Could AI Save Lives? AI for Suicide Prevention.
According to the Office for National Statistics, there were 5,691 suicides in the United Kingdom in 2019, with 11 people committing suicide per 100,000 population last year. With the Covid-19 pandemic leaving people feeling more isolated than ever before, and record levels of unemployment, this number is only expected to rise. It is therefore imperative that action is taken to identify suicidal individuals; to prevent them from committing suicide. Often, suicidal individuals may not confide in people, however, they may share content on the internet indicative of suicidal intent. While human moderators could be used to identify such content, they will, inevitably, miss cases. This is where AI comes in.
In 2017, following a number of incidents in which people committed suicide on live stream, Facebook started the “proactive detection” initiative. Facebook began using Artificial Intelligence to identify posts by users who may be feeling suicidal, allowing humans to take actions to prevent the suicide. Posts are ranked on a percentage scale of the probability of “imminent harm”, with 100% being the most likely. Comments on the post were also ranked, with content such as “has anybody heard from them” being seen as more urgent than “call me any time”. Speaking about the decision, CEO Mark Zuckerberg said: “Moving from reactive to proactive handling of content at scale has only started to become possible recently because of advances in artificial intelligence” and “In the last year, we've helped first responders quickly reach around 3,500 people globally who needed help”. However, critics have raised questions about the ethics of collecting and storing an individual’s mental health data, considering it excessive, as well as concerns about the legality; some claim that it breaches data protection regulations. These claims of illegality have led to the use of the “suicide algorithm” being banned in the European Union under the GDPR. Clearly, there is a careful balance between storing data to make the AI more effective at preventing suicide, and not storing too much data in order to protect a user’s privacy. Nevertheless, with regards to suicide prevention, the algorithm has been successful.
Facebook live lets you live-stream yourself to people all over the world.
Credit: CNET
The “Tree Hole Rescue Team” was started by Huang Zhisheng, an AI researcher at the University of Amsterdam. Zhisheng wanted to create an AI which could detect posts by suicidal individuals, so they can be actioned by the suicidal individual’s local authority, therefore saving lives. The AI scours “tree holes” (places where people post secrets) on the Chinese social media app Weibo looking for individuals likely to be in immediate danger of suicide, and sends an alert signal to a team of hundreds of volunteers, who then track down the individual and attempt to talk them out of it, while forwarding on information to the relevant authorities if needed. In an interview with BBC, Zhisheng spoke of how the AI has managed to save a man who had attempted to jump off a bridge and a 22 year old college student who planned to kill themselves in two days. Due to resource limitations, the AI can only read around 3,000 posts a day, but even with such a small sample of posts, the initiative has been effective, and has helped to save lives.
It’s not just on the front line that AI has been used prevent suicide, but AI has also seen increasing use more broadly for mental health purposes. A number of apps have been developed to support people’s mental well-being, therefore indirectly reducing the likelihood of suicide. The app Lantern uses AI to create a customised user experience, where they can receive help online with their mental health. Quartet Health uses machine learning technology to offer recommendations for those suffering from mental health problems. As mental health has been at the forefront of societal discussion in recent years, we may see more and more AI-powered assistance to those with mental health conditions.
It seems as if there is a place for AI in suicide prevention: with the capability to identify and respond to incidents faster than any human could, AI has undoubtedly saved lives. Of course, it is not the only respondent, a human aspect is required to make the correct inferences as to the severity of the situation and the best action to take. However, AI is an excellent indicator.
Thumbnail Credit: Ulrike Mai