Bayesian Artefact Design

  • Oulasvirta, Antti (Principal investigator)
  • Laine, Markku (Project Member)
  • Dayama, Niraj (Project Member)
  • Liao, Yi-Chi (Project Member)
  • Leiva, Luis (Project Member)
  • Nioche, Aurélien (Project Member)
  • Shin, Joongi (Project Member)
  • Todi, Kashyap (Project Member)
  • Peng, Zhenhui (Project Member)
  • Peng, Zhenhui (Project Member)
  • Nioche, Aurélien (Project Member)

Project Details

Description

The project advances computational design by establishing the methodological foundations for artificially intelligent design: human-level or supra-human generation of partial or full UI designs assuming access to behavioural data only (e.g., log data). The main objective is Bayesian Artefact Design, a reinforcement-learning-based formalism for AI agents that design artefacts for human use. AI agents learn to interpret people and predict the consequences of their interventions on them. Three challenging sub-objectives set this approach apart from existing approaches to computational design: (1) Agency: modeling the AI designer as an agent in the world that its designs are affecting; (2) Inference: the ability to observe and infer people's abilities, beliefs and intentions from behavioral data; (3) Speculativity: the ability to entertain possible designs by accurately predicting their consequences to people.
Short titleBAD
StatusActive
Effective start/end date01/09/201831/08/2023

Fingerprint

Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.