Scheduling in hospitals is a decision problem that relies on encoded information (such as room assignments or diagnoses) as much as on non-encoded information (medical knowledge). To support clinical scheduling with information systems, both types of information and their respective carriers have to be incorporated in combination, thus always requiring a "human-in-the-loop", which prohibits full autonomy. In contrast to other proposals in Autonomic Computing, this paper makes a case for a hybrid humancomputer decision support system to support scheduling in hospitals that combines humans' ability to improvise and to assess a situation, with computers' ability to communicate quickly to achieve emergent self-organization. Furthermore, the paper outlines an approach to evaluate such a system.