Is the Best Possible User Interface Predetermined?

Abstract:
This talk has humble beginnings, but it ends up by putting forward a grand question to human movement science. A few years ago, we started looking at biomechanical simulation as a means to solve ergonomics issues plaguing novel user interfaces. In a typical study, we would ask users perform interactive tasks in a motion capture laboratory and use the data in biomechanical simulation as well as performance analysis. However, we soon realized that even though this produces a wealth of useful data, it is inherently an *a posteriori* method limited to observed movements. The question emerged: What kind of knowledge would help identifying the best user interface design out of a larger design space *a priori*? The fundamental hypothesis I discuss in this talk is that the design of the best user interface is not a creative problem but a modeling problem. That is, the best design is (somehow) predetermined by biomechanical, neuromechanical, and cognitive propeties. The challenge to human movement science is to produce models that capture these aspects to make valid predictions about both musculoskeletal demands *and* movement performance (speed, accuracy) of a given movement. I present our early attempts with data-driven approaches (muscle coactivation clustering) and interactive visualizations. I then show first results from muscle synergy based models that, together with optimal control theory and generalized human body models, may solve this problem for non-contact movements. I discuss how such models may be used to optimize user interface designs computationally.