Gender Parity, Gender Equality, and Intersectionality : Public Perceptions of a ‘50:50’ Workforce

(1)

1

Gender Parity, Gender Equality, and

Intersectionality

Public Perceptions of a ‘50:50’ Workforce

Kara De Kretser

Supervisor’s name: Stina Backman, Gender Studies, LiU

Master’s Programme

Gender Studies – Intersectionality and Change Master’s thesis 15 ECTS credits

(2)

2 Abstract

Gender parity. Gender equality. Diversity and intersectionality. Are they understood to be one and the same thing? Whilst there is much public data and opinion on economic benefits to having gender parity within organisations and how it can help support women’s empowerment and inclusion in male dominated professional sectors, public perception on the topic may paint a different picture. In this thesis, the social media platform Twitter is used to collect data to conduct a content analysis in order to understand public sentiments in response to one company’s perceived success in their organisational gender parity initiative. That company is American tech organisation, Duolingo. In 2018, Duolingo posted via Twitter that they had achieved a 50:50 male:female ratio in their recruitment of new engineering hires. The response on Twitter reveals that whilst many Twitter users agreed with Duolingo that this was a success, many did not. The Tweets are classified and analysed according to sentiment and coded according to the core topic in their communication – gender parity, gender equality, and diversity and intersectionality - to gain an in-depth understanding into how the public understands and reacts to these concepts. By analysing 275 Tweets through textual and visual analysis, this thesis supports an investigation via case study as to whether or not gender parity is publicly perceived and understood as a positive organisational strategy towards gender equality. Or whether it is seen to be exacerbating gender inequalities and perpetuating gender and intersectional stereotyping, biases and norms.

Keywords

Gender parity, gender equality, diversity, intersectionality, Twitter, social media, organisational strategy, content analysis, sentiment analysis

Acknowledgements

Thank you first and foremost to my family and loved ones. This would not have been possible without your encouragement and support. To my supervisor, Stina Backman, I express my sincere gratitude for your guidance, direction, and kindness throughout this process. And last but not least, to all those persons whom have an opinion or thought on gender parity, gender equality and intersectionality: may our workplaces and social structures both today and in the future continue to learn and grow with your voices, so that they can be equitable, just and provide opportunities for all persons.

(3)

3

Introduction

Gender parity, gender equality, and diversity and intersectionality. Related concepts, but they are not all one and the same thing. The benefits of gender parity - a numerical concept used to compare relative numbers of men and women - has been recognised at the highest levels of the United Nationsas both a moral imperative and an operational necessity (2020). The benefits of gender parity have been advocated and backed by data to show the economic and organisational advantages of achieving parity in the workplace (McKinsey 2016). However, little research currently examines the public perceptions on the topic and what gender parity, gender equality and diversity and intersectionality mean for individual persons. In this case study, the Twitter social media platform is being utilised to provide an insight into these queries, and to understand how public individuals interpret and react to these concepts. Through the data analysis process, this thesis applies sentiment analysis and data coding, combining quantitative and qualitative approaches. The results are used to investigate public perceptions as to whether gender parity initiatives contribute towards gender equality, or if through the use of binary gender categorisations, they are perceived as reinforcing gender stereotyping, norms, and biases.

The concept of gender in organisations, and the investigation of strategies to analyse how targeted recruitment initiatives can create, or constrain, opportunities for women and men in the workplace, can generate many different stories (Alvesson and Billing 2009, p. 11). Through this case study, one story is investigated in order to produce knowledge on how public perception and sentiment regarding gender parity, gender equality and intersectionality can help inform strategy and/or policy development for the empowerment of women and marginalised groups in the workplace. The thesis will be analysing data posted on the Twitter social media platform in response to one organisation - Duolingo, an American tech company with a world-renowned language learning application - who promoted their own gender parity approach and its perceived success in achieving a 50:50 balance of men and women in a campaign for new engineering recruits. By examining the data generated from a social constructionist theoretical approach, integrating feminist concepts, and analysing the case from an intersectional point-of-view, this research initiative aims to produce knowledge and insights that will be useful to inform change and transformation towards equal opportunities in the future.

(5)

5 Purpose

Whilst this research study does not intend to establish any definitive conclusions as to the virtue of the gender parity strategy under analysis in this case study, it is the ambition of the researcher that this thesis will help inform future strategy and policy development towards gender equality and inclusion within organisations of all types. By unveiling insights and generating analytical data on public understanding and response to the concepts of gender parity, gender equality and intersectionality, it is anticipated that this will provide details as to how such a strategy is both perceived and received by individual persons. This is hoped to facilitate production of knowledge as to how members of the public see such strategies, whether they are received positively, negatively, or ambiguously, and whether the intended meaning of the language and terms we use to communicate these issues are properly understood. This knowledge will be applicable to organisations of all types that aim to contribute towards efforts to achieve gender equality. The data generated will be able to support the work undertaken at present, and in the future, by corporations, organisations (governmental and non-governmental) and any other bodies (all of which are referred to collectively as ‘organisations’) consciously trying to affect the balance of gender representation or the opportunities availed to minority groups within their workforce or membership base.

Understanding how the public perceive and react to such concepts is a critical factor to be considered by such organisations as they plan their own tailored approaches to contributing towards gender equality. Insights such as those presented in this thesis will help such organisations to consider these concepts, study their own approaches, and explore insights from an individual person's perspective. At a time when gender parity and the representation of women in organisations is gaining increasing attention and scrutiny, studies such as this are important in order to raise awareness as to public perception and understanding. It is the public that are most often impacted by organisational activities, yet little research exists as to how the public respond to initiatives such as these. Therefore, this research is undertaken with the aim to encourage further investigation of such work, ensure policies are inclusive, and to highlight the importance of the view of the individual as we develop strategies for change at even the highest levels of society.

(6)

6 Research Questions

The thesis will be guided by the following research questions:

i). What do Tweets in reply to Duolingo’s 50:50 gender parity strategy reveal about public understanding of gender parity, gender equality and intersectionality?

and,

ii). How do Twitter users react to their understanding of gender parity, gender equality and intersectionality?

Definitions

The following key terminology or concepts form a basis for the research and analysis in this thesis and are hereby defined accordingly. The definitions outlined below are based upon those provided by the European Institute for Gender Equality (2020).

Diversity

Differences between individuals in a group of people based on gender, skills, ethnic background, sexual orientation, knowledge, cultural background, values, or attitude.

Gender

The social attributes or the opportunities which are seen to be associated with being either a woman or a man1 (and the relation between those women and between those men, and within their own gendered groups) in a given context. This is part of a broader socio-cultural context which can change over time and is not related to the biological sex determined and differentiated through identification of a person's genitalia at birth.

Gender equality

Refers to the equal opportunities, responsibilities and rights for women and men and girls and boys. Gender equality does not mean that women and men and girls and boys will all be

1_{Whilst the researcher uses the terms ‘woman and man’ or ‘women and men and girls and boys’ etc. throughout} this thesis, this is to align with the gender groups used in the case study. The researcher fully acknowledges that an individual may identify with other gendered identities along the gender spectrum (such as cis women and men, transgender, and intersex for example), and not only binary female and male gender categorisations.

(7)

7 treated exactly the same, but should also take into consideration their different needs, priorities and interests.

Gender parity

The numerical concept used in concern of relative equal numbers (or proportions) of either men and women or boys and girls, of a particular indicator. In this case, gender parity is referred to as equal numbers of women and men in a group of individuals recruited for a specific job type and level.

Intersectionality

This analytical tool is used to understand and study how a person’s gender and sex intersect with other attributes of a person's identity (such as race, ethnic group, social status, sexuality, age etc.) with the result of creating or contributing to a unique experience of disadvantage or discrimination.

Literature Review

A review of existing research and literature on the topics of gender parity, gender in organisations and the implementation of gender parity initiatives revealed a selection of case studies conducted to examine these issues in the past. Studies of gender parity within Romanian academia and of women working in science in Colombia are two examples of prior research conducted on the topic which were informative for the purposes of this thesis. Furthermore, an investigation of past work utilising the Twitter social media platform and its utility as a data source was also conducted. This revealed articles such as the Hawaii International Conference piece on the relevance of social media and sentiment analysis in analysing responses to organisational communications. These select examples, further explored below, provide an insight to currently available scholarly works related to this thesis topic, whilst also supporting a positioning of, and a need for, this research within the field of gender studies and the broader social sciences.

The case study approach has been used in past academic writing by many authors to analyse gender parity in given contexts and settings. In one example, staffing data from Romanian academia was analysed to understand how gender parity can contribute towards sustainability in university management structures (Drumea et al. 2020). The analysis of quantitative data

(8)

8 based on male and female gender groups in management roles in Romanian universities was used to assess how the country, through its universities, is achieving their goal to demonstrate sustainability. By analysing gendered propensities in direct and secret voting processes for high-level managerial positions within academia, the researchers were looking at merit-based decision-making as well as gendered attitudes in their election processes. The statistical values and analysis presented in this study were all quantitatively based. Therefore, they were not able to provide qualitative understanding of the potential reasons why the country has a feminine majority in academia as a result of gender parity initiatives. They were also not able to present findings related to the public’s personal views on gender parity and how it contributes to gender equality, nor the representation of intersectional persons, within the universities.

A study of women in science focusing on gender parity throughout the twenty-first century in Colombia was also reviewed (López-Aguirre 2019). The findings of this study were interesting as they concluded that medical and health science were the only fields in Colombia where gender parity had been achieved to date. However, they also implemented a future projection model to understand how long it would take across the country to achieve gender parity in related fields. The findings revealed that even though medical and health science had reached male:female equality in so far as academic membership numbers, the data was anticipated to demonstrate a steady decrease of women’s representation over time to come. Lack of funding, gender biases and poor protection of women's rights were all recognised as possible factors for the anticipated decline. Again, this study was based on quantitative data and did not analyse public perception towards gender parity initiatives as an organisational strategy intending to contribute towards gender equality, nor how or where intersectionality is included in the results.

Both these articles demonstrate the recent emphasis on gender parity to assess how organisational structures should optimally be gendered numerically and to understand effort towards sustainable outcomes such as gender equality in the workplace and academia. Additionally, they demonstrate conclusive outcomes on the benefits of gender parity initiatives to contribute towards the strategic objectives of women's economic empowerment and right to equal opportunities in employment. But from the wider literature base reviewed, there has not been any evident analysis of the positive and negative perceptions of gender parity initiatives as seen or expressed from people concerned with the issue. The data generated by these and similar studies help us to understand the importance of gender parity as an initiative towards

(9)

9 achieving gender equality. However, they have not been able to reveal the sentiment or perceptions of the public to inform future actions in these areas. In this light, the significance of Twitter as a data source and the use of sentiment analysis in past literature must also be understood in order to inform this thesis.

The proliferation of academic research papers on the use of Twitter as a data source and its benefits for conducting sentiment analysis are apparent. There are a multitude of articles across many fields of study which acknowledge and demonstrate this. In a paper on education for sustainable development and climate change, researchers acknowledge that the traditional methods of data collection such as interviewing and participant observation could lead to biased results as it depends on the willingness of the participant to make honest disclosures (Goritz, Kolleck and Jörgens 2019). However, they claim that using data sources derived from the Twitter platform to conduct content analysis could be more reliable given that the user and their discourse are automatically incorporated into the analysis. They claim that Twitter is an important data source for a plentiful supply of short, mainly to the topic communications from and between users on any given subject, across the globe, and which can be traced back across years. Although they acknowledge that through Twitter, the ability to benefit through observation of real-life interactions in-person is compromised, it opens a new field of data and opportunities to investigate any topic. This is of importance for this study given its validation and positive conclusion in using Twitter as a data source.

An article produced for the Hawaii International Conference on System Sciences focusing on the sentiment mining of Twitter data was also relevant to this current study (Salehan and Kim 2020). The authors acknowledge that the world of social media has drastically transformed the way that human beings communicate with one another. The article focuses on how emotions are aroused through the social media context, then investigates the result that this emotional stimulation may have on retweeting of Twitter content. Their research model contributes to literature in this field from a theoretical perspective by exploring how Twitter users will use information revealed to them through the social media platform, and whether their sentiments and emotional arousal will affect how they go on to share this information. This is relevant in this study as it shows how sentiment analysis of data from the social media platform has in the past been used to validate research on organisational messages. It also provides an insight as to past work exploring perceptions of Twitter users and what result this may have on user’s subsequent thoughts or actions. This is a critical point of confirmation for this study

(10)

10 which aims to investigate the sentiments of Tweets in order to understand how users are interpreting and reacting to a set of defined concepts – gender parity, gender equality and intersectionality.

Although a review of past literature is not able to reveal any specific works on the use of Twitter data to conduct sentiment analysis related to the topic of gender parity, the lack of data reveals both a gap and an opportunity for this study to contribute in this field. The thesis will also be utilising a previous case study and research assignment undertaken by the researcher themselves, entitled ‘Tweeting towards 50:50 - What can Twitter reveal about public sentiment and gender parity strategies?’ (De Kretser 2020) that was submitted during the Linkoping University course ‘Analysing Change’. This previous case study focused on two key research questions; firstly, what were the Twitter reactions to Duolingo’s 50:50 gender parity strategy, and secondly, what do the Twitter reactions reveal about public sentiment towards gender parity initiatives. That case study noted that the data analysis could be taken some steps further in order to reveal greater insight into and between the concepts of gender parity, gender equality, intersectionality and how these are perceived and understood by the public through a social media platform such as Twitter. Therefore, this thesis will be using that case study as a point of departure, using the previous research as a platform to commence further exploration, re-analysis and consideration of the Twitter data. This is to better understand how the public perceive gender parity strategies such as that implemented by Duolingo, and what consequences such strategies may be viewed to have for gender parity, gender equality and intersectionality.

Methods, Theoretical Framework and Ethical Considerations

Methodology

A case study approach that incorporates and utilises mixed methods has been selected as the research methodology for this study. As Martin Denscombe explains, a case study focuses on one instance of a particular phenomenon with the objective to generate an in-depth analysis and account of the processes and findings occurring in that instance (2007, p. 35). The researcher aimed to investigate a real-life case of the application of a gender parity strategy and the sentiments, understandings and reactions to strategy implementation in an organisational context, and the case study approach was found to be appropriate to discover findings

(11)

11 following an inductive logic (Ibid, p. 38). This thesis utilises a discovery led case study to explore and attempt to understand the issues identified within the particular setting, in this case, a thread of tweets in response to a statement made on Twitter by a company regarding the outcomes of the implementation of their gender parity recruitment strategy. This case study is to be distinguished by its defined boundaries of the original Twitter post published by Duolingo, and all tweets made in reply to the post at the date of original data collection (17 March 2020).

Although each case study is an example of a unique set of data and content, Denscombe advocates for the notion that it may still be perceived as an “example of a broader class of things” (2007, p. 43). The Duolingo case was selected as it provided an example of an organisational gender parity strategy implementation scenario, whilst still being able to present data to be analysed which may be indicative of trends of thought and understanding which could help produce knowledge on public perceptions on the topic. The researcher acknowledges the challenges in generalising all findings from a case study (see alsosection on Limitations). The applicability of the findings from this study are interpreted and analysed from a broader public sentiment and social constructionist theoretical lens. This is with the ambition to find similarities in how this case study may be able to produce findings that can to some degree be applied to other organisational contexts.

A mixed-method approach - defined as the utilisation of quantitative and qualitative methods within the one research project (Denscombe 2007, p. 108) - has been utilised in order to improve the accuracy of interpretation of the subjective findings of this study. It is also used to corroborate the data and knowledge being produced. The case study approach lends itself to the application of a mixed-method study, providing a more complete picture of results and allowing alternative perspectives of content analysis to be applied (Ibid, p. 110). Gayle Letherby, in her text ‘Feminist Research in Theory and Practice’, notes that many feminists have been critical of the use of quantitative methods, arguing for the application of qualitative tools in order to fully represent and produce knowledge to understand and represent both women and men’s lives (2003, pp. 80-81). Yet Letherby acknowledges that method research is an “innovative method” (Ibid, p. 81). In this study, the utilisation of mixed-methods was determined to be appropriate, and innovative, in that it provided alternative sets of data for the researcher to deal with contextual and content-based subtleties of a complex social situation of the issue of gender equality in an organisational context. Furthermore, by

(12)

12 combining qualitative content analysis together with quantitative data on numerical trends in sentiment analysis, the researcher aimed for improved accuracy of findings, as well as a more complete picture (Denscombe 2007, p. 109).

This research study begins from a point of departure from a previous, albeit more limited case study, conducted by the researcher themselves (De Kretser 2020). This study continues with the utilisation of the method of sentiment analysis to classify and understand the data derived from the Twitter platform. However, the previous case study classified data using three sentiment categories; positive, negative and ambiguous. This thesis intends to further analyse data by subsequently coding each Tweet in order to understand the core conceptual content. The coding was undertaken according to categorisation as either predominantly focused on gender parity, gender equality, diversity and intersectionality, or as non-classifiable.

The empirical data used in the thesis research was downloaded from Twitter user’s public posts (and paraphrased for ethical concerns), in addition to statements and articles released by Duolingo as well as other relevant empirical literature on the topic of gender parity, gender equality and intersectionality. A total of 275 Tweets were publicly posted in response to the original Tweet published by Duolingo on 11 October 2018. All 275 Tweets were downloaded and then categorised using the two-stage content analysis approach. This was firstly through the sentiment analysis data classification system developed by the researcher, and then subsequently through the data coding process.

Content Analysis

As part of the case study, and together with the mixed-method approach as described above, an interpretative content analysis technique was applied to the Twitter data. Content analysis is a research method to support the study of communications, in this instance comprising both text and visual content (Graphics Interchange Format (GIF) and emojis) from posts made on Twitter as per the boundaries of this case study. As defined by Drisco and Maschi, this research technique supports the process of generating inferences from the specific messages being studied, in this case, the tweets (2015, p. 2). In this thesis, content analysis is used in a descriptive manner to identify the attitudes, sentiments and interests of individuals (the Twitter users). This is to be able to evaluate and compare the tweet communication content with one another and also with previously published materials (Ibid), in this case on the topics of gender parity, gender equality and intersectionality. Given the social constructionist theoretical

(13)

13 approach being applied in this study, and the recognised subjective nature of the research, the content analysis technique is interpretive in nature. This means that the researcher is “systematically identifying specific characteristics of messages” (Ibid, p. 3) in order to derive meaning from the content. As tweets commonly include everyday language, humour, irony, and sarcasm (both written and visual), the interpretation of the communications is a necessary component of the content analysis process. This is because it is not possible to simply analyse the manifest content (the literal communication) in the data set (Ibid).

Both stages of content analysis – the sentiment analysis and the coding processes - relied on interpretation by the researcher to generate findings. Whilst applying a systematic process as detailed further below, the researcher recognises that the method applied remains subjective and is cognisant that objectivity should in no way be assumed in the application of the research methodology.

Sentiment Analysis

Sentiment analysis was the pre-dominant method applied through content analysis in this case study, in order to determine the emotion, opinion and/or attitude of the Twitter users through their tweets. Despite the availability of options to purchase sentiment analysis software online, the researcher manually downloaded and conducted individual sentiment analysis for each of the Tweets in the thread being analysed. The major challenges of conducting Twitter sentiment analysis, as recognised by Nabizath Saleena Ankit (2018) in their article on the topic, is threefold. Firstly, the fact that Tweets are generally in informal language, secondly, that Tweets are so short (maximum of 140 characters), and thirdly, that abbreviations and acronyms are commonly used (Ibid, p. 1). Although the manual sentiment analysis process was time-consuming, the process allowed the researcher to ensure the accuracy of classification and became familiar with the data set in the process, both of which an online data analysis software program would not have made possible.

Each Tweet in response to the original post by Duolingo was read by the researcher (and often re-read multiple times) and then the sentiment interpreted from the Tweet was classified according to a system comprising three categories. The categories are outlined in Table 1 below which also provides the guiding indicators used to support consistency in the researcher’s analysis process.

(14)

14

No. Sentiment Analysis Classification

Indicators

1 Positive Expresses positive sentiment through one or more of the following:

- positive, optimistic, and/or congratulatory language (i.e. use of word/s such as ‘great’, ‘well done’, ‘excellent’, ‘awesome’, ‘fantastic’)

- direct statements or discussion threads on the perceived benefits to gender parity and gender equality initiatives (i.e. directly supporting the Duolingo strategy implementation where other users are criticising the same; argumentation as to why we need proactive approaches to get more women in a workforce, etc.)

- use of emojis or GIFs which express positive imagery (i.e. hands clapping, cheering, smiling, thumbs up, etc.).

2 Negative Expresses negative sentiment through one or more of the following:

- negative, unoptimistic, critical, and/or derogatory language, or statements which are mocking in nature (i.e. should hire the best person for the job not just hire a woman; they should hire on

qualifications not gender; claiming sexism or bias against men, etc.) - obvious sarcastic and negative tone (i.e. ‘only hire a female because she is a female’; ‘have you ever heard of the anti-discrimination act?’) - direct statements or discussion threads on the perceived disadvantage to gender parity initiatives (i.e. claims that company productivity will be reduced if they do not hire best person for the job but hire based on gender, etc.)

- use of emojis or GIFs which express negative imagery (i.e. boo’ing, disgusted or unhappy facial expression, thumbs down, etc.).

3 Ambiguous Statement on the users view towards gender parity, gender equality or

intersectionality generally but may not have been directly related to the original post, and was classified as vague, open to interpretation and/or could neither be clearly classified as distinctly positive nor negative in their sentiment. Includes but is not limited to:

- requests for further information to determine the user's opinion (i.e. request for details on company recruitment process; statements on women's participation in the workforce generally)

- intersectionally focused (i.e. requesting data on statistics of intersectional or diverse persons employed by the company)

- replies as part of a thread which discussed gender issues generally but not the original post specifically (i.e. discussion on how to remove gender identification of all applicants)

- completely off-topic (i.e. review of user experience of

the Duolingo app generally; messages from a conversation thread then went on to ‘troll’ or to judge or criticise another user)

- consisted of an emoji or GIF image that was unclear in its intended meaning (i.e. emoji of a face crying with laughter)

- were not able to be understood and interpreted by the researcher (i.e did not make linguistic sense; in a language other than English) - tagging another user to re-Tweet the original post.

(15)

15 As both a researcher and a human, the author was not able to remove oneself from these constraints, and fully acknowledges the influence that their personal situatedness may impact on the interpretation of each Tweet. With this in mind, the researcher has attempted in each and every analysis of tweet sentiment, to be conscious of any potential sways or biases in personal perspective. As such, the researcher made efforts to refer to the classification table diligently and repetitively to ensure consistency in the sentiment analysis classification process.

Data Coding

As part of the interpretive content analysis process, and as the second stage in the analysis of the data set, data coding was conducted on all the tweets within this case study. Data coding was used as an organisational technique, applied to allow the researcher to ‘code’ or conduct specific identification of relevant materials for the future analysis in this study (Drisco and Maschi 2015, p. 72). The coding process is primarily the identification and tagging for future utilisation of content which is interpreted to be most relevant to the research questions. As such, the following code list was used in the content analysing process to identify those tweets that were relevant to the concepts being further analysed in the thesis:

• Gender parity, • Gender equality,

• Diversity or intersectionality.

A separate coding category of ‘non-classifiable’ was used for all remaining data that was interpreted to not show any relevance to the three key conceptual categories requiring identification for the analysis stage of this study.

The coding process was a critical step in the interpretative content analysis method applied to this research. It helped to ensure the researcher could understand the content and compare through quantitative methods (number of tweets), which concepts were most commonly being discussed. It also allowed the researcher to identify and group tweets containing communication on the same concepts, and compare views, sentiment, and opinions. It must be noted that in the coding process, the researcher did not rely solely on specific words (‘gender parity’, ‘gender equality’, ‘diversity or intersectionality’ as examples), but rather coded based on the “overall or symbolic meaning of phrases or passages” relevant to the study (Ibid).

(16)

16 Theoretical Framework

Social Constructionism

In the 1960s, feminists argued against the conventional, gendered norms and stereotypes persisting for women and men (Dicker 2016). The traditional division of labour in both society and the home saw men perceived as the breadwinner, and the women cast as homemakers. The emphasis on men going into the public sphere to earn an income emphasised their need for skills such as strength, dominance and even competitiveness, that would help them get ahead in the workplace (Ibid). Contrastingly, with the expectation of the role to care for the family and the home, women were encouraged to show characteristics that exhibited a nurturing, cooperative and dependent nature (Ibid). But feminists argued that these roles were not biological - we were not born this way - but that they were socially constructed. Feminist theory of the social construction of gender emphasises that the gender that each individual identifies with is contingent on time and place, and that our culture and the people around us and our interactions with them all play a role in defining our place on the gender spectrum. Yet social constructionism also emphasises how gender relations are impacted by power and status, identifying a perceived lower rank for women as an attribute of their gender, and how this may be used to create differences between females and males (Berenbaum, Blakemore and Liben 2009, p. 192).

The social constructionist theory of gendered identity, it should be noted, contrasts with the genetic perspective on the role of gender development. Genetic theories on the development of gender are based upon the differences between the sexes as resulting from the genes on a person's sex chromosomes (Berenbaum, Blakemore and Liben 2009, p. 138). Because of a feminist outlook on social constructionist theory of gender being socially constructed, and of women’s status being something that is completely separate from their biological sex, this was found to mean that women’s rank could be both challenged and changed (Dicker 2016, p. 13). With this approach in mind, the second wave feminists - a notable period of feminism beginning in 1960s USA and defined by its aim to ensure equal rights for women - were particularly concerned with the notion of ‘gender socialisation’ (Ibid). Gender socialisation is the process by which “people learn the behaviours and attitudes that are considered appropriate for their sex” (Ibid). This theory of gender being interpreted through social categorisation provides an opportunity to analyse systems for division of labour as well as economical structures and how they use domination in regard to gender differences (Schmidbauer

(17)

17 and Wischermann 2011, p. 216). Considering the social constructionist theoretical approach to the case study within this thesis, this is of particular importance. It provides an opportunity to read and analyse the data with a focus on how organisational strategies such as gender parity strategies can impact upon and be reacted to by the public.

Social constructionism supports the analysis of data which is anticipated to raise questions on the social construct of an individual’s gender vs. an organisational strategy aiming to promote change towards gender equality, but which might be considered by many to reinforce binary gender definitions/gender based on biological sex. Gender parity with its emphasis on a dual classification, male:female, may potentially be found to be at odds with the social construction of gender. The impact of the social construction of gender is important to this study, not only when assessed from the viewpoint at the organisational level, but similarly from the individual’s perspective. In its attempts to investigate understanding and reactions to feminist concepts from public individuals' communications on Twitter, this thesis will be required to analyse how the individual perceives and understands the social construction of their identity in terms of gender. By analysing the topic from a social constructionist framework, it is anticipated that gender may be seen on different and oftentimes opposing levels - from the organisational, from the researcher’s, from the historical and from the Twitter users standpoint. This is of critical relevance in our attempts to understand and learn more about gendered behaviours and attitudes, even stereotyping, and the challenges to them (from both the organisational and the individual perspective) when it comes to a person's gender and its impact on career choices and opportunities.

The social constructionist approach is also being utilised given that it lends itself to research methods which may be seen to differ from traditional quantitative studies of feminist issues (Berenbaum, Blakemore and Liben 2009, p. 192). Social constructionists have been noted for the utilisation of discourse or content analysis and the use of narratives to address research questions (Ibid). As a mixed-method case study approach, this research, its process of analysis and interpretation of Twitter data, and the necessity to reflect upon the fact that knowledge in this instance is considered to not be objective - “objective knowledge is never truly possible” for this post-modern perspective (Ibid, p.191) - lends itself to the social constructionist approach. The researcher sees this as a critical factor to be considered in the analytical framework when interpreting statements of a personal nature being made on a social media platform such as Twitter, as is the case for this thesis.

(18)

18

Intersectionality

Intersectionality will be used as a foundation for the theoretical and analytical framework of this thesis to guide those specific components of the analysis that aim to investigate how gender parity and gender equality are understood in light of other defining characteristics based on social and political identities. Intersectionality as a theoretical approach is particularly relevant in the analysis of women’s identities and the interplay between sex, gender, and other traits such as race, sexuality, class etc., in situations where what is seen to be efforts to create a ‘diverse’ workforce are being analysed. Intersectionality as a theoretical concept came to prominence in the 1990s in the USA, and is noted for its challenge to the notion of ‘women’ as an all-encapsulating categorisation of persons, and its emphasis on the need to instead focus on the individual (Dicker 2016, p. 7). This notion of women not fitting into one broad category was already raised back in 1979 when activist Barbara Smith spoke at the National Women’s Studies Association conference. As Dicker notes, Smith had stated that “women come in all shapes and sizes, with all kinds of concerns; to talk about “women” as one broad category is thus impossible, since a black woman, for example, cannot separate her race and her sex - these axes of her identity intersect and are always present” (2016, p. 7). This statement highlights the third-wave movement’s focus on the different backgrounds and needs of all women, and the necessity to keep this factor in mind when advocating for women’s rights and gender equality. The questioning of what gender equality therefore means is integral in the analysis of strategies such as gender parity which claim to work towards the achievement of gender equality, but which use the binary norms of male and female, leaving out of the equation for potential consideration any other identifying attributes which both men and women may relate to. Given that gender parity is used by many corporations as a strategy to promote diversity and inclusion, using intersectionality as a theory to analyse the empirical data of this study is intended to aid the researcher. It is anticipated that its use will support understanding of how public sentiment towards inclusion and parity might be able to inform future efforts towards change in organisational culture in relation to diversity beyond gender binary norms, by looking into social power relations and structures.

Situatedness

As Ramazonoglu and Holland state, “no social researcher starts from scratch in a state of social, intellectual or political isolation” (2002, p. 142). In embarking on such a research process, it is integral to a researcher that they reflect on the baggage that they carry with them, in terms of

(19)

19 values, politics, identity and theory (Ibid, p. 144). One can consider that it is essential to reflect upon their own situatedness within the topic, the sphere of gender studies, and the world at large, in order to understand their own reflexivity and process of interpretation as conducted through this thesis. Donna Haraway (1988) defined this as ‘situated knowledges’; a reference to the fact that all of our knowledge is created and emerges from our own positional perspective. The researcher fully supports this notion and has reflected on this throughout the research process. The interpretation of each Tweet throughout this study was a subjective process based upon the researchers own personal attachments to meaning and tone based on their own history, upbringing, and socio-cultural norms. This also aligns with the postmodern social constructionist research approach which supports the notion that “knowledge is not objective” (Berenbaum, Blakemore and Liben 2009, p. 191). All knowledge we produce is impacted by our histories, our understandings of the world, and the lens through which we see and interpret everything around us.

The knowledge thus produced through the content analysis process is reflected by the researchers own social construction and identifications as a white, Australian, heterosexual woman, a parent, someone who identifies as a feminist and has worked in gender mainstreaming in the development sector over the last 7 years. As such, it has not been possible for the researcher to “view from above, from nowhere” (1988) as Haraway stated in response to the notion of a guise of neutral thinking, of being objective and universal in our approach to our topics as researchers. It is recognised and acknowledged that objectivity in a topic which revolves around the researcher’s personal interpretation of written and visual data often including comedy, innuendo, sarcasm and other linguistical nuances will be seen through a subjective lens, and that all findings will be based from the researchers situatedness as both person and student.

Ethical Considerations

As stated in Twitter’s Privacy Policy “most activity on Twitter is public”, and this includes profile information, username, and each users Tweets (2020). This is unless the Twitter user has selected to have a ‘protected’ account, in which case their tweets can only be viewed by their followers. All tweets used in this study were public with the researcher not having access to, nor knowledge of, any protected tweets made in the thread. Nonetheless, the researcher felt it imperative to protect the privacy and retrievability of personal user details and their respective tweets in the name of ethical research. In an ideal research situation, the researcher

(20)

20 would have been able to contact all users whose tweets are included in this study in order to request their informed consent to participate. However, this has not been possible given the large scale of data. Additionally, in the absence of an established academic guideline on ethics and the ethical utilisation of data obtained on social media platforms in academic writing, the researcher has established their own ethical guideline and process to protect the anonymity of each of the Twitter users whose tweets are utilised in this study.

The guideline for ethical use of data produced through extraction of tweets which informs the process followed in this essay is that all personal users, in the absence of provision of informed consent, deserve the right to have their identity protected. Additionally, all personal Twitter users concerned deserve the right to have prevented the ability of others to search online for their tweet content in order to discover their original tweet or retrieve any personal information provided in their Twitter accounts as publicly available details. To abide by this chosen guideline and to protect the anonymity of user details, each tweet quoted in this paper has not had the Username of the author stated in order to protect their identity. Furthermore, where tweet content is quoted in this essay, the content has been paraphrased to avoid any opportunity for others to discover or search for the original tweet online. Most of the references to Twitter content utilised in this thesis are selected abstracts from the original tweets, reproduced in paraphrased form (i.e. most quoted Twitter content is not the complete tweet but a specific selection of the relevant words have been paraphrased). It must be noted that the original tweet that this study refers to – the statement and image published by Duolingo on 11 October 2018, has been referred to and quoted in its original form, with a weblink provided to the original post. As the post was made by the Duolingo corporate communications Twitter account and not an individual person, and given the fact that their account is public, has over 350,000 followers within its global Twitter network, and their full history of tweets is publicly available online, anonymity has not been provided to the company in this study. The researcher's interpretation of the Twitter privacy policy and its statement that activity on Twitter is public and is publicly available (i.e. no Twitter account is required to access the Duolingo post, nor the public tweets posted in response to the original post) provides the basis for their decision to use the post in its original format.

Limitations

The utilisation of the case study approach in any research project may be subject to criticism regarding the “credibility of generalisations made from its findings” (Denscombe

(21)

21 2007, p. 45). Case studies may be subject to scrutiny in so far as claims to apply the findings and knowledge generated from one case, and how they can be applied beyond the scope of the case boundaries. Furthermore, the reliance on interpretative methods of analysis, as opposed to an emphasis on statistical methods, may be challenged in their ability to produce objective results that can be applied to future research. In order to produce research with rigour and aligned with the aims of the thesis to produce knowledge that may be capable of informing future organisational strategies on the topics of gender parity, gender equality and intersectionality, the researcher has attempted to mitigate these limitations to the study by ensuring transparency (within ethical considerations) of the data utilised, clarity in the analytical concepts being applied, and reliance on established theory to support the interpretation of the case study results.

Twitter, as a publicly available social media platform (where even non-registered users can view Tweets and read conversation threads), provides a readily available (subject to internet connection) forum for investigation of any number of topics. However, there are some obvious limitations to conducting a study using Twitter as the basis for discovery of the empirical data set used. First to be noted is the representativeness of the data. The particular group of users which have replied to the Tweet in focus may be followers of the company, or have come across the original Tweet by having it ‘re-Tweeted’ (shared via @ to them) by a friend or follower. The group of users cannot be seen as representative of the entire global nor American national (given it is an American company) population. Nor can it be representative of Twitter users as a group. As Ahmed, Bath and Demartini write in their article focussing on Twitter as a data source, “not all users will tweet about a topic of interest” (2017, p. 20). As a result, the data set used in the thesis cannot be considered to be representative of the larger population, and instead, must be considered to be an example of the views of a group of people who were interested at the time in the particular topic of the Tweet or the ensuing conversation thread. It is also not possible to assess the authenticity of Twitter user profiles. Like other social media platforms, usernames can be selected at will upon registration and user profile photos do not have to be actual images of the user themselves. A user can create multiple user accounts if so desired and given an ability to disguise oneself behind an unofficial name or image, there is no way to confirm if user account information is authentic or not. This may potentially have the implication of users tweeting false data and inauthentic information or viewpoints. Therefore, all data included in this study is considered to be representative of this particular group of Twitter users, but not of a wider population.

(22)

22 An additional limitation involves the ethical concerns of reproducing Tweets in their original form. Given the multitude of Twitter users included in the research study (i.e. by responding to the tweet in focus) the researcher was not able to be in communication with each user to obtain their informed consent to participate in the study. Due to this, in addition to the researchers desire to protect the anonymity of all users whose Tweets are included in the data set, the researcher has made their best effort to paraphrase all Tweets ‘re-produced’ in this thesis. The fact that the Tweets are not being reproduced in their original wording may be a limitation of the study insofar as the researcher may have unintentionally misinterpreted their understanding of the original Tweet, sarcasm, or visual emojis or GIFs used based on their own cultural background and knowledge.

The Case Study

Duolingo, Inc. (Duolingo) is a private American company who provides online language courses through their website and mobile application. The company was launched in 2011 and is estimated to have over 300 million users worldwide (Lardinois 2018). In 2018, the company was named by Inc. Magazine as the Best Workplace of the year (Inc 2018). In the same year the Duolingo company used its Twitter account to communicate to its followers that it had achieved gender parity, a 50:50 male:female gendered result for its new software engineer hires (see image below).

(23)

23 The Tweet thread in focus is a reference to an achievement the company had made earlier in the same year. On 11 January 2018, a company writer named Jeesoo Sohn published a piece on the Duolingo webpage entitled ‘How Duolingo achieved a 50:50 gender ratio for new software engineer hires'. On their website, Duolingo used this piece to explain why this was such an achievement “in light of the historical gender imbalance in the tech industry” (Sohn 2018). The piece goes on to explain how this “didn’t happen by chance” but that the 50:50 result was part of a new recruitment strategy implementing both internal and external changes when conducting their university recruitment cycle (for best new graduates) (Ibid). Duolingo explained that following a thorough review of their company’s recruitment approaches, they had decided to make a more pro-active effort to approach women’s groups at university and career fairs they attended, both for networking and to learn about what it would take to have more women join the company. These external efforts to approach more women through university recruitment rounds were combined with internal efforts such as ensuring there was a woman on all hiring panels and removing gender identifying details from applications to reduce the possibilities of unconscious biases. The company was proud of the success of their efforts to “making gender balance a priority” and had thus shared this story through the Twitter social networking platform (Ibid).

Being a tech company with such a widespread user base and broad Twitter following makes this case study one of great interest in light of public reaction to corporate efforts to achieve gender parity in new recruitment initiatives. Twitter, with the ability for users to engage in conversations with people of whom they are not familiar, on such a wide variety of topics of individuals’ personal interests, makes it an intriguing source of data to analyse. In its simplest interpretation by the researcher, we see that this is a case study of one company in the still male-dominated tech industry who wants to share “a milestone that (they) are very proud of” (Sohn 2018) with their followers. This Tweet from Duolingo had received 275 responses at the time of data collection. A quick scan over the replies shows that whilst many were, not all Twitter users reacting to the Duolingo post were positive about the gender parity strategy.

Whilst the strategies employed by the Duolingo company and the background to their decision to initiate a plan to achieve gender parity in new recruits is very interesting, this thesis focuses on how the particular case provides an opportunity to investigate and analyse public understanding and reactions to gender parity as a strategy and tool to achieve

(24)

24 change towards gender equality. Any change can be met with both positive and negative sentiments. Twitter, with its brief statements from users as to their views on such a topic, provides an opportunity for these sentiments to be consolidated and interpreted with a goal to help inform future change towards gender equality.

Data Analysis

In order to analyse the research questions forming the basis of this thesis, both quantitative and qualitative methods provided strong utility for both interpretation and understanding. The first phase of content analysis, the sentiment analysis, was an expansion of a similar sentiment analysis process conducted by the researcher in a case study also based on the Duolingo Twitter response thread (De Kretser 2020). The sentiment analysis conducted in this thesis expanded on that process of classification of sentiments, further analysing written communications as well as the visual communication of GIFs and emojis posted by Twitter users in the data set. The first phase of the content analysis here, through the sentiment analysis, corroborated the quantitative findings of the original case study conducted by the researcher in so far as the same number of tweets were classified in accordance with the classification system outlined under Methodology. The data, once aggregated into the specific classification groups, provided an opportunity to analyse and review the sentiment of each Tweet as part of a collective assessment. Image 2 below shows the quantitative comparison of the sentiments of the aggregated tweet data.

As stated in the earlier case study conducted by the researcher, the majority of tweets, 129 in total, were interpreted as ambiguous, followed by 117 classified as positive, and 29 which were classified as negative (Ibid). The sentiment analysis of the Twitter data provides an indication as to public opinion in reaction to reading of the Duolingo post regarding their corporate perception of a successful implementation of a gender parity strategy. The analysis of the data in this Positive-Negative-Ambiguous classification system provides an overview as to public reaction to such strategies generally, and the immediate public response as to acceptance or rejection of strategies of this nature. As the researcher stated in the earlier case study, if the tweets classified as ambiguous are disregarded, a comparative analysis between positive and negative sentiment shows a strong positive majority of 80.1% of tweets in comparison to 19.9% classified as negative. From a quantitative outlook, this simplified data set shows most of the public sentiment expressing acceptance of gender parity initiatives as conducted by

(25)

25 Duolingo. This information can be used to inform further application of such strategies, however, in this study, the second stage of content analysis - data coding - provides further insight and knowledge as to public understanding of and reactions to the concepts of gender parity, gender equality and intersectionality.

Image 2: Number of tweets as classified through sentiment analysis

The data coding system was applied to this case study to organise the tweets in a manner that identified the Twitter data relevant for further qualitative analysis. Through the coding process, the four categories of the code list were applied to each tweet in order to understand which concepts were most discussed across the Twitter thread. Image 3 below provides an overview of the aggregated data coding findings, depicting the total number of tweets per code. As the graph indicates, the majority of the tweets (53%) were not able to be classified in accordance with the content analysis coding system applied in this study. This was reflective of the large number of tweets which used emojis or 5 or less words which may have been classifiable according to sentiment (i.e. emojis of clapping hands, or tweets consisting of singular words such as “congratulations” etc.) thus could be analysed by the researcher according to sentiment, but which were not interpreted as communicating a reaction, insight or opinion for the coding process. The data in Image 3 does present an overview on whether the majority of tweets which could be coded were actually communicating a response to gender parity - the topic of the Duolingo post to which the Twitter user was responding - or one of the other core concepts being investigated by the researcher (gender equality or diversity and intersectionality). The coding process, from a quantifiable perspective, shows that when comparing only the classifiable tweets, more tweets made a conceptual reference to gender equality (41%), than to gender parity (32.5%). An interesting outcome of this numerical

(26)

26 comparison is the finding that of the classifiable tweets once coded, approximately 36% of those tweets were communicating on the concepts of diversity and intersectionality. This highlights the necessity to analyse the data from a qualitative standpoint in terms of these three conceptual topics to fully understand the trends of thought and their corresponding individual reactions.

Image 3: Number of tweets as classified through the data coding process

Of interest is the comparative analysis of both the sentiment analysis findings correlated with the coding system based on tweet content. As presented in Image 4 below, it is evident that of the tweets coded as primarily communicating on the topic of gender parity, the majority (with almost 60%) were negative in sentiment. When compared with the other sentiment classifications, we see that this is an overwhelmingly large cluster of tweets, providing evidence that the majority of Twitter users that were commenting specifically in relation to the concept of gender parity, were not in favour of the strategy as per the implementation by Duolingo. The same was found, albeit in a smaller percentage group with 33%, to be coded as communicating on gender equality, again with negative sentiment. The tweets coded as communicating on the topic of gender equality were strong across all sentiments, positive, negative and ambiguous, indicating that whilst gender parity as a concept may be more polarising, gender equality is perceived and interpreted across the sentiment scales in not-so contrasting opinion groups. Diversity and intersectionality were coded with a large majority of the relevant tweets having been classified as ambiguous in sentiment. 22% of the total of the ambiguous sentiment tweets were communicating on the concept of diversity or intersectionality. This provides evidence that the topic must be considered where gender parity

(27)

27 is of issue, but that Twitter users were largely using the platform to bring attention to the issue, rather than to express clearly classifiable positive or negative sentiments on the topic.

The quantitative analysis of the data revealed a few core concepts requiring further investigation. In alignment with the research questions, the data when classified and then compared on numerical basis emphasised a necessity to investigate further through content analysis on a qualitative level the three core concepts being communicated within the data set – gender parity, gender equality and diversity and intersectionality – in order to better understand the perspectives of the Twitter users whom had commented through the Duolingo twitter feed.

Image 4: Comparative analysis of sentiment classified data with data coding classification

Gender Parity

Understanding of Gender Parity

Whilst gender parity is a numerical concept related to gender equality, it is understood by the researcher that it is not the same thing as gender equality, as per the definitions derived from the European Institute for Gender Equality (2020). In the context of this case study, the male to female 50:50 ratio relates to the contract offers as accepted by new hires through a targeted recruitment round of engineering graduates – as specified by Duolingo in the article by Sohn (2018). The content analysis of the tweets in response to Duolingo’s Twitter post on their perceived success of the 50:50 male:female recruitment ratio does however reveal some

(28)

28 contention amongst public understanding of what gender parity means. From a definitional standpoint, tweets are interpreted to analyse whether or not Twitter users understood the concept of ‘gender parity’ in accordance with the European Institute for Gender Equality definition of the term. A number of tweets reveal a correct interpretation of the concept. One tweet included the statement “it's a quest for a 50% male and a 50% female world”. Another stated “the hiring ratio finally meets the population ratio”. These tweets are examples of those that show an understanding that gender parity does refer to the relative equal numbers for a specific indicator, again, in this case, the genders of those persons from a specific recruitment round.

In contrast however, a number of tweets demonstrated an incorrect interpretation of the concept of gender parity. One twitter user stated that “I doubt a company would really hire based on gender just to achieve equality”. Their interpretation of the meaning of gender parity is not clear, however their tweet reveals that their understanding is not aligned with the concept of a 50:50 gender ratio as they question the company’s strategic origins and intentions. Several Twitter users expressed statements such as “so they didn’t give men a job simply because they are men”, “this is discrimination” and “this is sexism against men”. Such tweets reveal another avenue of interpretation of the concept of gender parity not aligned with the understanding of parity as equal numerical ratios of a given indicator. These tweets provide an insight that for many persons, their understanding of a gender parity strategy or affirmative action and targeted recruitment policies are not about taking steps to help ensure biases do not go against women, but rather that it is being used to simply refuse men employment contract offers. As Duolingo stated, achieving gender parity involved working with universities that prioritised gender balance, and thus, whom had higher ratios of female engineering students (Sohn 2018). This is one example of the process changes made by the company, of which they state was successful, as it helped their effort to reach more female students. As a result, this change gave more female students a chance to learn about Duolingo and be encouraged to apply to join the company (Ibid).

Even without a consideration or expectation for Twitter users to have read the more detailed account provided by Duolingo on their webpage in the article written by Jeesoo Sohn (2018), we still see that public understanding of what gender parity is does cause some confusion. Whilst some tweets state outright that they see gender parity and the strategy as “discrimination” and a bias “against men”, others praised “targeted recruitment” to help

(29)

29 women be considered for the jobs. One tweet included the statement “skilled women were out there all along – they didn’t know to apply beforehand”. Another stated “active recruitment from colleges with more women automatically leads to getting lots of great applications from qualified women”. Tweets such as these provide an insight that there is public understanding of how gender parity can be operationalised as an organisational strategy for targeted recruitment activities such as that implemented by Duolingo. These tweets show recognition that for whatever reason, before the targeted recruitment campaign and proactive steps being taken to reach-out to more female students, less women may have applied for such technical engineering positions in the company. Through the corporate decision to take steps both internally and externally to aim to hire more women, more women actually applied, allowing the concept of gender parity and the relative numerical indicators of numbers of men and women, to be calculated and compared. This is indicative of a correct interpretation amongst some Twitter users of the concept and its purpose in providing a mechanism to compare, with the use of binary gender categorisation, the numerical data on applications for these positions. This then allowed them to use the data to ensure that with all qualifying details and requirements being met, they could make sure as many women were hired as men.

Reactions to Gender Parity

“A very basic example of the social construction of gender is the view that there are two genders” state Berenbaum, Blakemore and Liben (2009, p. 191). This simple division of gendered roles is often provided for in contemplation of labour posts and workplace participation, as it can be used to emphasise the stance that men are the breadwinners and women the child bearers (Ibid). In the Duolingo gender parity strategy case, we see the so-called ‘simple division’ in action – the strategy, despite its proclaimed aim to work in advocacy for gender equality, uses the dual gender system of men and women in its ambition to recruit equal numbers of both genders into its workforce. But the user reactions to this, as we see through the Twitter thread, reveal a challenge to the simple, two-group, gender classification system. Tweets that include statements such as "how does this work for non-binary people?” reveal this. They emphasise a public recognition that gender equality cannot be achieved through a simplified cultural construction that all persons be categorised into an either/or grouping of men or women.

The use of the two-gender classification in the job application and recruitment process can be interpreted to serve the cultural context within which the company is conducting

Gender Parity, Gender Equality, and Intersectionality : Public Perceptions of a ‘50:50’ Workforce