The digital age has produced an abundance of automated sensing mechanisms for collection of very large data streams which are beyond human capability to analyze, evaluate and derive decisions about the causal physical changes in the dynamic system. In addition, users have ubiquitous access to human generated contextual information from a number of sources which can greatly improve situational understanding and interpretation of dynamic change in the system. The user is, thus inundated with the processing and interpreting of this growing volume of variable data to extract actionable information in support of time critical needs. Examples include urban area surveillance, geospatial intelligence, introduction and adaptive marketing of competitive products, adaptive sensing and command and control of natural resources such as watershed and agriculture, and internet and infrastructure protection. In all these examples, machine level intelligent processing of low fidelity sensor data must be augmented by high-level understanding and human cognition of situational awareness for extracting contextually relevant actionable intelligence from big data. A unified mathematical framework is formulated so that humans and machines may collaborate to realize the potential for improved in-situ decision-making, contextual understanding, and event prediction without imposing a need for increased human workload in dynamic, unknown and uncertain operational environments.
There are fundamental differences in human cognition and machine cognition of data. These differences stem from incompatible internal representations of information, processing methods, structural differences in logic and reasoning, and learning and inference mechanisms. Humans are better at fusing and interpreting high level textual or visual information in low dimensions while machines can more easily contend with the curse of dimensionality, compressing and processing low fidelity electronic sensing data. Many of the differences in logic and reasoning stem from the hitherto undisputed use of probabilistic logic in machine cognition which has been shown to be incompatible with human decision making. Quantum probability theory based models of human cognition will be presented which resonate with deeply rooted psychological intuitions of human cognition and fusion of information. On the other hand, machine cognition models will draw on recent work in Symbolic Dynamic Filtering to formulate the fundamentals of an information science for discovery of causal patterns in asynchronous streams of multivariate spatio-temporal data. These causal patterns, represented as finite state automata, fully capture the generating dynamics, i.e. they preserve the statistical characteristics of the original data streams. Hence, they form the alphabet of a formal machine language for distributed machine learning, situation awareness, statistical prediction and distributed control. Machine learning of this formal language has been shown to be orders of magnitude faster than classical methods based on Bayesian networks, neural networks, hidden Markov models or particle filtering. Innovative research concepts for developing analytical transformations between human and machine cognition models will be discussed that enable human operators to gain insight and foresight into abstract machine interpretations of big data while assessing and enabling them to leverage rich experiential and un-modeled domain knowledge and perceptions of human supervisors.
Shashi Phoha earned the Ph.D. degree in Mathematical Statistics from Michigan State University, East Lansing, MI, M.S. degree in Operations Research from Cornell University, Ithaca, NY, and M.A. degree in Mathematics from Punjab University, Chandigarh, India. Dr. Phoha joined the Pennsylvania State University in June 1991, and is currently a Professor of Electrical Engineering and the Director of Information Sciences & Technology Division, Applied Research Laboratory. Prior to joining Penn State, Dr. Phoha held academic positions at Bucknell University and Tulane University as well as research and management positions at Computer Science Corporation, ITT Defense Communications Division, and MITRE Corporation.
Dr. Phoha has diverse experiences in establishing and leading collaborative advanced research programs and laboratories in major U.S. Government Labs, industry, and academic institutions. She has competitively won and successfully managed several multi-million dollar programs of national importance and attracted prominent scholars to conduct this research. Dr. Phoha has supervised over 25 doctoral students and several MS and BS Honors students.
Dr. Phoha was awarded the 2004 IEEE Computer Society Technical Achievement Award. She has published over 200 research papers, two books, and holds several patents licensed to industry. She has held leadership positions in national and international organizations and given about 25 invited keynote addresses at conferences and symposia. Dr. Phoha chaired the Springer-Verlag Technical Advisory Board during 2001-02 for publishing the Dictionary of Internet Security. She was the Guest Editor of Special Issues of IEEE Transactions (TMC), associate editor of the IEEE Transactions on Systems, Man, and Cybernetics for five years and is editor of the International Journal of Distributed Sensor Networks. From 1992 to 2001, she was on the board of directors of the International Consortium CERES Global Knowledge Network along with representatives of 13 other international universities. She was a member of the National Software Strategy Group and co-chair of its Committee on Innovation. She was also a leader in launching the National Cybersecurity Grand Challenge for Critical Infrastructure Protection.