Multi Agent Student Examination System Over Internet

There is a vision supported by surveys and scientific research that one day Computerized testing (CT) will virtually replace paper and pencil testing. More than five decades ago, the earliest forms of CT systems were focused on certification testing for the Information Technology industry. Today, at hand are many generally accepted models of CT systems serving elements of practically every market once dominated by paper and pencil testing. However, problems arise while choosing a CT implementation model for an educational institution. Existing CT systems have been implemented following either Computer based testing (CBT), Web based testing (WBT) or Software agent based testing (SABT) approaches. Each approach offers unique pros and cons over the other. On the other hand, in an educational institution, the comparative disadvantages for each approach outweigh its unique advantages when that particular implementation approach is used from one discipline to another. Specifically CBTs are limited to a local coverage; WBTs are vulnerable to unpredictable network delays and link/node failures; while SABTs use incompatible interfaces. Moreover, all existing CT systems are designed around one test deliver model (TDM), that is, linear testing, linear on the fly testing, testlet, mastery model or computerized adaptive testing, yet within a discipline different test delivery model maybe desired. In this thesis my approach to educational institutional testing is different. Our approach is based on combining agent-oriented approach with existing web-interface technologies. The output of the thesis is a Multi Agent Student Examination System over Internet (MEST), a system that blends the strengths of CBT, WBT and SABT. MEST is designed to offer any to the test delivery models.