Determining Internet Users' Values for Private Information

We examine the problem of determining a user’s value for his/her private information. Web businesses often offer rewards, such as discounts, free downloads and website personalization in exchange for information about the user, such as name, phone number and e-mail address. We present a technique that helps the user determine whether such an offer is acceptable by computing its value in terms of the consequences that could occur as a result of such an information exchange. Bayesian networks are used to model dependencies in the user’s utilities for such consequences, and utility elicitation is used to reduce the uncertainty of these utilities. We also derive a “bother cost”, which is used by the elicitation engine to determine the optimal time to stop the question process. A simple example experiment demonstrates the effectiveness of the technique by significantly improving the user’s expected utility in a simple privacy negotiation.