Abstract
The use of Bayesian networks for behavioral analysis is gaining attention. The design of such algorithms often makes use of expert knowledge. The knowledge is collected and organized during the knowledge acquisition design task. In this paper, we discuss how analytical games can be exploited as knowledge acquisition techniques in order to collect information useful to intelligent systems design. More specifically, we introduce a recently developed method, called the MARISA (MARItime Surveillance knowledge Acquisition) Game. The aim of this game is to ease the elicitation from domain experts of a considerable amount of conditional probabilities to be encoded into a maritime behavioral analysis service based on a multi-source dynamic Bayesian network. The game has been deployed in two experiments. The main objectives of such experiments are the validation of the network structure, the acquisition of the conditional probabilities for the network, and the overall validation of the game method. The results of the experiment show that the objectives have been met and that the MARISA Game proved to be an effective and efficient approach.
Highlights
Due to the significant amount of information available and its associated uncertainty, there is a growing attention towards expert systems in support of human Situational Awareness able to perform probabilistic reasoning [1], such as the ones based on Bayesian networks (BNs)
With respect to the first objective, the MARISA Game allowed the collection of the players’ belief assessments for the different knowledge structures. Those beliefs have been mapped to subjective probabilities (Section 3.4) and used to populate the Multi-Source Dynamic Bayesian Network (MSDBN)
As an example of such process, we can consider Figure 8. It shows four different configurations of the same knowledge structures included in a knowledge card of the Smuggling of Goods (SMG) Island card deck and the corresponding belief stated by a player during EXP2 (Section 3.5)
Summary
Due to the significant amount of information available and its associated uncertainty, there is a growing attention towards expert systems in support of human Situational Awareness able to perform probabilistic reasoning [1], such as the ones based on Bayesian networks (BNs). Knowledge acquisition (KA), which is a step of primer importance in the design of many systems, is the process of extracting, structuring, and organizing domain knowledge from experts [4]. Two widely used techniques are unstructured or structured interviews (i.e., protocol-generation techniques) [4]. Additional techniques include protocol analysis techniques, hierarchy-generation techniques, matrix-based techniques, sorting techniques, limited-information, and constrained-processing tasks and diagram-based techniques [5]. Details on the different techniques and their use can be found in the abundant literature on the topic (e.g., [6,7,8,9])
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
