Using Sequences of Life Events to Predict Human Lives

This piece, by Onno Berkan, was published on 12/10/24. The original text, by Savcisens et al., was published by Nature Computational Science on 12/18/23.

This University of Denmark research briefing, based on Savcisens et al., introduces a new way to understand and predict human life outcomes using detailed data from Denmark's national records. The researchers developed an AI system called "life2vec" that analyzes day-by-day information about people's lives, including health records, education, jobs, income, and where they live. What a time to be Danish.

The study used comprehensive data covering approximately six million Danes over an eight-year period (2008-2016). What makes this study unique is its approach to treating human lives like a language that AI can analyze. Just as AI systems can understand patterns in written text, this system learns patterns in how people's lives unfold.

The researchers converted life events into a special "synthetic language" where each event becomes part of a life sequence. We won’t go into too much detail. Still, the system could understand statements like "In September 2012, someone received twenty thousand Danish kroner as a guard at a castle" and convert them into this synthetic language. This allows the AI to process complex life information in a structured way.

The system combines both health and labor records, creating a comprehensive view of each person's life journey. The model can also predict various life outcomes, including mortality risk, based on many variables such as age, education, income levels, and more. By doing so, the system organizes different aspects of life (health, jobs, location, etc.) to reveal how these elements relate to each other. For example, the system found that being male, working class, or having mental health diagnoses correlates with a higher mortality risk.

The potential applications of this research are significant. The system opens new possibilities for understanding how different aspects of life are connected and could help in:

• Understanding how social factors impact health outcomes

• Exploring connections between various life events

• Developing more personalized interventions

However, the researchers emphasize that the model is currently a research prototype and is not ready for real-world deployment. Before any practical use, it would need careful auditing to ensure fairness and explainability, and its accuracy would need to be improved.

Looking to the future, this framework could incorporate additional types of information, such as online behavior and social relationships, potentially leading to an even more comprehensive understanding of human lives. The researchers suggest this could create new opportunities for collaboration between social and health sciences, offering a more complete picture of how different aspects of our lives interact and influence each other.

Want to submit a piece? Or trying to write a piece and struggling? Check out the guides here!

Thank you for reading. Reminder: Byte Sized is open to everyone! Feel free to submit your piece. Please read the guides first though.

All submissions to berkan@usc.edu with the header “Byte Sized Submission” in Word Doc format please. Thank you!