SCHOLAR Intelligent Tutoring System Developed by Jaime Carbonell Sr.
Around 1970, Jaime Carbonell Sr. at Bolt Beranek and Newman (BBN) developed SCHOLAR, a computer-assisted instruction program that used a semantic network of geographical knowledge to conduct mixed-initiative tutorial dialogues, marking an early attempt to build adaptive, knowledge-based educational software.
Background
By the late 1960s, computer-assisted instruction had been around long enough to have a clear limitation. The programs that existed were essentially elaborate drill sheets. A student would see a question, type an answer, and the program would compare that answer against a list of expected responses. Right or wrong. Next question. The software followed a script, and the script never changed.
There was no room in that model for a student to ask something. If you did not understand a concept and wanted to know more before moving on, the program could not help. It only knew which questions came next, not why you might be confused. The knowledge embedded in these systems was shallow, locked into fixed paths that a student had to walk in order.
Researchers in artificial intelligence had been developing richer ways to represent knowledge, using structures called semantic networks. A semantic network organises facts as a web of related concepts: not just “the Amazon is a river” as an isolated item, but a connected set of relationships between rivers, countries, regions and climates. The question was whether that kind of structured knowledge could drive something more flexible than a quiz.
What happened
Jaime Carbonell Sr., working at Bolt Beranek and Newman (BBN) in Cambridge, Massachusetts, built a system called SCHOLAR to try exactly that. The subject was South American geography, and instead of storing a list of facts, SCHOLAR stored them in a semantic network that linked concepts to one another. When a student asked about Peru, the system could draw on related nodes: its neighbours, its climate, its major cities. The knowledge was not a script. It was a structure the system could reason over.
What set SCHOLAR apart from earlier programs was its mixed-initiative dialogue. In most computer-assisted instruction, the machine asks and the student answers. SCHOLAR allowed it to go both ways. The student could ask a question at any point, and the system would generate a response by querying its network rather than looking up a pre-written reply. Control of the conversation could shift back and forth. Carbonell described the approach in his 1970 paper in IEEE Transactions on Man-Machine Systems, one of the first places this kind of knowledge-driven tutoring appeared in the published literature.
SCHOLAR was a research demonstration, not a product that reached classrooms at scale. What it showed was that the techniques being developed in AI research, specifically how to represent and connect knowledge, could be applied to teaching software in ways that made genuine dialogue possible. A further description of the system appeared in a 1972 ERIC report, which documented its structure and the thinking behind it. Together, those publications gave later researchers a concrete example to build from.
Why it mattered
SCHOLAR was among the first systems to use a structured semantic network as an underlying knowledge representation for a tutoring application, moving beyond the rigid, scripted question-and-answer formats of earlier computer-assisted instruction. Its mixed-initiative dialogue structure, allowing the student to ask questions as well as answer them, established a design template that subsequent intelligent tutoring system research built upon for decades. The work demonstrated that AI knowledge-representation techniques could be applied directly to educational software, bridging the expert-systems tradition and pedagogy.
People
Jaime Carbonell Sr
Organisations
Bolt Beranek and Newman, Massachusetts Institute of Technology
Sources
- AI in CAI: An Artificial Intelligence Approach to Computer-Assisted Instruction.IEEE Transactions on Man-Machine Systems (ACM/IEEE).Primary source
- SCHOLAR: A Computer-Based Tutorial System for Instruction in Geography (ERIC ED061595).ERIC / U.S. Department of Education.Secondary
Cite this page
AI Achievements. (1970). SCHOLAR Intelligent Tutoring System Developed by Jaime Carbonell Sr.. Retrieved 2026-08-22, from https://achievements.ai/milestone/scholar-program-developed-jaime-carbonell
@misc{achievements_scholar_program_developed_jaime_carbonell,
title = {SCHOLAR Intelligent Tutoring System Developed by Jaime Carbonell Sr.},
author = {{AI Achievements}},
year = {1970},
url = {https://achievements.ai/milestone/scholar-program-developed-jaime-carbonell}
}