Researchers at the Technion are developing methods for monitoring suicidal tendencies using artificial intelligence

In recent years, with the development of artificial intelligence (AI) and the proliferation of machine learning models, methods based on computational learning have been developed in suicide research. Because of this, it was decided in the Faculty of Data and Decision Sciences at the Technion to hold a datathon on the topic of developing tools for early monitoring of suicidal tendencies

Suicide is a tragic human phenomenon in which a person chooses to take his own life. Every 40 seconds on average a person in the world takes their own life. More than 500 Israelis commit suicide every year, and thousands more suicide attempts are recorded every year in Israel. Since 2003, the date September 10 (yesterday) has been marked worldwide as World Suicide Prevention Day. Studies show that in many cases it is possible to alleviate people's suffering and prevent suicide attempts by early identification of the suicidal tendency and the allocation of appropriate treatment. Many research efforts have been invested in understanding suicide and its characteristics, but many of the risk factors that may be used for early warning are still unclear. A meta-study conducted over five decades shows that when using traditional research methods to identify suicide risk, predicting suicides is only slightly better than random prediction.

In recent years, with the development of artificial intelligence (AI) and the proliferation of machine learning models, methods based on computational learning have been developed in suicide research. These methods have been found to be much more effective than traditional methods based on theoretical concepts. In view of these findings, the Faculty of Data and Decision Sciences at the Technion chose to hold a dataathon on the topic of developing tools for early monitoring of suicidal tendencies.

Datathon is an artificial intelligence development competition focused on huge data and its analysis. The Technion competition participants developed tools and applications for monitoring suicidal tendencies, based on a textual database collected in the research group of Prof. Roy Reichert from the faculty with researchers Shir Lisk and Ilanit Sobol. The datathon lasted for three consecutive days and was attended by about 70 students who were divided into 16 groups.

"Every year we hold an event that exposes our students to the real world, with real data and social challenges," he said Head of the Data and Information Engineering program at the Technion Prof. Avigdor Gal. "This time we dedicated the event to monitoring suicidal tendencies. This is the students' first experience in meaningful work with data, and we emphasize to them not only the technological aspects but also the ethical aspects of data responsibility."

in the first place The competition was won by the students Ziv Barzilai, Liad Domb, Omri Lazuber and Yonathan Voloch who used a data set that contained information about people who displayed suicidal tendencies on social networks. With the help of mentors from associations that deal with mental health and suicides, the group members developed a system that identifies suicidal tendencies among users and warns the people close to them about the fear of self-harm. According to them, "We hope that this system will help save human lives and raise awareness of the issue of suicide."

The second place was shared by two teams:

The first, in which students Idan Horowitz, Liane Fichman, Shir Gisler and Ariel Cohen are members, harnessed tools of machine learning and artificial intelligence in combination with the opinion of experts in the field of mental health. According to the members of the group, "the system we developed uses facial recognition models and tools from the field of natural language processing to analyze a video and speech of psychological therapy. In this way, it is possible to provide in real time or after treatment conclusions regarding mental illnesses or a tendency to suicidal thoughts, which will help the therapist make an accurate diagnosis about the patient."

The second, in which Ariel Novominsky, Vladislav Komentani and Alexander Freidin are members, proposed a method for quantitative measurement of the abstract concept "emotional process". The group members wanted to understand and characterize the emotional process that a person goes through in difficult moments. "We were looking for an approximation to the emotional state, which allows for a mathematical analysis of it. We used the texts we received as part of the data on suicidal people, and thus quantified the process in a way that will help in locating people who are at risk of self-harm."

In third place Eden Hindi, Leon Halika and Kafir Eliyahu won. Their goal was to identify suicidal symptoms in order to create a control group that would allow statistical comparisons to be made between people at risk and people not at risk, under the assumption that most users are not suicidal. They matched the control group to the treatment group according to the type of content and other parameters. Using this data, they built a classification model that makes it possible to determine if the user's content contains any signs of suicidal tendencies. Despite the small amount of data, their classifier produced promising results.

The dataathon, led by Dr. Gila Molcho, was accompanied by Jeremy Attia from Data for Good Israel and three graduate students in the faculty, who also served as mentors in the competition: Shir Lisk, Ilanit Sobol and Tom Yubiler. Lisk and Sobol are investigating identifying suicidal tendencies on social media as part of their thesis and PhD, and they allowed the students to use the data collected as part of their research.

Representatives of associations joined the competition as mentors

The Datathon was enthusiastically joined by associations that provided the students with the information and motivation to develop solutions that might save lives. Rabbi Shalom Hamer, founder of the "Darcha Shel Gila" association, gave the students a glimpse into the life of a suicidal person and the process of deterioration to suicide and allowed them to openly ask relevant questions. Mira Farbstein harnessed the ARAN organization and even donated her time to take part as a mentor and judge in the event. Dr. Shiri Daniels, a national professional director in Aran, gave a lecture to the students and Uri Rachman, a volunteer at the "For Life" association, served as a mentor and judge at the event.

"We have tools to try and help," said Dr. Daniels, "but even psychologists have difficulty identifying suicidal tendencies ahead of time. Relatives of people who tried to commit suicide sometimes say that there were early signs, but it was difficult to recognize them."

With the help of these organizations, two important goals were achieved: the students could approach this complex topic with maximum information and no less important - start a dialogue on this difficult topic in an open and sensitive way, be alert for signs from people suffering from suicidal tendencies and thus possibly save lives.

The event is supported by the organization IDSI (Israeli Data Science Organization funded by VT) led by Prof. Paul Feigin, the Social Incubator at the Technion led by Ronit Piso and Tech.AI - the Technion's umbrella organization for artificial intelligence.

More of the topic in Hayadan:

• Forecasting floods using machine learning with an accuracy of over 90%
• A computerized optical nose that can detect smells and turn them into an image on the computer screen
• save new brain cells
• The facial recognition systems fake in estimating the age of the person photographed