The background of the research project is to research models of individual decision making about information that 1) are readily applicable to data and 2) allow for behavioral biases. In the first step, this project obtains a model of stochastic information choice. In this model, the decision maker chooses a sequence of search keywords. The model accounts for a crucial difference between information choice as commonly modeled using information partitions and information choice in practice. Information about whether an event in an information partition is true is equivalent to information whether its negation is true. However, for lists of keywords (for example, a search engine query) there is no meaningful negation.