Artificial intelligence (AI) is increasingly of tremendous interest in the medical field. How-ever, failures of medical AI could have serious consequences for both clinical outcomes and the patient experience. These consequences could erode public trust in AI, which could in turn undermine trust in our healthcare institutions. This article makes 2 contributions. First, it describes the major conceptual, technical, and humanistic challenges in medical AI. Second, it proposes a solution that hinges on the education and accreditation of new expert groups who specialize in the development, verification, and operation of medical AI technologies. These groups will be required to maintain trust in our healthcare institutions.

In this paper a case study on improving requirements engineering is presented. Improving requirements engineering was initiated in a department at Ericsson Eurolab after an analysis had shown that many of the problems in software development had their root cause in insufficient understanding of the customer and in unclear requirements. A method-in respect to Tom Gilb who supported us in this field-called Gilb Style was introduced. This method focuses on: quality (or non-functional) requirements; quantification; strict separation between design and requirements; constraints and assumptions. After a year of experience and several projects using this method, the findings are presented. The biggest benefit in using this method is a change of culture towards requirements. This change is not limited to requirements specifications for software but includes e.g. requirements on internal service functions. The common understanding of requirements has drastically increased. Several positive side effects include more effective inspections, introduction of weekly reviews, simpler definition of test cases. The biggest problems were communication problems to (internal) customers, who did not participate in the introduction of the method.

CoDecide is an experimental user interface toolkit that offers an extension to spreadsheet concepts specifically geared towards support for cooperative analysis of the kinds of multi-dimensional data encountered in data warehousing. It is distinguished from previous proposals by direct support for drill-down/roll-up analysis without redesign of an interface; more importantly, CoDecide can link multiple views on a data cube for synchronous or asynchronoous cooperation by multiple analysts, through a conceptual model visualizing the problem dimensions on so-called tapes. Tapes generalize the ideas of ranging and pivoting in current data warehouses for the multi-perspective and multi-user case. CoDecide allows the rapid composition of multi-matrix interfaces and their linkage to underlying data sources. A LAN version of CoDecide has been used in a number of design decision support applications. A WWW version representing externally materialized views on databases is currently under development.