Abstract. Sequence clustering is a technique of bioinformatics that is used to discover the properties of sequences by grouping them into clusters and assigning each sequence to one of those clusters. In business process mining, the goal is also to extract sequence behaviour from an event log but the problem is often simplified by assuming that each event is already known to belong to a given process and process instance. In this paper, we describe two experiments where this information is not available. One is based on a real-world case study of observing a software development team for three weeks. The other is based on simulation and shows that it is possible to recover the original behaviour in a fully automated way. In both experiments, sequence clustering plays a central role.

Abstract. While personal data is a source of competitive advantage, businesses should consider the potential reaction of individuals to certain types of data requests. Privacy research has identified some factors that impact privacy perceptions, but these have not yet been linked to actual disclosure behaviour. We describe a field-experiment investigating the effect of different factors on online disclosure behaviour. 2720 US participants were invited to participate in an Amazon Mechanical Turk survey advertised as a marketing study for a credit card company. Participants were asked to disclose several items of personal data. In a follow-up UCL branded survey, a subset (N=1851) of the same participants rated how they perceived the effort, fairness, relevance, and sensitivity of the first phase personal data requests and how truthful their answers had been. Findings show that fairness has a consistent and significant effect on the disclosure and truthfulness of data items such as weekly spending or occupation. Partial support was found for the effect of effort and sensitivity. Privacy researchers are advised to take into account the under-investigated fairness construct in their research. Businesses should focus on non-sensitive data items which are perceived as fair in the context they are collected; otherwise they risk obtaining low-quality or incomplete data from their customers.

To assess the risk of a loan applicant defaulting, lenders feed applicants" data into credit scoring algorithms. They are always looking to improve the effectiveness of their predictions, which means improving the algorithms and/or collecting different data. Research on financial behavior found that elements of a person"s family history and social ties can be good predictors of financial responsibility and control. Our study investigated how loan applicants applying for a credit card would respond to questions such as "Did any of your loved ones die while you were growing up?" 48 participants were asked to complete a new type of credit card application form containing such requests as part of a "Consumer Acceptance Test" of a credit card with lower interest rates, but only available to "financially responsible customers." This was a double-blind studythe experimenters processing participants were told exactly the same. We found that: (1) more sensitive items are disclosed less often -e.g. friends" names and contact had only a 69% answer rate; (2) privacy fundamentalists are 5.6 times less likely to disclose data; and (3) providing a justification for a question has no effect on its answer rate. Discrepancies between acceptability and disclosure were observed -e.g. 43% provided names and contact of friends, having said they found the question unacceptable. We conclude that collecting data items not traditionally seen as relevant could be made acceptable if lenders can credibly establish relevance, and assure applicants they will be assessed fairly. More research needs to be done on how to best communicate these qualities.

No abstract

No abstract