A group of researchers, psychologists and psychiatrists have developed Artificial Intelligence algorithms that, by analyzing texts, images and activities on the social network Twitter, detect patterns of suicidal behavior with 85% accuracy.
This is the STOP (Suicide Prevention in Social Platforms) project, whose objective is to search for and analyze patterns of suicidal behavior by applying Artificial Intelligence in social networks.
The project is directed by Ana Freire, a researcher at the Department of Information and Communication Technologies at the Pompeu Fabra University (UPF) in Barcelona, and has had the collaboration of the UAB Computer Vision Center and the Parc Hospital Taulí de Sabadell, in Catalonia (northeast).
The researchers highlight that in Spain, where there are about 3,000 suicide victims every year, “the taboo associated with this phenomenon, the scarce education in mental health and the difficult access, sometimes, to psychological consultations means that people with mental problems do not receive neither a diagnosis nor an adequate treatment “.
The World Health Organization has calculated that each suicide has an emotional impact on at least six people around the victim.
In this scenario, according to the researchers, who have published their contribution based on artificial intelligence techniques in the journal ‘Journal of Medical Internet Research’, “social networks have been shown to be an effective means of detecting problems such as depression or disorders of eating behavior that, in very extreme cases, can generate suicidal ideas “.
According to the researchers, around 8,000 tweets are published on Twitter per second, “which contain very valuable information for various fields, but also for analyzing issues related to mental health.”
“In our case, we train artificial intelligence algorithms so that they can distinguish patterns of high risk and low risk of suicide, with data labeled by experts in mental health and completely anonymous, to respect the privacy of users”, they explain in their work , which is the first to address this problem by analyzing texts in Spanish while taking into account the history of publications (tweets) of each user.
The technique also generates a comprehensive methodology for collecting suicide data and analyzes images and texts.
According to the researchers, “the main contribution of this work is that for the development of the models images are explored, together with aspects that are normally taken into account by specialists in diagnosis such as interactions between users, analysis of sleep patterns and the existence of risk factors “.
“This work has allowed us to learn differential characteristics between the groups of ‘high risk’ of suicide and ‘free of risk’ and to see that the first group tends to speak more in the first person and to use denials and terms related to feelings, among which highlights anxiety “, they detail.
They have also observed that they tend to have fewer friends (accounts that they follow), write texts with fewer characters and are more active on weekends and at night.
Likewise, they have been able to demonstrate “that there may be a certain correlation between the content of the images shared on social networks with the mental health of the user who shares them,” according to Jordi González, researcher at the Center for Computer Vision (UAB), who has participated with his team in the project.
Ricardo Baeza-Yates (UPF) has highlighted the importance of algorithms such as the one now published “to find in social networks new factors derived from the use of digital media that can help an effective diagnosis and contribute to making suicide stop being a taboo subject in our society”.
As a project, STOP has a crowdfunding initiative to expand research to other mental health problems, such as eating disorders.