With a high degree of accuracy, artificial intelligence could be used to forecast life events by analyzing registration data on people’s domicile, education, income, health, and working conditions.
Image Credit: NicoElNino/Shutterstock.com
It is possible to use artificial intelligence created to simulate written language to forecast people’s future events. A collaborative research effort involving DTU, University of Copenhagen, ITU, and Northeastern University in the US reveals that by harnessing extensive datasets on individuals' lives and employing “transformer models”—similar to ChatGPT—specifically designed for language processing, it becomes possible to systematically structure the data.
Remarkably, these models can forecast future events in a person's life, including the potential estimation of their lifespan.
In the recently published scientific study in Nature Computational Science, scientists examined health information and labor market attachment for six million Danish dogs using a model called life2vec. The model has been demonstrated to outperform other sophisticated neural networks and predict outcomes like personality and time of death with high accuracy after it has been trained in the initial phase, i.e., learned the patterns in the data.
We used the model to address the fundamental question: to what extent can we predict events in your future based on conditions and events in your past? Scientifically, what is exciting for us is not so much the prediction itself, but the aspects of data that enable the model to provide such precise answers.
Sune Lehmann, Study First Author and Professor, Technical University of Denmark
Predictions of Time of Death
Life2vec’s forecasts provide generic responses to topics like “death within four years.” Upon scrutinizing the model's outputs, researchers find alignment with established conclusions in the social sciences. For instance, under similar circumstances, individuals in leadership roles or with substantial incomes tend to have higher survival rates.
Conversely, being male, possessing certain skills, or having a mental health diagnosis correlates with increased mortality risk. Life2vec utilizes a vast array of vectors to encode this data, employing a mathematical framework to organize various information. It strategically positions data points related to birth timing, education, income, housing, and health within the model.
Whats exciting is to consider human life as a long sequence of events, similar to how a sentence in a language consists of a series of words. This is usually the type of task for which transformer models in AI are used, but in our experiments we use them to analyze what we call life sequences, i.e., events that have happened in human life.
Sune Lehmann, Study First Author and Professor, Technical University of Denmark
Raising Ethical Questions
The life2vec model raises ethical concerns, according to the scientists, including privacy concerns, safeguarding sensitive data, and the influence of bias in data. Before the model can be used, for example, to determine a person’s risk of catching a disease or experiencing other avoidable life events, these challenges need to be better understood.
The model opens up important positive and negative perspectives to discuss and address politically. Similar technologies for predicting life events and human behavior are already used today inside tech companies that, for example, track our behavior on social networks, profile us extremely accurately, and use these profiles to predict our behavior and influence us. This discussion needs to be part of the democratic conversation so that we consider where technology is taking us and whether this is a development we want.
Sune Lehmann, Study First Author and Professor, Technical University of Denmark
The researchers suggest that the next stage would include additional kinds of data, such as text and photographs or details about the social networks. The social and health sciences can now interact in completely new ways because of this usage of data.