Drawing on well-known examples of serendipity in scientific discovery, we develop a set of criteria that can be applied to model and evaluate serendipity in computational settings. We use design patterns, and the growth of a pattern language, as a way to describe the processes of discovery and invention that comprise serendipitous encounters. We show how several earlier patterns of serendipity can be applied in a Writers Workshop for computational systems, and include related recommendations for practitioners.
When we have addressed the outstanding issues pointed out by previous reviewers, we plan to submit the revised paper to Minds and Machines.