When mental health moves through social media
Institutional Communication Service
23 January 2023
Can a Tweet or a post on Instagram tell us something about our mental health? How meaningful is the relationship between well-being and the use of these platforms? To answer these questions, Marta Fadda, a researcher in bioethics at Università della Svizzera italiana, Oliver Grübner, a health geographer, and Marcus Wolf, a psychologist, at the University of Zurich conducted a series of studies to demonstrate the usefulness of social platforms for mental health research. The team of three researchers won the first UZH Postdoc Team Award. This interdisciplinary research investigates the links between social media posts and the population's mental health. Dr Marta Fadda tells us about it.
What does the research consist of? What tools does it use?
The research analyses many tweets for which the user's geographic location was available when sharing, using innovative emotion recognition tools. As you can imagine, the extraction, storage, analysis, and sharing of the data and results of the various studies we conducted have important ethical and legal implications. For example, users do not provide informed consent for the analysis of their tweets for specific research purposes, as is usually the case in other studies, and the preservation of these tweets could expose users to risks concerning their privacy. I was in charge of analysing the ethical and legal issues related to the various studies we conducted and coordinating the team in identifying and implementing ethically and legally justifiable solutions. In a nutshell, I asked myself whether it was ethical and legal to implement the technologies available to answer our research questions with the type of data collected.
How do you determine a person's mental health from what they write on social media?
Our emotions are indicators of our mental health status. The analysis we used consists of EMOTIVE software that detects and measures emotions in posts on social platforms. This software extracts eight emotions: anger, disgust, fear, happiness, sadness, surprise, shame, and confusion. Studying the distribution of these emotions over time and space provides valuable information concerning how groups of people react to certain events.
What practical implications does this research have on the ground?
For example, Dr Grübner found that the adverse emotional reactions identified in the tweets of people in New York City during and after Hurricane Sandy were concentrated in specific neighbourhoods of the city, particularly Staten Island. This information makes it possible to plan and implement targeted interventions to address emergencies where they hit hardest.
The fact that we have cutting-edge technologies available today does not necessarily mean that their use is ethical or legal. So what are the ethical and legal implications of these studies?
The fact that people freely share their opinions, experiences, and emotions on social platforms does not mean that research workers (or others) can take hold of this information and use it as they see fit. Users who generate content online, even on entirely open platforms that anyone can access, have rights with respect to their privacy. Imagine yourself walking down the aisle of a supermarket with your shopping cart. While you may expect to encounter people and that they may take a look at your groceries, you would probably be annoyed to discover that someone has been following you from the moment you enter to the moment you leave the store and has been recording all your movements, your purchases, and your exchanges with the sales staff. In addition, the analysis of large amounts of posts, if not conducted scientifically robustly, can generate erroneous conclusions that can lead to the spread of stereotypes and negative judgments among the population.
How have you addressed these issues?
The ethicality and legality of our studies have always been at the forefront. In addition to engaging within the team, we confronted the scientific literature that addressed similar studies' ethical and legal challenges. We interfaced with various institutions, including data protection authorities at the cantonal level. However, we realised that the available guidelines did not consider a unique aspect of our data: the geographic location of the collected posts. So we developed new guidelines that would take this additional aspect into account. For example, there are critical privacy issues when considering that the tweet we analyse contains the person's exact location at the time the tweet was shared. One of our recommendations was to modify the geographic location data and extend it to a larger area so that the possibility of precisely locating the person would be more limited.
Fadda M, Sykora M, Elayan S, Puhan MA, Naslund JA, Mooney SJ, Albanese E, Morese R, Gruebner O. Ethical issues of collecting, storing, and analyzing geo-referenced tweets for mental health research. Digit Health. 2022 Apr 12;8:20552076221092539. doi: 10.1177/20552076221092539. PMID: 35433020; PMCID: PMC9008807. (https://pubmed.ncbi.nlm.nih.gov/35433020/)