Research mentor: A tool to aid in the training of scientific investigators

We describe an intelligent mentor for teaching the ability to think scientifically. The student is given an arbitrary starting place in the matrix of knowledge surrounding an area of biomedical research. He/she then proposes hypotheses and supporting experiments which are checked against the knowledge base for agreement, consistency or contradiction. Agreement or consistency results in the report of successful experiments, thus advancing the student's ''state-of-the-art.'' Contradiction results in failure of the experiment to support the hypothesis. In either case, a new hypothesis can then be proposed and tested, each step being potentially contingent on results of the last. The knowledge base upon which the system operates is a frame-based implementation of the Biomatrix, augmented with pointers to literature citations. Each object (hypotheses, experiments, cells, animals, etc.) is described in terms of its properties and its relations to other objects. Thus, the matrix is represented as a semantic network, Other objects create the relations among the hypotheses, subhypotheses, experiments and other parts of the knowledge base. This system provides experiential learning at a rate determined by the student, while saving costly resources.