Ever since William of Ockham declared that in situations accounted for by multiple explanations one should prefer the simpler explanation, scholars have struggled with how to use this advice.

Chance discovery is about investigations at that boundary of deduction and induction, where a situation described with partial Possible "explanations." The boundary between deduction- n where our explanations are provably correct, and induction---where our explanations are plausible but improvable, is proving to be broader and more we first thought. complicated than .

Our search for plausible methods of "jumping to conclusions" has produced everything from methods that reveal a complex tradeoff t unexplainable but apparently effective application of heuristics, to over¬formalized logical systems which leave no scope for new forms of knowledge.

This volume captures what could be argued as the complete repertoire of the three most important aspects of understanding this landscape between deduction and induction. Even the concept of "chance discovery" captures the problem: what knowledge leads, perhaps by "chance" to improved selection of explanations, without denying the possibility of other as-yet-missing knowledge?

 Included in this three segment volume are first a broad repertoire of various knowledge-driven heuristics which demonstrate improved identification of "chance" across diverse applications. The examples of different heuristics include the identification and use of emotional responses, tracking of eye movements, chance discovery to understand multilangua9e browsing behaviour, as well as more conventional application of analogical reasoning methods. The second segment provides a sampling of formalization methods for chance discovery, which walk the fine line between identifying new knowledge forms that can be brought within formalized reasoning, without closing the door to the richness of partial information situations. Included here are those expected candidates, including extend formal deductive reasoning with the probability and e various the tools that partial knowledge. The final segment provides a sampling of some of the existing tools for conducting experiments with chance discovery systems, again driven by general applications (e.g., interactive knowledge acquisition, opponent modeling, and customer data mining).

Perhaps the best summary is that the walk through the space between complete and partial knowledge situations is complex, and there are many different methods available to approach that walk. This volume provides a broad and interesting repertoire of sign posts on that walk, and will help future explorers improve their understanding of a very difficult problem.

Randy Goebel, Department of Computing Science, University of Alberta