New research on diversity at AI conferences

The research analyses the development of diversity indicators from 2007 to 2023 and is accompanied by a dashboard where the data can be explored further. 

The paper focuses on seven core AI conferences, and how indicators for gender, geographical location and organisation type has changed over time. Based on the papers published in the conference proceedings, it shows a slight upwards trend in the overall diversity index. There are however differences between the conferences, with NeurIPS and ICML seeming to be the less diverse ones.

Specifically on gender, the proportion of female authors has increased over time, which may relate to the growing number of initiatives to foster inclusion in AI communities, particularly led by women and other minorities in the field. A selection of such initiatives is included in in the paper, the earliest being WiML (Women in Machine Learning), established in the same year as the paper’s analysis begins, 2007.

For the full analysis, read the paper here, and explore the data in our new dashboard.