Generalized Representational Information Theory

Generalized Representational Information Theory:


Just as researchers today struggle to find working definitions for words like consciousness and intelligence, they struggled to find a standardized meaning for the word information in the early 1900s. Ralph Hartley, a research for Bell Laboratories, first introduced a theory of information based on the idea that information consisted of strings of symbols, a reasonable idea in the age of telegraph, telephones, and radio. Shannon and Weaver moved things along in the 1940s, resulting in Shannon-Weaver Information Theory (SWIT). While Hartley's theory was concerned primarily with sets of symbols, SWIT was concerned with the probability or uncertainty of events (the likelihood a particular structure or sequence of symbols are meaningful). Both theories fall far short of describing what a modern cognitive scientist or AI researcher means when they talk about information. A newer theory was developed in the field of psychological research, Representational Information Theory (RIT). The idea behind RIT is that communication between animals and their environment is mediated by concepts. The only drawback of RIT is that it only supported binary dimensions. RIT looks at information in terms of complexity rather than uncertainty like SWIT. In a new paper published in the journal Information, researchers described a generalized version of RIT, called GRIT that may be useful in the fields of AI and robotics:

"concepts live in the mental space of organisms ranging from aplasia to insects and from dolphins to humans. Some may argue that they also live in the mental spaces of intelligent robots and expert systems. Regardless, the point is that only by using concepts as mediators can information as a measurable quantity reflect human intuitions as to what is informative."

The paper includes a technical appendix with mathematical examples of Generalized Representational Information Theory (GRIT) showing examples such as the one above that includes three dimensions (shape, color, and size). For all the details, read the paper, titled, "Complexity over Uncertainty in Generalized Representational Information Theory (GRIT): A Structure-Sensitive General Theory of Information" (PDF format). The paper was written by Ronalda Vigo of the Center for the Advancement of Cognitive Science, Psychology Department, Ohio University. Hartley's 1928 paper is available online as Transmission of Information (PDF format). Shannon's 1948 paper (on which the later Shannon-Weaver book was based) can be found as "A Mathematical Theory of Communication" (PDF format).

DIGITAL JUICE