Skip to main content
DA / EN
RESEARCH

New AI technology can detect suicidal thoughts on social media

Researchers can now predict suicidal thoughts based on social media behaviour using artificial intelligence. The hope is that this technology can help save lives.

By Sune Holst, , 8/20/2024

"Social media has become an important place for people to express their innermost thoughts and feelings anonymously. We can use AI to analyze these expressions in real-time and detect potential suicidal thoughts before they develop into tragic actions."

These are the words of Serkan Ayvaz, an associate professor at the SDU Center for Industrial Software (CIS). Together with Mohamed A. Allayla from the Department of Computer Engineering at Yildiz Technical University in Turkey, he published a scientific article, where the researchers present an advanced big data analysis tool that uses social media platforms as an early indicator of suicidal thoughts.

A massive global tragedy
Every year, more than 700,000 people take their own lives, according to the World Health Organization (WHO). This means that a suicide occurs every 40 seconds. WHO emphasizes that suicide can be prevented.

By analyzing social media, researchers can now identify when a person begins to show signs of suicidal thoughts—and thus intervene before it is too late. The researchers' big data analysis tool can become a vital resource here.

"The unique aspect of our system is its ability to operate in real-time," explains Serkan Ayvaz.

"We have created a platform that can continuously collect and analyze data, and when the model detects a post with potential suicidal thoughts, it is visualized in the system so that relevant authorities or organizations can take action."

The technology utilizes advanced AI algorithms, including natural language processing (NLP), to analyze text content from social media platforms like Twitter and Reddit.

Using a large dataset containing both suicidal thoughts and regular posts, the researchers have trained AI models to distinguish between these two types of content with high accuracy.

Potential for widespread application
The researchers behind the technology have two specific suggestions for how it can be beneficial.

"Healthcare professionals and authorities can use the system to identify and assist individuals at risk, but we can also create a system where people expressing a need for help are automatically offered resources and contacts for support. This could be through online forums or organizations like Livslinjen."

Ethical considerations and the future
Regarding ethical considerations, Serkan Ayvaz explains that users' privacy has been taken into account.

"Our system only uses publicly available data, and we do not target specific individuals. It is important that any future use of this technology also considers ethical aspects."

The researchers now plan to extend the framework to support multiple languages, improve the model to achieve even higher performance, and develop interfaces that non-technical end users can easily interact with.

Editing was completed: 20.08.2024