“Dave...I can assure you ...that it’s going to be all right ...” A Definition, Case for, and Survey of Algorithmic Assurances in Human-Autonomy Trust Relationships

摘要

As technology becomes more advanced, those who design, use and are otherwiseaffected by it want to know that it will perform correctly, and understand whyit does what it does, and how to use it appropriately. In essence they want tobe able to trust the systems that are being designed. In this survey we presentassurances that are the method by which users can understand how to trustautonomous systems. Trust between humans and autonomy is reviewed, and theimplications for the design of assurances are highlighted. A survey of existingresearch related to assurances is presented. Much of the surveyed researchoriginates from fields such as interpretable, comprehensible, transparent, andexplainable machine learning, as well as human-computer interaction,human-robot interaction, and e-commerce. Several key ideas are extracted fromthis work in order to refine the definition of assurances. The design ofassurances is found to be highly dependent not only on the capabilities of theautonomous system, but on the characteristics of the human user, and theappropriate trust-related behaviors. Several directions for future research areidentified and discussed.