It could become a scientist’s best friend: software that incorporates everything you read, write, speculate, and infer, along with the whole body of literature and data relevant to your field, and comes up with helpful suggestions, unseen connections, and even resource budgets to help guide your research.
The same kind of software could aid in non- science areas, too, from crop rotation planning for agriculture to decision support for logistics. SFI Visiting Scientist Stephen Racunas is leading work under a two-year U.S. Office of Naval Research grant, “Contradiction-based logic for information fusion,” to develop the required theory and algorithms.
“Contradiction-based logic is a direct formalization of the scientific method,” explains Stephen, who is also a senior research engineer at Stanford University’s Medical School, where he is working toward biomedical applications of the technology. “[It] aims to nd the rela- tionships between new ideas, old ideas, and unexplained or unpublished data.”
CBL is not a statistical model of scientific inference, nor is it an artificial intelligence scheme. “In a way it’s the opposite of A.I.,” he says. “In no way does it attempt to replace, model, or supplant human thought. At most it’s Artificial Anti-Stupidity...it’s more like a spell-checker for the realm of ideas.”
Rather than attempt the Quixotic task of capturing the entire structure of scientific knowledge in an authoritative über-database or expert system, Stephen’s approach is to glean that structure from the bottom up, using the questions scientists ask, the conjectures they pose, the data they collect, and the conclusions they favor as raw material.
By translating these myriad thoughts and observations into the formal language of model theory, CBL creates a mathematical mapping of the ever-evolving state of scientific knowledge. This mapping can then be plumbed using well-established mathematical optimization techniques.
“Because of the way this new logic is set up,” Stephen says, “each optimal hypothesis is a boundary point” in an abstract space in which theories, data, and other constraints — “including time and budgetary concerns, if the scientist wishes” − are connected along high- dimensional surfaces.
“The best hypotheses are the ones on the border − where constraints meet possibilities,” he says.