In telerobotic systems, human beings are required to provide input to the telerobot. It is very difficult to provide accurate and smooth input paths quickly through the master arm and avoid collisions. These problems are further compounded by time-delay when the distances involved are large (for example, in the case of space exploration). If an intermediate VR system can modify the human input to provide smooth, collision-free paths, the intuitive capabilities of human beings to identify a feasible path can be quickly combined with the machine's ability to modify the colliding segments to noncolliding ones. We call this synthesis shared intelligence . This paper develops the strategies and mathematics to implement this concept of shared intelligence in path planning, using a VR interface in which human and machine abilities are combined appropriately. Shared intelligence unburdens the human operator from precisely and accurately inputting the planned path. With the help of the free space developed for a simple telerobot, it is shown how the human input is converted into parameterized paths in the configuration space, checked for collisions, and automatically modified into collision-free paths using repulsive fields introduced in the configuration space. Simulations and experimental results are also shown to illustrate these concepts.