Usable automated data inference for end-users

Nowadays, data are more easily accessible than ever, yet support for correlating base data into interesting consequences is often unavailable, or too expensive, or too technical for many users. Lacking access to automated tools, these users are forced to infer the information they need from the raw data manually. This is time consuming, error prone, and often results in bad decisions, suboptimal plans, or missed opportunities. We have designed a possible solution to this problem, which revolves around extending the spreadsheet paradigm to enable users to define useful forms of correlation among their data. Spreadsheets are widely available to users, already allow them to routinely define complex custom calculations on numerical data, and are easy to use productively with little or no training. However they are unable to deal with the kind of recursive relational calculations needed for many common types of data correlation problems. The mentioned design exploits techniques from logic programming to solve these issues. One aspect that remains to be fully developed is a matching user interface for these functionalities. A specific challenge is to retain the cognitive simplicity, ease to use and gentle learning curve of today’s spreadsheet applications. We propose to develop an advanced prototype of this design and perform extensive usability studies to understand how to best present the extended functionalities to interested users. Usability is the focal point of this work.