On understanding, modeling and predicting user behavior in web search

Web search engines provide quick and easy access to information available online. In the early days of the Internet, web links were the most valuable source of information for predicting result usefulness. Nowadays, the whole web search stack relies on user behavioral data, starting from crawling policies to optimizing presentation of the results. However, accurately interpreting user interaction behavior is not straightforward due to various types of bias. For example, users tend to click more on results ranked on top positions (position bias) and visually salient results (attention bias). In this thesis, we study existing tools for modeling and predicting user interactions with a search engine, improve them and develop new ways of gaining insights about user behavior.