Sign In
International
Media release
From:
Computing: Machine learning approach may predict aspects of human life
A machine learning approach that may accurately predict different aspects of human lives, including the likelihood of early mortality and personality nuances, is described in a study published in Nature Computational Science. The model may be able to provide a quantitative understanding of human behaviour.
The question of whether human lives are predictable has been debated by social scientists. While the socio-demographic factors that play an important role in human lives are well understood, accurately predicting life outcomes has not been possible.
Using data on education, health, income, occupation and other life-events for around 6 million people from a national register in Denmark, Sune Lehmann and colleagues designed a machine learning approach to construct individual human life trajectories. The authors adapted language processing techniques and represented human lives in a way that was similar to language within the model. This approach allowed them to generate a vocabulary for life events in a similar way to how language models capture complex relationships between words. The proposed model — called ‘life2vec’ — establishes complex relationships between concepts such as health-related diagnoses and place of residence to income levels, encoding individual lives with a compact vector representation that forms the foundation for the prediction of life outcomes. The authors demonstrate that the model can predict early mortality (specifically, the likelihood of individuals from their cohort in the age range 35–65 years surviving the 4 years following 1 January 2016) and capture personality nuances better then state-of-the-art models and baselines, outperforming them by at least 11%.
The findings demonstrate that, by representing the complex linkage between social and health outcomes, accurately predicting life outcomes may be possible. However, the authors stress that their research is an exploration of what may be possible, and it should only be used in real-world conditions under regulations that protect individual rights.
