Can Analysis of Tweets Inform Interventions to Prevent Gender-Based Violence?
The Internet and its increasing availability, even in low- and medium-income countries, has given rise to the use of social media to better understand a broad array of health issues. MEASURE Evaluation, funded by the United States Agency for International Development (USAID), recently explored the feasibility of using large social media datasets to track changes in attitudes on gender norms regarding sexual relationships between younger women and older men and attitudes on gender-based violence (GBV) against women and girls.
The study encompassed 10 sub-Saharan African countries (Kenya, Lesotho, Malawi, Mozambique, South Africa, e-Swatini [Swaziland], Tanzania, Uganda, Zambia, and Zimbabwe), all of which have programs under the umbrella of USAID’s DREAMS program, which focuses on adolescent girls and young women (AGYW). DREAMS, which stands for Determined, Resilient, Empowered, AIDS-free, Mentored, and Safe, aims to curtail HIV transmission among young women in the countries with the highest HIV burden.
Estimates are that from one-third to two-thirds (28 percent to 60 percent) of women ages 15–49 years in sub-Saharan Africa have experienced physical violence. However, reliable and valid data on GBV are hard to find, particularly in low-resource settings, due to the lack of an agreed-upon definition of GBV, methodological challenges to measuring frequency and duration of GBV, and under- or nonreporting because of stigma. Gender norms are a frontier for social media research. Although many national and international programs seek to change harmful gender norms to achieve equities in health, few collect data on social attitudes.
Could social media postings help us understand harmful gender norms? MEASURE Evaluation explored if social media might offer an alternative strategy for data collection. Our rationale was that social media platforms could be useful because of the velocity, volume, and variety of data that people share about their lives and behaviors—information difficult and costly to obtain through surveys. For our study, we considered Facebook, Snapchat, Instagram, and Twitter. We selected Twitter for its data availability and ease of access.
The data were collected twice over two years, from January 2016 to January 2018, and analyzed using quantitative content analysis. We found that social media data can be useful. We obtained a picture of the conversations around popular hashtags, discovered how hashtags change (quickly), and—because the data are 100 percent user-generated—we knew there was minimal researcher bias in the data collected. We also saw how news articles and “viral” events, like the 16 days of Activism against Gender-Based Violence campaign, influenced the online discussion.
Our data collection timing in 2017 overlapped with this campaign, sponsored by the United Nations each year from November 25–December 10, to focus on violence against women and girls. During this period, our dataset showed a marked decrease in sentiment defending GBV and a spike in negative comments decrying such violence. However, shortly after the campaign, the sentiments expressed about GBV returned to the original, less impassioned attitude. The lesson we learned is that “hashtag advocacy” can be fleeting in its success at changing opinion and that follow-on work is needed to fan the flames of change that may begin during social media campaigns.
That said, though the data can be useful, we found multiple challenges using Twitter to seek understanding of GBV and gender norms—which are widespread social phenomena that are by no means limited to the millennial age group that typically frequents the platform. For example, while our search encompassed 10 countries, most Twitter traffic in the dataset came from South Africa in the areas around Johannesburg and Cape Town—areas more affluent and better-connected to the Internet than rural areas. Another challenge was that—because of the affluent profile of most Twitter users—rural and poorer communities were under-represented and so the findings are not generalizable to the larger population. A third challenge was that data collection was dependent on searching specific terms and hashtags, but those vary from place to place and change fluidly.
An overarching and ever-evolving challenge was the ethics of using social media for data collection. Privacy policies for individual social media platforms can change at any time. While international ethical standards exist for collecting and using data on sensitive subjects, such as GBV, how those translate to online data that are “public” remains unclear. In our report, Using Social Media Data to Understand Changes in Gender Norms, and accompanying guidance, we discuss possible ways of ensuring proper handling and reporting of these types of data.
Where does this leave us?
For those implementing and researching social and behavioral change communication interventions among certain populations that use social media, Twitter may be an informative place to start. It can help to identify the current conversation on a topic, so that an intervention reflects current terminology and norms. Understanding how social media is used to describe and communicate social experiences, such as GBV, can help program managers tap into these trends. Effective use of this data in development policymaking and advocacy could improve the lives of women and men, boys and girls by resulting in more efficient services and programs.
For more information on MEASURE Evaluation’s work on gender, visit https://www.measureevaluation.org/our-work/gender, or visit https://www.measureevaluation.org/our-work/evaluation to learn more about innovative evaluation approaches the project uses to understand important public health issues and solutions.
