Preventing Suicide Risk with Artificial Intelligence

Experiments are carried out to detect suicide candidates using big data and artificial intelligence. The latest experiment of this kind comes from the United States, the universities of Tennessee and Florida.

Researchers have worked with an artificial intelligence program capable of accurately detecting at-risk patients in a given population. Result: AI could predict a suicide attempt two years in advance with 80-90% success rate. This rate would climb to 92% for a suicide that could occur within a week.

How to predict suicide?

The principle of artificial intelligence is to analyze a large amount of data – anonymous – in order to deduce rules. For this, an AI must train. US researchers have therefore submitted to the program the medical records of more than 5,000 people who had been admitted to the hospital for attempted suicide.

Among them, they distinguish

Experiments are carried out to detect suicide candidates using big data and artificial intelligence. The latest experiment of this kind comes from the United States, the universities of Tennessee and Florida.

Researchers have worked with an artificial intelligence program capable of accurately detecting at-risk patients in a given population. Result: AI could predict a suicide attempt two years in advance with 80-90% success rate. This rate would climb to 92% for a suicide that could occur within a week.

How to predict suicide?

The principle of artificial intelligence is to analyze a large amount of data – anonymous – in order to deduce rules. For this, an AI must train. US researchers have therefore submitted to the program the medical records of more than 5,000 people who had been admitted to the hospital for attempted suicide.

Among them, they distinguished more than 3,000 of them who had made a real attempt at suicide and 2,000 who had injured themselves but without going as far as the dramatic end.

The computer has molded these data (age, sex, addresses, drugs were taken and medical diagnostics …) and deduced some rules. However, caution must be exercised, and not all researchers have revealed this, but the researchers put forward sleep disorders as a precursor to suicide.

Avoiding suicides through social network analysis

The Tennessee and Florida experimentation is limited to persons who have been admitted to the medical community. To go further, it would be necessary to analyze profiles of people outside the medical framework.

This is what Cincinnati researchers have been doing since last year, based on totally different criteria. For their part, they have developed an algorithm that focuses on the tone of the voice and the words used (morbid vocabulary, the length of silences, etc.).

Finally, scientists would like to be able to identify suicide candidates on social networks in order to be able to intervene on time. Social networks constitute a formidable amount of information connected to the mood of the populations. Moreover, Facebook is working on the development of its own algorithm to detect the warning signs of suicide.