Stanford Heuristic Programming Project founded by Edward Feigenbaum
In 1970, Edward Feigenbaum founded the Heuristic Programming Project (HPP) at Stanford University, establishing a dedicated research group to investigate the construction of knowledge-based expert systems and their application to scientific and medical domains.
Background
By the late 1960s, AI researchers had spent more than a decade chasing general intelligence: programmes that could reason about anything, given the right rules. The results were underwhelming. Systems like the General Problem Solver, developed by Herbert Simon and Allen Newell at Carnegie Mellon, could handle carefully constructed toy puzzles but fell apart when given real problems with the kind of messy, incomplete information that experts actually work with.
The field was beginning to ask a different question. What if, instead of building a machine that knew a little about everything, you built one that knew a great deal about a very narrow domain? Early work on DENDRAL, a programme that helped chemists identify the molecular structure of unknown compounds from mass spectrometry data, suggested this might work. Edward Feigenbaum at Stanford University had been part of that work, and what he saw there pointed toward a different way of thinking about what AI was actually for.
The central problem was institutional as much as intellectual. Researchers working on knowledge-based systems were scattered, and there was no dedicated home for the kind of sustained, applications-focused work the approach needed. Building a system that could reason like a specialist required encoding large amounts of specialist knowledge, testing it against real cases, and refining it over years. That was not the kind of work that fitted neatly into a single research grant or a single PhD project.
What happened
In 1970, Edward Feigenbaum founded the Heuristic Programming Project at Stanford University. The HPP was set up specifically to investigate how knowledge could be represented inside a computer in a way that allowed the machine to reason with it, and to build working systems in scientific and medical domains that would test those ideas against real problems.
The project took the view that the knowledge an expert carries, the rules of thumb, the patterns learned from experience, the shortcuts that come from years in a field, was itself the thing worth capturing. A programme built this way could not do everything, but within its domain it might do something genuinely useful. DENDRAL, which had been under development since the mid-1960s, continued at the HPP and went on to be used by working chemists. MYCIN, which was developed at Stanford during the 1970s and advised physicians on bacterial infections and antibiotic dosages, became one of the most studied programmes in the history of AI.
What the HPP provided was a structure: a place where the hard, slow work of knowledge engineering, which meant sitting with domain experts, extracting what they knew, and encoding it in a form a programme could use, could be treated as a serious research problem in its own right. Feigenbaum described this approach in a 1977 paper at the International Joint Conference on Artificial Intelligence, where he laid out the themes and methods that the project had developed. The HPP did not produce a single dramatic result in 1970. What it produced, over the decade that followed, was a coherent research programme that changed what people thought AI could be built to do.
Why it mattered
The HPP became the principal institutional home for expert-systems research in the 1970s and 1980s, producing landmark programmes such as DENDRAL and MYCIN that demonstrated a machine could reason reliably within a bounded domain by encoding specialist knowledge explicitly. By framing knowledge representation and inference as central engineering problems rather than incidental features of general intelligence, the project set the agenda for commercial expert-systems development in subsequent decades and shaped how AI was funded and understood by governments and industry through the early 1980s.
People
Organisations
Stanford University, Stanford Heuristic Programming Project
Sources
- Stanford Heuristic Programming Project, Feigenbaum Papers Finding Aid.Stanford University Libraries.Primary source
- The Art of Artificial Intelligence: Themes and Case Studies of Knowledge Engineering (Feigenbaum, 1977 IJCAI).ACM Digital Library.Secondary
Cite this page
AI Achievements. (1970). Stanford Heuristic Programming Project founded by Edward Feigenbaum. Retrieved 2026-08-22, from https://achievements.ai/milestone/heuristic-programming-project-hpp-by-ed
@misc{achievements_heuristic_programming_project_hpp_by_ed,
title = {Stanford Heuristic Programming Project founded by Edward Feigenbaum},
author = {{AI Achievements}},
year = {1970},
url = {https://achievements.ai/milestone/heuristic-programming-project-hpp-by-ed}
}