AI World Models for Memory, Meaning, and Human Futures

  • Juwon Jung, Microsoft Research

Interactive world models are emerging as a new design material, enabling people to become creators by transforming their personal data into explorable visual worlds. Moving beyond traditional world-model applications in gaming, robotics, and autonomous-driving simulations, this project explored how these technologies could enable personally meaningful, everyday experiences. It examined this question through the lens of ‘memory experiences’ – an inherently relational domain in which people preserve, interpret, and pass experiences among families and friends.

Across two rounds of focus groups with 17 participants recruited from four relational cohorts, the research explored enduring values around remembering. Participants also imagined future memory interactions using world-model capability cards and collaborative creation scenarios. They valued intergenerational discovery and legacy, but resisted literal reconstructions of people or memories. Acceptance of AI in both the process and outcomes of world-model creation was found to depend on clear communications of uncertainty, making AI contributions and their provenance visible, and preserving human judgement and agency.

These insights were brought into a series of short video envisionments in which a grandfather and his family co-create an evolving digital legacy world. The research contributes design principles for meaningful, human-centred AI and world-model experiences. It argues that the promise of world-models (and other generative AI) lies not only in what they can generate, but also in how the (co-)creation process is configured to help people reflect, connect, and make meaning together.

Contact the researcher: Juwon Jung
Email: juwonjung.design@gmail.com
LinkedIn: www.linkedin.com/in/juwon-jung-057a731b3 (opens in new tab)

Watch Next