What if you could know the date of your death? A Danish study uses artificial intelligence to accurately predict certain aspects of human life, including when a person is likely to die. Explanations.
Imagine that Artificial Intelligence could predict your future and determine when you will die… This is what researchers wish to demonstrate in their research project “Using Sequences of Life-events to Predict Human Lives”. The goal ? Show that it is possible to predict the future using data from the past.
A correct algorithm 78% of the time
“With this data we can make any type of prediction” says Sune Lehmann, the lead author of the study. To carry it out, the researchers rely on data from the National Register of Denmark, which concerns 6 million Danes and provides information on many different subjects. The information is then integrated and processed by the Life2vec algorithm model which predicts certain aspects of a person’s life.
To test the quality of the algorithm, the researchers selected 100,000 individuals from the Danish registry. They are aged 35 to 65 because, according to researchers, mortality is more difficult to predict at this period of life. The information retrieved concerns education, health, income and occupation; and extend over a specific period: between January 1, 2008 and December 31, 2015. Among the participants, half survive after 2016 and the other die. However, only researchers have this information since the algorithm does not know who dies after 2016.
Life2vec then interprets past events to predict certain information about people’s future lives. He can then individually guess their way of thinking, feeling and behaving in their future life. It goes further: it even deduces the probability that a person will survive the four years following 2016.
According to the results and in comparison with reality, the algorithm is correct 78% of the time. Impressive ! Furthermore, according to the study report, Life2vec outperformed other algorithm models by at least 11% in predicting individual mortality more accurately.
Results consistent with those existing in the social sciences
According to the analyses, the mortality of the people studied is consistent with what the social sciences reveal. For example, experience shows that men would be more likely to die after 2016, as would those suffering from mental health problems like depression or anxiety. Conversely, those who occupy a management position or who have a high income tend to survive.
Further research is needed to complete the study
Although this study provides interesting results, it does have some limitations. First, the algorithm model can only predict the next four years. Life2vec is limited because it is currently only a research prototype. Future algorithms could then provide further details on the predictions.
The study could be based on data extended over a longer period. These could be more precise, in particular by taking into account essential data, for example, whether people take care of their health or not.
It is also important to note that this research was carried out in a developed country that has a strong healthcare system. The results cannot therefore be applied generally.
Finally, this study also raises ethical questions. Indeed, knowing the date of one’s death could lead to psychological problems. Some people think that removing this element of mystery would make life too predictable and take away some of its value.