Potential for social media to challenge gender-based violence in India: a quantitative analysis of Twitter use

ABSTRACT

Gender-based violence (GBV), a global issue that disproportionately affects women, is especially pervasive in India and reinforced by existing gender norms. Starting late 2012, India experienced increased GBV-related media attention, when a young Delhi woman was fatally gang raped. Public outcry ensued through Twitter. Social media platforms, such as Twitter, can erode social boundaries and permit persons to challenge social norms and the status quo. Moreover, Twitter may provide a virtual safe space in which women in India can voice their opinions about GBV and press for social change. This article shares findings from research exploring ways in which men and women used Twitter in the aftermath of the Delhi rape case, focusing on the conversations around GBV that took place, and the opportunities that Twitter offers for more strategic engagement of individuals, especially women, to press for social change.

La violence sexiste, problématique mondiale qui touche les femmes de manière disproportionnée, est tout particulièrement omniprésente en Inde et renforcée par les normes de genre existantes. Lorsqu’une jeune femme de Delhi est morte suite à un viol en réunion fin 2012, l’Inde a commencé à faire l’objet d’une attention médiatique accrue en matière de violence sexiste. Ces événements ont suscité un tollé général sur Twitter. Les réseaux sociaux comme Twitter peuvent éroder les frontières sociales et permettre aux gens de mettre en question les normes sociales et le statu quo. Qui plus est, Twitter peut fournir un espace virtuel sûr dans lequel les femmes de l’Inde peuvent exprimer leurs opinions sur la violence sexiste et réclamer des changements sociaux. Cet article présente les conclusions tirées de recherches examinant les manières dont les hommes et les femmes ont utilisé Twitter à la suite du viol de Delhi, en se concentrant sur les conversations sur la violence sexiste qui ont eu lieu et sur les occasions qu’offre Twitter en ce qui concerne une mobilisation plus stratégique des personnes, et en particulier des femmes, pour réclamer des changements sociaux.

La violencia basada en el género (vbg), un tema de preocupación mundial que afecta de manera desproporcionada a las mujeres, es especialmente generalizada en India, donde se ve agravada por las normas de género prevalecientes. Desde finales de 2012, y a raíz de la violación tumultuaria de una joven de Delhi que murió posteriormente, el tema de la vbg ha recibido mayor atención por parte de los medios del país. A través de Twitter se expresó un clamor público al respecto. En este sentido, las plataformas de redes sociales como Twitter han servido para erosionar las fronteras sociales, permitiendo el cuestionamiento de ciertas normas sociales y del status quo. Además, Twitter proporciona un espacio virtual seguro a través del cual las mujeres de India pueden expresar su opinión sobre la vbg y promover el cambio social. El presente artículo da cuenta de los hallazgos surgidos de una investigación que abordó cómo, tras dicha violación en Delhi, hombres y mujeres utilizaron Twitter. Dicha investigación se centró en los intercambios sobre la vbg y en la oportunidad que representa este medio para lograr que los individuos, en especial las mujeres, se involucren estratégicamente con el propósito de impulsar el cambio social.

KEYWORDS:

• Gender-based violence
• India
• social media
• Twitter
• content analysis

Notes on contributors

Tilly A. Gurman is a Senior Research and Evaluation Officer at the Johns Hopkins Center for Communication Programs, Bloomberg School of Public Health. Postal address: 111 Market Place, Suite 310, Baltimore, MD 21202, USA. Email: [email protected]

Catherine Nichols is a Data Analysis Advisor at the Office of HIV/AIDS, U.S. Agency for International Development. Email: [email protected]

Elyssa S. Greenberg is Senior Associate, Business Development & Project Management, East & Southern Africa at Chemonics International Inc. Email: [email protected]

Notes

1 At the time of the current study, tweets were limited to 140 characters. In November 2017, the limit was extended to 280 characters.

2 Hootsuite bought out uberVU in 2014.

3 A total of 11 coders engaged in four approximately three-hour training sessions about the definitions of each content category and the specific procedures for coding. Each session included instruction and discussion about the coding procedure along with time to practise applying the coding guidelines. In between training sessions, coders independently practised coding and sent their data to the study's first author, who identified discrepancies across coders. Coders practised applying the guidelines to tweets that came from a time period near but outside the specific date range for the current study. This approach provided GBV-related tweets from males and females in India very close to the time period, but did not affect the integrity of the actual data-set analysed. During three of the training sessions, the study team discussed discrepancies and questions about content categories. These discussions led to clarifications and modifications to the coding guide, based on group consensus. The last training session yielded a finalised coding scheme. At that point, coders received unique data-sets as well as specific instruction to refrain from discussing further coding decisions with each other. This instruction – considered an essential component of rigorous content analysis – ensured that coders did not make additional modifications to the coding guidelines and strengthened the validity of the ultimate data-set (Neuendorf 2008). The final data-set that each coder received comprised two types of data. First, each coder received a unique set of tweets that only he/she would code. The second component included a non-unique set of tweets (n = 130) that every coder would assess and would be later used to determine the inter-rater reliability. The reliability data-set represented a 5 per cent random subset of tweets extracted from the complete data-set. The sample size for the reliability data-set fell within a sufficient range for content analysis (Neuendorf 2008). Upon completion of data collection, the first author used ReCal3 online software to calculate two measures of inter-rater reliability – percentage agreement and Cohen's Kappa. Percentage agreement is the most widely accepted test of inter-rater reliability in content analysis, with 80 per cent agreement considered satisfactory (Neuendorf 2008). Although percentage agreement is often used alone in content analysis, it fails to consider chance as a factor in inter-rater agreement. Therefore, rigorous content analyses will also calculate a conservative measure of reliability that adjusts for the possibility of chance agreement across coders, such as Cohen's Kappa. With Cohen's Kappa, values greater than 0.75 reflect excellent agreement beyond chance and values between 0.40 and 0.75 reflect acceptable agreement beyond chance – especially in more exploratory content analyses on novel topics (Banerjee et al. 1999). Results from the reliability analysis conducted in this study classified the final coded data-set as satisfactory. Percentage agreement across the study variables ranged from 84.5 to 99.0 per cent, with an average of 92.3 per cent. Cohen's Kappa ranged from 0.26 to 0.98, with an average of 0.61.

4 Chi-square analysis determines whether the observed distribution of frequencies for one or more category of interest differs from the expected frequency distribution. Basically, the statistic assesses whether the observed difference in frequencies is due to chance alone. For more information, see https://www.youtube.com/watch?v=VskmMgXmkMQ or https://www.mathsisfun.com/data/chi-square-test.html (last checked by authors 2 May 2018).

5 The @mention tool is one in which a user tags another user in a tweet by using the other's username anywhere in the body of the tweet. It is considered a way to engage in two-way communication, since it connects the other user to the tweet thereby also connecting the tweet to users that are following the other user.

6 When there is little variation in coding and one category is over-represented in coding decisions, inter-rater reliability statistics that adjust for chance agreement can be substantially negatively impacted. In addition, over-representation of coding decisions – such as the case in the current study where jokes and personal/vicarious experience were hardly present – make calculating statistics like Cohen's Kappa unsolvable (Freelon 2010). Looking at the overall patterns of coding in the current study, this reality was likely the case. It is important to note that the percentage agreement across coders was generally satisfactory to excellent. Given the large number of coders, it is less likely that agreement occurred due to chance than if the study had used fewer coders (Potter and Levine-Donnerstein 1999). In addition, the current study applied a rigorous approach to content analysis, with a well-tested coding scheme, a thorough inter-rater reliability assessment, and no discussion between coders once training ended (Potter and Levine-Donnerstein 1999; Neuendorf 2008). This rigour, in the end, supports the validity and reliability of this study.