Experiments have shown that artificial intelligence can give us insight from brain scans
Scientific research involving artificial intelligence has made serious progress in helping us understand brain activity, from our use of language to elements of mental health. Now, AI is able to spot suicidal thoughts and tendencies just from brain scans.
In a recent study, published in the Nature Human Behaviour Journal, two groups of young adults (with a mean age of 22) – one with suicidal thoughts, one without – had their brain activity observed in an fMRI scan. The volunteers were tasked with thinking about 30 words, some were positive (“praise”), some negative (“trouble”) and the rest were related to death (“suicide”).
An algorithm was then trained using the ‘neural signatures’ of the patients and tasked with identifying, from scan results, whether someone had experienced suicidal thoughts. Incredibly, the program identified individuals with 91 per cent accuracy and went further in determining whether someone had attempted suicide with 94 per cent accuracy. However, the study itself is limited by a small sample size (34 people), the sheer expense of fMRI scans and the fact that any clinical test would need to be perfectly accurate.
Almost 800,000 people die due to suicide each year, and the signs leading up to it can be hard to catch and predict. This research adds to the school of thought that says the biological signs are really there for people thinking of suicide.
The current method for identifying suicidal thoughts is to, of course, ask. But this method is flawed: another study suggests that 78 per cent of those who die by suicide deny having these thoughts in their last contact with a psychiatrist. We need only consider how taboo talking about suicide still seems to see how someone might be ashamed or in denial about their thoughts. More so, would the introduction of AI to pinpoint these thoughts take us further away from the emotional – and human – support people suffering with mental health still need?
This new research provides hope for a medical test to identify and help protect vulnerable people in future. The lead author of the study, Marcel Just, has profound ambitions: “This isn't a wild pie in the sky idea,” since, “We can use machine learning to figure out the physicality of thought. We can help people,” he told the Daily Beast. Just however outlined that it’s not altogether a practical solution, given the immense cost these scans are – he hopes to conduct the study again with equipment that provides a much cheaper testing option. He adds that more research needs to be done to see if the AI can accurately predict suicide attempts too.
As Wired reports, research in Florida State and Vanderbilt have created and trained an algorithm that correctly predicts future suicide attempts 85 per cent of the time, working off of medical records of those who had in the last two decades. Instead of brain activity, it detects via age, sex and medical history.
While researchers work to expand the use of AI in healthcare, it’s growing exponentially in other industries. Sophia the Robot became the first to receive citizenship recently in Saudi Arabia, the art world has seen some curious works created by AI, and – worryingly – autonomous lethal weapons could be in our future. Elon Musk continues to assert that AI and robotics could cause existential risk to humans too.
Social networks have also stepped up in recent times. Facebook monitors wall posts and live videos to pick up on anything that indicates self harm. Relying on users, Instagram launched a suicide prevention tool that allows people to anonymously flag posts about self-harm and suicide.