Decision Learning: Learning with Strategic Decision Making

Опубликовано 27 июля 2016, 2:29

With the increasing ubiquity and power of mobile devices as well as the prevalence of social networks, more and more decisions and activities in our daily life are being recorded, tracked, and shared. This abundant and still growing real life data, known as �big data�, provides a tremendous research opportunity in various fields whereas traditionally data collected only in controlled and laboratory environment are available for analysis. To analyze, learn and understand such user-generated big data, machine learning has been an important tool and various machine learning algorithms have been developed. However, most existing machine learning algorithms focus on optimizing a global objective function at macroeconomic level, while totally ignore users� local interests at the microeconomic level. Since the user-generated big data is the outcome of users� decisions, actions and their socio-economic interactions, which are highly dynamic, without involving users� local behaviors and interests, the results learned at a certain time instance may not be applicable in a future time instance. In this talk, we present a new tool called �decision learning� that can involve users� behaviors and interactions by combining learning with strategic decision making. We will discuss some examples to show how decision learning can be used to better analyze users� optimal decision from a user� perspective and design a mechanism from the system designer�s perspective to achieve a desirable outcome.