An artificial intelligence predicts with 78% accuracy the death of a person in the next four years
Artificial intelligence is revolutionizing all spheres of society and in the field of health, although with caution, its deployment opens a range of possibilities in optimizing hospital management and even in the diagnosis and detection of diseases. But a group of researchers from the Technical University of Denmark (DTU) and Northeastern University (USA) go one step further and develop an AI capable of predicting what will happen in a person’s life and even estimating the moment of his death.
It seems like science fiction, but they are the conclusions drawn from the study ‘Using sequences of life-events to predict human lives’, published this week in ‘Nature Computational Science’. By analyzing the sociodemographic data of six million people, from the Danish center ‘Statistics Denmark’ and the National Patient Registry (LPR), the artificial intelligence model called ‘life2vec’ can predict the death of a person with 78 percent accuracy. person in the next four years.
It does this by examining large amounts of data about people’s lives, such as schooling, education, salary, housing and health. The data set contains, for example, records of visits to health professionals or hospitals, as well as patient type, diagnosis and degree of urgency. The deep learning model, used in the field of natural language processing, such as ChatGPT, systematically organizes this amalgam of data and predicts events in people’s lives.
“From a scientific point of view, what is exciting for us is not so much the prediction itself, but the aspects of the data that allow the model to give such precise answers,” says lead researcher Sune Lehmann, a professor at DTU.
Individuals at risk
The database has information that covers from 2008 to 2020, although ‘life2vec’ uses data from the period 2008–2016 to be able to predict situations for the following four years and then check if it was correct in its forecast. In the study, the researchers organized people into pairs, between a group of between 35 and 65 years old, and to the question of “will he die in four years?”, the model was correct in 78 percent of the cases, 11 percent more than similar models.
According to the scientists involved, the results agree with existing findings in the social sciences. For example, all things being equal, individuals in leadership positions or with high incomes are more likely to survive, while having a mental illness, being involved in manual labor, or generally being a man is associated with a higher risk. die. This opens a new interaction between social and health sciences.
Ethical debate
This type of predictive models also raises ethical debates. The researchers themselves are open to these approaches, such as the protection of sensitive data, privacy and the role of biases in data. The leader of the study, Sune Lehmann, believes that “this debate must be part of the democratic conversation so that we consider where technology is taking us and whether it is a development we want.”
In the words of the researcher, “the model opens up important positive and negative perspectives to debate and address politically. “Similar technologies for predicting life events and human behavior are already used today by technology companies that, for example, track our behavior on social networks, create extremely precise profiles of us and use them to predict our behavior and influence us.”