Dare2Del—An Explanatory Interactive Learning Assistant to Support Individual Strategies to Identify Irrelevant Digital Objects

Abstract Digital hoarding can lead to rising economical, environmental, as well as personal costs. In private as well as in working life, a large amount of files stored on personal desktops as well as in the cloud are outdated and are never again opened. However, irrelevant files are rarely deleted because employees fear that decisions to delete might be wrong and cannot be revoked. Furthermore, such decisions might require a lot of cognitive effort. In an interdisciplinary collaboration between artificial intelligence (AI) research and work psychology, we developed and empirically evaluated an intelligent assistant system to support identification of irrelevant files. The AI assistant Dare2Del combines explicit knowledge representation and machine learning. The core of the system is realized in Prolog and Inductive Logic Programming (ILP). To allow for human control and oversight, Dare2Del relies on explanatory interactive methods. Users can ask for explanations why a file is considered irrelevant by the system and they can correct system classifications as well as explanations. Both types of information are used for model adaptation. In an experiment, we could show that persons with low cognitive inhibitory control benefit from Dare2Del in terms of mental effort.