AbstractComputer dissemination of information offers significant advantages over manual dissemination because the computer can use strategies that are impractical and in some cases impossible for a human. This paper describes the Ames Laboratory Selective Dissemination of Information1 system with emphasis on the effectiveness of user feedback. The system will accept any document, abstract, keywords, etc., in a KWIC2 or Science Citation Index3‐Source format. User profiles consist of words or word clusters each with an initially assigned significance value. These values are used in making the decision to notify a user that he may be interested in a particular document. According to responses, the significance values are increased or decreased and quickly attain an equilibrium which accurately describes the user's interests. The system is economical compared to other existing SDI systems and human intervention is negligible except for adding and deleting profile entries.