Biblio

Explaining reasoning and behaviour of artificial intelligent systems to human users becomes increasingly urgent, especially in the field of machine learning. Many recent contributions approach this issue with post-hoc methods, meaning they consider the final system and its outcomes, while the roots of included artefacts are widely neglected. However, we argue in this position paper that there needs to be a stronger focus on the development process. Without insights into specific design decisions and meta information that accrue during the development an accurate explanation of the resulting model is hardly possible. To remedy this situation we propose to increase process transparency by applying provenance methods, which serves also as a basis for increased explainability.

One essential requirement for supporting analytics for Big Medical Data systems is the provision of a suitable level of traceability to data or processes ('Items') in large volumes of data. Systems should be designed from the outset to support usage of such Items across the spectrum of medical use and over time in order to promote traceability, to simplify maintenance and to assist analytics. The philosophy proposed in this paper is to design medical data systems using a 'description-driven' approach in which meta-data and the description of medical items are saved alongside the data, simplifying item re-use over time and thereby enabling the traceability of these items over time and their use in analytics. Details are given of a big data system in neuroimaging to demonstrate aspects of provenance data capture, collaborative analysis and longitudinal information traceability. Evidence is presented that the description-driven approach leads to simplicity of design and ease of maintenance following the adoption of a unified approach to Item management.