University of São Paulo: Affective computing: system created at USP identifies signs of depression on social networks
Therapists, psychologists and psychiatrists will be able to count on a new tool to assist in the treatment of their patients: a system that uses Artificial Intelligence to identify signs of depression on social networks. The technology, developed during the doctoral research carried out by Felipe Giuntini, at the Institute of Mathematics and Computer Science (ICMC) at USP, in São Carlos, is able to assess the mental health of users through the interpretation of texts and emoticons that are published on the platforms.
The new system is composed of a series of computer codes (algorithms) that were tested to analyze the behavior of 415,000 users of the social network Reddit. This step served to “train” the technology to determine if a person exhibits depressive behavior. Post histories of up to ten years were evaluated, allowing for a robust preparation of the algorithms, which can be applied to any social network.
The main results obtained in the analyzes were the following: users with depression tend to get closer to those who have the same problem, commenting and interacting, for example, with posts with negative content. The system also identified, from the words and expressions that were shared, the feelings most present among depressives, which were: shame, guilt, sadness and nervousness. Finally, it was possible to observe that users with the disorder spent at least three days without posting on social media after having shared some content that showed feelings of guilt.
“The constant growth in the use and sharing of data on social networks has provided opportunities for the development of intelligent solutions capable of understanding human behavior online, as users share social aspects, feelings and opinions on a daily basis, facilitating studies in the area. known as affective computing”
Unlike the methods known in the scientific literature, which focus only on the classification of isolated posts to determine whether they are depressive, the system created by Giuntini allows the observation of a set of publications over time, which is essential for professionals to health monitor the evolution of their patients during treatment, checking for possible changes in mood. In addition, the new algorithms developed at USP also consider, in the analyses, aspects of the user’s personality, as well as the context of the publication, to classify their emotional characteristics, allowing for a more accurate assessment of the published content. The platform is also able to predict with 83% accuracy what will be the behavior trend of people in future posts.
“The constant growth in the use and sharing of data on social networks has provided opportunities for the development of intelligent solutions capable of understanding human behavior online, as users share social aspects, feelings and opinions on a daily basis, facilitating studies in the area. known as affective computing”, explains the researcher, who was supervised by professor Jó Ueyama, from the ICMC. According to the institute’s scientists, the new system could be used as a cell phone application or as a computer program. They are open to possible partnerships for the technology to be implemented in the market.
The research An approach to sequential evaluation of emotional behaviors of depressive users on social networks in groups and individually is available for consultation at this link .
The work was funded by the National Council for Scientific and Technological Development (CNPq), the Coordination for the Improvement of Higher Education Personnel (Capes), the Foundation for Research Support of the State of São Paulo (Fapesp), through the Center for Mathematical Sciences Applied to Industry (CeMEAI), headquartered at ICMC, and Sidia Institute of Science and Technology. The study also had the collaboration of scientists from the Psychology Department of the Federal University of São Carlos (UFSCar), Dartmouth College, in the USA, in addition to professors Agma Traina and Mirela Cazzolato, both from ICMC.
According to the last National Health Survey (PNS), carried out in 2019 by the Brazilian Institute of Geography and Statistics (IBGE), Brazil has 16.3 million people with depression. The number represents an increase of about 34% compared to the previous survey, produced in 2013. The age group with the highest proportion of confirmed diagnoses was 60 to 64 years old, while the lowest percentage was obtained in 18 to 29 years old. age.
Another survey, published earlier this year and which included the participation of researchers from the School of Arts, Sciences and Humanities (EACH) at USP, in São Paulo, evaluated, between June and August 2020, anxiety rates and population depression in 11 countries: Brazil, Bulgaria, China, India, Ireland, Northern Macedonia, Malaysia, Singapore, Spain, Turkey and the United States. The objective was to analyze how mobility restrictions imposed by the covid-19 pandemic were impacting people’s mental health. The results showed that the Brazilian population had the highest number of confirmed cases among the countries participating in the study. In all, 63% of the inhabitants interviewed had anxiety records, while 59% had depression.