CIF: EAGER: Towards an Information Theory of Attention

University Of Illinois At Urbana-Champaign, Urbana IL

Investigators

Abstract

Information theory has set forth ideals of communication system performance that have driven technological development for decades. Successes in technologies like telephony and the Internet now allow an individual to reliably receive a great deal of information. Indeed, a major problem facing society today is too much information rather than too little: the problem of information overload. A wealth of information creates a scarcity of human attention and a need for engineering designs that consider both information and attention. Problems, however, are not just technical in scope but fundamentally socio-technical, with solutions requiring explicit consideration of human behavior. This early-concept exploratory research aims to understand the gain and loss of attention in communication, drawing on mathematical models of people. The plan first considers the fundamental trade-off between information rate and Bayesian surprise, a psychological measure of human attention, over timing channels and in multiple-access and multicast settings modeled after social media systems. The plan next considers the best ways to communicate when the receiver may lose attention, either by getting "worn out" or by not being surprised enough. Methods to obtain the best fidelity over channels that die will also be developed. the project integrates with an education/outreach plan also focused on attention, which aims to inculcate a culture of engagement in the college of the future. Educational materials for undergraduates across campus, in new courses on big data and on heroic engineering will be developed. Public outreach will draw on the proposer's media presence and IEEE initiatives. The proposed work will lay new foundations for an information theory of socio-technical systems that brings mathematical rigor and concern for fundamental limits to problems and information systems that are becoming increasingly prominent in society: problems that must dispense with the assumption that humans can be abstracted away. The significance lies in the introduction of a new dimension to communication, attention, whose importance in information-theoretic terms remains untapped. This is expected to lead to novel trade-offs between information and surprise in various network settings, as well as novel formulations of communication in the presence of both noise and signal-dependent failure, providing concrete formulations of reliable communication using finite block lengths.