FindHer: a Filter to Find Women Experts

FindHer: a Filter to Find Women Experts Gabriela Ferraro CSIRO Data61, Canberra, Australia Australian National University, Australia gabriela.ferraro@data61.csiro.au Zoe Piper CSIRO, Canberra, Australia zoe.piper@csiro.au Rebecca Hinton CSIRO, Canberra, Australia rebecca.hinton@csiro.au Abstract Women are underrepresented in many spheres of our societies, including research. A com- mon excuse for exclusively male line-ups is that suitable women could not be found. One way of promoting visibility of women in in- dustry and academia is to explicitly provide solutions to find them. Expert Connect is a publicly searchable database of Australia’s re- searchers that now includes FindHer, a filter to find women experts in any field of research. In this industry paper, we evaluate Natural Lan- guage Processing and Computer Vision tech- nologies for gender determination within the aim of automating gender profile tagging for FindHer. We found current off-the-shelf tools are highly effective in detecting gender from names and photos. Nevertheless, a human- in-the-loop approach should be preferred to a fully automatic one, since ethical concerns might arise. 1 Introduction Women representation remains critically low in a range of fields, including the research sector (Lar- ivire et al., 2013; West et al., 2013; Mihaljevic- Brandt et al., 2016; Bonham and Stefan, 2017). A common excuse for exclusively male line-ups is that suitable women could not be found. This has led to a proliferation of initiatives encourag- ing women to list their details in various skills and expertise related directories. These directories are then promoted to conference organisers, company boards, investor groups, the media and beyond. These directories do promote women and im- prove their chance of discovery; they also put the onus on women to create and update their pro- files across multiple platforms. This can be time- consuming and repetitive. What is more, it may or may not result in extra opportunities since of- ten the only people looking at these directories are people who already know they want to engage a woman. We can be smarter about how we man- age the public data that already exists. We want to meet the challenge of someone saying that “a suitable woman could not be found”. In short, we need to put gender-based information in places where people are already looking. Expert Connect1 is a publicly searchable database of Australian research expertise designed to boost industry-researcher collaboration. Since International Women’s Day in 2019, Expert Con- nect can now be filtered to find women experts in any field of research using FindHer.2 Currently, only 15 percent of the Expert Connect data is in- cluded in the filter (with over 4,500 women pro- filed). It’s not a perfect process, but we are contin- ually working on improvements with the number of women profiled continuing to grow. Work like this raises ethical considerations. In Section 3.1, we discuss some of the issues we considered in the process of submitting our ethics approval for this work. We present experiments on automatic gender determination using off-the- shelf Natural Language Processing (NLP) and Computer Vision (CV) technologies. Our aim …

Content truncated to 500 words