Developing theories behind different types of data mining (for example, lower bounds on complexity of mining sequences) would enable us to determine if our data is mineable. We could use this to develop ``profiles'' of hard-to-mine data (for example, a characterization of data that is likely to contain association rules). This will enable us to build systems that provide the data (or access to the data) in a way that ensures that mining is either unlikely to find interesting results, or will find ``planted'' results that do not concern us.