The Role of Virtual Reality in Human-AI Interaction (HAII)

A growing number of daily-life activities are surreptitiously outsourced to intelligent systems, powered by the latest advancements in the field of Artificial Intelligence (AI): from smart spam filters (e.g., Gmail) and online shopping recommendations (e.g., Amazon), to ride-sharing applications (e.g., Uber) and smart personal assistants (e.g., Alexa); a great deal of daily chores are streamlined by AI. In other words, we have already begun to trust AI in making decisions on our behalf about whose messages we receive, from whom we buy, with whom we share a ride, and even from where we inform ourselves. In fact, AI is set to automate increasingly fundamental daily-life facets, such as commuting (i.e., autonomous driving), and even law enforcement (e.g., “coplink”).Consequently, the deeper AI weaves itself into the fabric of our societies, the greater the volume of training datasets that are required, and the more diverse and multidimensional they must be. However, certain training datasets are not immediately available; plus they have to be collected in an ecologically-valid fashion. For example, achieving fully autonomous driving (level 5), inevitably assumes the capacity of AI to perform ethical decision-making (i.e., The “Trolley Dilemma”). Data on ethical decision-making is scarce, and cannot simply be aggregated in an ecologically-valid manner by traditional means (e.g., surveys).Virtual Reality (VR) emerges as a viable alternative for amassing diverse datasets for tra...