Computational systems and methods are disclosed that learn about and assist with appropriate structure-based design choices. The systems and methods autonomously explore design states, and suggest to a user one or more optimize design states. Constraints are provided to limit exploration to valid design states. Systems and methods are disclosed that assist groups of users with coordinating their efforts in producing a cohesive design. Systems and methods are disclosed for learning from past optimizations in order to provide more rapid convergence on an optimized design, avoid local maxima and other hurdles to optimization, avoid undesired optimizations, and so on.