Social media has become a popular platform for millions of users to share activities and thoughts. Many applications are now tapping on social media to disseminate information (e.g., news), to promote products (e.g., advertisements), to manage customer relationship (e.g., customer feedback), and to source for investment (e.g., crowdfunding). Many of these applications require user profile knowledge to select the target social media users or to personalize messages to users. Social media user profiling is a task of constructing user profiles such as demographical labels, interests, and opinions, etc., using social media data. Among the social media user profiling research works, many focus on analyzing posted content. These works could run into the danger of non-representative findings as users often withhold some information when posting content on social media. This behavior is called selective self-disclosure. The challenge of profiling users with selective self-disclosure behavior motivates this dissertation, which consists of three pieces of research works.

The first work is that of profiling silent users in social media. Silent users (or lurkers) are the users who choose not to disclose any information. In this work, we examined 18 weeks of tweets generated by two Twitter communities consisting of more than 110K and 114K users respectively. We find that there are many lurkers in the two communities. We also show that by leveraging lurkers' neighbor content, we are able to profile their attributes with accuracy comparable to that of profiling active users.

The second work is that of profiling users with selective topic disclosure. Social media users may choose not to post some of their interested topics. As a result, their posting and reading topics can be different. To better determine and profile social media users' topical interests, we conducted a user survey in Twitter. In this survey, participants chose the topics they like to post (posting topics) and the topics they like to read (reading topics). We observe that users' posting topics differ from their reading topics significantly. We find that some topics such as "Religion", "Business" and "Politics" attract much more users to read than to post. With the ground truth data obtained from the survey, we show that predicting reading topics can be as accurate as predicting posting topics using features derived from posted content, received content and social networks.

The third work is that of profiling users with selective opinion disclosure. In social media, users may not disclose their opinions on a specific issue i even when they are interested in i. We call these users issue-specific silent users or i-silent users. This work investigates the opinions of i-silent users. We conducted an opinion survey on a set of users for two popular social media platforms, Twitter and Facebook. We analyzed the survey results together with their social media data. We find that more than half of our users who are interested in issue i are i-silent users in Twitter. The same has been observed for our Facebook users. The survey results also show that i-silent users have opinion distribution different from the users who post about i. With the ground truth user opinions from the survey, we show that predicting i-silent users' opinions can achieve reasonably good accuracy from user posted content that is not related to issue i, and achieve better accuracy when we utilize user opinions on other issues as features.