficulties finding places to pray (2.7.h). There are several issues parents face, such as a need for broad childcare support (2.6.b), challenges with evening events like posters and parties (2.6.c), and the overall choice of the dates of the conference (2.7.f); the last of which also impacts educators (2.2.f). Registration was a large problem this year, which impacted people based on where they live and their time zone (2.7.d), and their economic status and liquidity (2.7.d).

Cost, Location and Travel: Attending the conference is expensive, which especially impacts students and people with particular jobs (2.2.a), from different parts of the world (2.4.a), and those who are low-income or do not have significant liquid assets (2.7.c). In addition to cost, travel is also made difficult by visa & immigration issues (2.2.d, 2.4.b), which in some cases target specific ethnic groups (2.4.c). Parents, especially those who have to travel long distances, face difficult decisions about attending due to both cost and time (2.6.a). All problems are exacerbated as locations selected for the conference remain low in number.

Transparency, Communication, and Inclusion Activities: The can conference improve it’s communication with the community, and the transparency of it’s processes. This includes having a formal, transparent structure for how one progresses from author to reviewer to AC to board (to avoid information asymmetry between “in-group” and “out-group”), and having explicit governance documents and elections (2.7.m). Communication can be improved around the Code of Conduct, it’s precise meaning, and its enforcement (2.7.l). There was a desire for the organization to work with the media to downplay AI hype (2.7.m). Regarding inclusion efforts, there was a disbelief that the conference would do anything substantial, and a desire to aim for equality of opportunity (2.7.i); to avoid identity politics (2.7.i); to see a public, explicit statement from the conference about inclusion, and to acknowledge past problems (2.7.m); varied opinions about affinity group events and a concern that inclusion efforts are going too far (2.7.j); and to have open discussion and offer training sessions on how attendees can create an inclusive atmosphere (2.7.j).


Results of the NeurIPS 2018 Demographics & Inclusion Survey







No releases published


No packages published