Data Science technologies are poised to take over the next phase of global digital disruption. Despite this, the general understanding of the workings, limitations, advantages, disadvantages and especially the outputs of machine learning algorithms are lacking. More specifically, up to a couple years ago the algorithms employed by Automated/Algorithmic Decision Making (ADM) software were often treated as black boxes. ADM software is fairly new and mostly based on the use of machine learning algorithms in conjunction with large amounts of data. This kind of software can be difficult to understand, but due to the significant usage of ADM technologies today, we must be able to better explain their workings across the full range of stakeholders. Focusing on how ADM is used in the public sector, this project examined the main drivers and obstacles, and how the innovation bottlenecks can be addressed efficiently.
Firstly, the current status quo and trends in data science applications in the public sector were identified. From these findings a report with recommendations for scientific libraries in Switzerland was compiled. The objective of this report is to underline the role of scientific libraries in the newly emerging big-data world and increase awareness around data quality and robust innovation practices.
Secondly, from the same research a potential user group who could benefit from the development of a Data Science toolkit was identified. For this user segment an initial toolkit design concept was developed through an iterative participatory process. The toolkit design concept offers an intuitive, accessible approach to teaching data science topics. The concept is meant to be further developed, prototyped, and employed. The application is designed to improve communication and understanding in data science to ensure timely, effective and robust innovation of ADM software applications in the public sector.
1. March 2020 – 30. November 2020