Abstract:
Information overload and an abundance of choices lead to situations where selecting one option becomes extremely difficult or even worse, a guessing game. Collaborative ranking systems are widely used to alleviate this problem by creating intelligent rankings of items, based on an aggregation of user opinions. Current ranking systems fall short in a number of areas, including accuracy, transparency and flexibility. Wiki Ratings, the system developed in this thesis, is the first completely user-definable, multi-criteria ranking system that creates a consensus-driven decision-making culture that can work over any grouping of entities, each defined by a set of value dimensions. By using a paradigm that allows users to digitize their cognition and define which value dimensions are important and how important they are, Wiki Ratings can produce personalized rankings and intelligent, value-driven recommendations with higher accuracy, transparency and flexibility. The system can show why an item is ranked the way it is, which values play a bigger role in influencing the rankings, and even which values are preferred by different demographics. This will consequently enable service and product providers to dig deeper and understand why their services or products are ranked the way they are. A proof of concept is developed and evaluated that allows users to digitize their belief systems and rate entities over an arbitrary set of value dimensions. A ranking of entities is then created based on an additive multi-criteria distance metric and a merger of each individual's belief system. The paradigm used to define entities allows powerful, cross boundary and transparent recommendations. An evaluation study and survey showed that the algorithms in use are more accurate than a number of existing ranking and recommendation systems.