Active Formalization in Artificial and Human Reasoners
Abstract
Cognitive agents see the world through representations—the set of concepts, symbols, rules, relations, and other structures used to construct an understanding of the world. Although not an everyday task, a remarkable feature of human reasoners is that they can take informal or partially formal representational systems, reason over (as opposed to with) those representational systems, and in some cases produce entirely new systems. When the newly created representational system is formal, we call this process formalization. Some components of the formalization process can be automatic, non-deliberate, and akin to a process of what might be called mere translation. However, ensuring that the resulting formal representational systems are “good” involves more than mere translation: significant deliberate and justified reasoning about some problem space, highly creative thinking, the creation of new abstractions, and iterative refinement are just some of the processes used by some of the greatest reasoners in history to produce formalizations. I use the term active formalization to refer to the aspect of formalization that goes above and beyond mere translation, and produces “good” formal representational systems. I propose a research program focusing on two questions: What is involved in the process of creating, restructuring, and reasoning over representational systems? Second, can artificial reasoners achieve active formalization, either autonomously or with the help of humans?
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Apr 09, 2018
- Source ID
- FA95501810052
Entities
People
- John Licato
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- University of South Florida