Apps, Gadgets & Technology, News

App & Artificial Intelligence Predicts Suicide Risk

App & Artificial Intelligence Predicts Suicide Risk
About Author
Christian Boas

Researchers at Florida State University have developed an algorithm that predicts thousands of patient records with an accuracy of 92 percent, whether a person will commit suicide in the next two weeks after the investigation. This method can also be applied to the reading of smartphone data.

suicide plans
The problem is that suicide plans are a very complex phenomenon that is composed of many different events that a person has experienced. However, an artificial intelligence (AI) can understand this complexity, and that very quickly. For this purpose, different factors must be linked very precisely.

This process exceeds the abilities of a human brain so that AI can always be a step ahead. The researcher psychologist Jessica Ribeiro and her team have now succeeded in predicting whether someone will commit suicide within the next two years using an AI at an accuracy of 80 percent.

A forecast for the next two weeks showed an accuracy of 92 percent. These exact results could be achieved by machine learning. An algorithm scanned the patient records of 3,200 people who had committed suicide. This algorithm is therefore able to learn which factors are a significant signal for a high probability of an imminent suicide.

factors
The used patient records spread over many years and even decades, which is why the algorithm perceives even bizarre factors that might indicate a suicide. These factors escape human thinking in many cases, since the direct comparison with thousands of data can not be carried out.

“As people, we want to understand what we are looking for. But this is just the same as asking which brush stroke is the most important in a painting, “Ribeiro explains the complexity of the algorithm.

However, it is difficult that not every patient has good records of doctors and hospitals. Therefore, another artificial intelligence is to analyze additional smartphone data.

“We do not listen to the phone calls, but look at how many calls are made and how many are missed. We use thousands of data sets and combine them in a way that people are not capable of, “the researchers said.

With the app “Companion” under development, this should be possible. The data of the app can be passed on to a doctor in the form of a risk list in case of an alarming deviation. The experts also praise the approaches of Facebook to quickly identify promises of suicide.

To Top