Using a Feminist and Inclusive Approach for Gender Identification in Film Research

Although there is a scientific consensus that gender is not binary, immutable, and physiological, it is still common to operationalize it in such a way. This can be said of both traditional as well as new sources of digitized or digitally born data. Not long ago, population surveys were treating sex and gender interchangeably requiring interviewers to ask a question about “sex/gender” only if it was not obvious from person’s voice (Bradburn et al., 2004). Big data and Internet sources imposed a challenge of large amounts of messy data, where gender is often operationalized and coded with the help of automatic detection methods (e.g., based on names or face recognition). Recently, there have been more attempts to critically assess and change these exclusive practices in both traditional (Bauer et al., 2017) and digital methods (Keyes, 2018). This contribution joins these efforts by describing our attempt to measure gender of film directors by relying on their own chosen self-representation. We also compare our results to alternative findings when binary manual and automatic gender detection methods are used. In communicating the comparisons, we visualize the data to invite others to critically reflect on current practices of gender operationalization.
The research we report on in this poster stems from a study that explores a substantive question of temporal and spatial patterns of festival runs (movement of films through the films festival networks) from a film studies perspective (Loist 2011; Loist 2016). The aim of the project is to explore festival runs of 1727 films selected from a sample of six relevant international film festivals starting from their premiere and ending with their commercial release or lack thereof. Gender of directors constitutes an important piece of information due to known discriminatory practices in film industry (Aylett, 2016). Our data sample collected from six festival programs included 1727 international director teams (that corresponds to 1899 directors). In our operationalization of gender, we focused on directors’ ways to use personal pronouns on available web resources (e.g., personal websites, Wikipedia pages). The web search was implemented manually by project assistants. If a person used male or female pronouns, a female or male gender was assigned accordingly. However, if a person used certain cues to indicate a self-identification with non-binary gender, a non-binary category was assigned. We report data on used cues, missing data, as well as challenges and limitations of the implemented web search (e.g., limited data on the authorship of the web resources). In addition, we compare the findings to manual non-consensual assignment of gender based on names and photos as well as automatic detection methods commonly used to assign gender in large databases. For the latter, we applied two widely used tools: Sexmachine implemented in Python (Elmas, 2013) and Genderize implemented in R (Wais, 2016). While Sexmachine primarily relies on name and country data, Genderize is a commercial application that also uses social media data (Karimi et al., 2016).
Results are communicated via visualizations guided by a feminist approach to information visualization and practice (D’Ignazio and Klein, 2015). This approach stresses the importance of examining the entire life cycle of the project against such principles as challenging binary categories, transparent communication of project’s decision making, embracing horizontal systems of knowledge transmission, embedding data into a context, as well as visibility of labor. We conclude by discussing possible non-binary gender measurements at scale as well as their ethical and data privacy challenges.

Bibliography

Aylett, H. (2016).
Where Are All the Women Directors? Report on Gender Equality for Directors in the European Film Industry.
EWA Women Directors in Film – Comparative Report.
https://www.ewawomen.com/wp-content/uploads/2018/09/Complete-report_compressed.pdf (accessed 1 November 2018)
.

Bauer, G. R., Braimoh, J., Scheim, A. I., and Dharma, C. (2017). Transgender-inclusive measures of sex/gender for population surveys: Mixed-methods evaluation and recommendations.
PLOS ONE, 12: e0178043.

Bradburn, N., Sudman, S., and Wansink, B. (2004).
Asking questions: The Definitive Guide to Questionnaire Design - For Market Research, Political Polls, and Social and Health Questionnaires. San Francisco (CA): Jossey-Bass.

D’Ignazio, C., and Klein, L. F. (2016). Feminist data visualization. In
IEEE Workshop on Visualization for the Digital Humanities (VIS4DH).
Baltimore: IEEE.
http://www.kanarinka.com/wp-content/uploads/2015/07/IEEE_Feminist_Data_Visualization.pdf (accessed 1 September 2018).

Elmas, F. (2013).
Sexmachine [Python package].
https://pypi.org/project/SexMachine/ (accessed 1 October 2018).

Karimi, F., Wagner, C., Lemmerich, F., Jadidi, M., and Strohmaier, M. (2016). Inferring gender from names on the web: a comparative evaluation of gender detection methods. In
WWW ’16 Companion, Proceedings of the 25th International Conference Companion on World Wide Web. Republic and Canton of Geneva, Switzerland: International World Wide Web Conferences Steering Committee, pp. 53–54.

Keyes, O. (2018). The misgendering machines: trans/HCI implications of automatic gender recognition.
Proceedings of the ACM on Human-Computer Interaction, 2: 88:1
‐88:22.

Loist, S. (2011). On the relationships between film festivals and industry. In Lee, Y. (eds),
Busan Cinema Forum: Seeking the Path of Asian Cinema: East Asia. BIFF - Busan International Film Festival, pp. 381–402.

Loist, S. (2016). The film festival circuit: networks, hierarchies, and circulation. In de Valck, M., Kredell, B., and Loist, S. (eds),
Film festivals: History, Theory, Method, Practice. London: Routledge, pp. 49–64.

Wais, K. (2016). Gender Prediction Methods Based on First Names with genderizeR. The R Journal, 8(1): 17-37.
https://journal.r-project.org/archive/2016/RJ-2016-002/RJ-2016-002.pdf (accessed 1 October 2018).