Predicting life expectancy or the future of an individual’s social relationships: that’s what an algorithm developed by Sun Lehmann of the Technical University of Denmark in Lyngby and described in the journal Nature Computational Science has been able to do. This artificial intelligence system learned to make predictions about individuals after being trained on data from 6 million people, turning it into a useful tool for predicting human behavior, but it also raises new ethical questions.
“This is a very interesting development, an application in an important context, of one of the things that artificial intelligence does best, namely prediction, analyzing large amounts of data, and identifying patterns and repetitions,” a legal expert on robotics and humans told ANSA. -machine interaction by Andrea Bertolini, from the Scuola Superiore Sant’Anna in Pisa. In recent years, artificial intelligence has shown how it is possible, with the help of sophisticated algorithms, to almost perfectly imitate some of our abilities that until recently were considered a human prerogative, such as writing, language, or artistic expression. Now, using the same algorithms that are used to emulate human language, Danish scientists have tried to test whether artificial intelligence can also predict some aspects of human life.
Just as relationships between words are encoded in language, researchers have encoded relationships between aspects of daily life related to health, income, or place of residence. From data concerning 6 million Danes, the algorithm tried to predict some future aspects, mainly early mortality, especially survival longer than 4 years for people aged 35 to 65, and to determine some personality characteristics associated with the social sphere, such as the duration of romantic relationships.
“At the level of pure scientific research – added Bertolini – the work of Danish researchers does not represent a technological paradigm shift, but rather an application in a new sector of traditional linguistic models that we are now learning to know in many fields.”.
In addition to the results, which the researchers say were fairly accurate, the study primarily highlights the need to openly discuss the real-world use of these types of apps and the potential impacts on individual rights. Many applications can be imagined for a forecasting model of this type, such as the possibility of having a kind of personalized oracle (ethically no different from tarot forecasts, but scientifically accurate), more reliable tools for planning retirement expenses, or for determining the value of buying a freehold property. “From a legal point of view – the Italian researcher concluded – it will be important to really evaluate the predictive capacity of these models and then define more precisely the context of use, of course, it will also be important to define how and what personal data could be used”.