Date of Award
Doctor of Philosophy (PhD)
Electrical Engineering and Computer Science
Varshney, Pramod K.
decision making, distributed inference, Human decision making, human machine intelligence
Computer Sciences | Electrical and Computer Engineering | Engineering | Physical Sciences and Mathematics
With the advent of the internet of things (IoT) era and the extensive deployment of smart devices and wireless sensor networks (WSNs), interactions of humans and machine data are everywhere. In numerous applications, humans are essential parts in the decision making process, where they may either serve as information sources or act as the final decision makers. For various tasks including detection and classification of targets, detection of outliers, generation of surveillance patterns and interactions between entities, seamless integration of the human and the machine expertise is required where they simultaneously work within the same modeling environment to understand and solve problems. Efficient fusion of information from both human and sensor sources is expected to improve system performance and enhance situational awareness. Such human-machine inference networks seek to build an interactive human-machine symbiosis by merging the best of the human with the best of the machine and to achieve higher performance than either humans or machines by themselves.
In this dissertation, we consider that people often have a number of biases and rely on heuristics when exposed to different kinds of uncertainties, e.g., limited information versus unreliable information. We develop novel theoretical frameworks for collaborative decision making in complex environments when the observers may include both humans and physics-based sensors. We address fundamental concerns such as uncertainties, cognitive biases in human decision making and derive human decision rules in binary decision making. We model the decision-making by generic humans working in complex networked environments that feature uncertainties, and develop new approaches and frameworks facilitating collaborative human decision making and cognitive multi-modal fusion.
The first part of this dissertation exploits the behavioral economics concept Prospect Theory to study the behavior of human binary decision making under cognitive biases. Several decision making systems involving humans' participation are discussed, and we show the impact of human cognitive biases on the decision making performance. We analyze how heterogeneity could affect the performance of collaborative human decision making in the presence of complex correlation relationships among the behavior of humans and design the human selection strategy at the population level. Next, we employ Prospect Theory to model the rationality of humans and accurately characterize their behaviors in answering binary questions. We design a weighted majority voting rule to solve classification problems via crowdsourcing while considering that the crowd may include some spammers. We also propose a novel sequential task ordering algorithm to improve system performance for classification in crowdsourcing composed of unreliable human workers.
In the second part of the dissertation, we study the behavior of cognitive memory limited humans in binary decision making and develop efficient approaches to help memory constrained humans make better decisions. We show that the order in which information is presented to the humans impacts their decision making performance. Next, we consider the selfish behavior of humans and construct a unified incentive mechanism for IoT based inference systems while addressing the selfish concerns of the participants. We derive the optimal amount of energy that a selfish sensor involved in the signal detection task must spend in order to maximize a certain utility function, in the presence of buyers who value the result of signal detection carried out by the sensor. Finally, we design a human-machine collaboration framework that blends both machine observations and human expertise to solve binary hypothesis testing problems semi-autonomously.
In networks featuring human-machine teaming/collaboration, it is critical to coordinate and synthesize the operations of the humans and machines (e.g., robots and physical sensors). Machine measurements affect human behaviors, actions, and decisions. Human behavior defines the optimal decision-making algorithm for human-machine networks. In today's era of artificial intelligence, we not only aim to exploit augmented human-machine intelligence to ensure accurate decision making; but also expand intelligent systems so as to assist and improve such intelligence.
Geng, Baocheng, "Augmented Human Machine Intelligence for Distributed Inference" (2021). Dissertations - ALL. 1399.