Developing new methods for analyzing large quantities of player profile data, such as hundreds of thousands of profile images or chat logs.
Looking for new methods to analyze player profile data and tease out interesting relationships. For example, do people who like to play sports games create a certain style of Mii avatar? Finding these relationships requires automatic methods for extracting and analyzing patterns, because humans can't coherently process tens or hundreds of thousands of examples at once. But once these algorithms find potential patterns, human insight is required to reveal their meaning, so patterns identified in broad data sets can be confirmed by followup experiments with smaller groups. Whereas the initial analysis relies on algorithm development and quantitative statistics, confirming those findings requires careful experiment design and qualitative methods.
