Augmented reality (AR) applications have gained prominence due to recent advancements in both software and hardware solutions. This paper advocates the use of augmented reality agents (AuRAs) for AR applications, by trying to answer three main questions: Why should AuRAs be researched? When are they appropriate for use? How should they be developed?

The authors motivate the need for AuRAs by considering them as both a design and interface paradigm, justifying the tradeoffs between benefits and added computational overhead. They conduct a navigation experiment in a simulated AR setting where users are instructed to pick up four products from a simulated supermarket before proceeding to the checkout. They replicate the experiment using three variations: using a traditional bubble to indicate the target; using multiple software agents visualized as arrows; and using a virtual avatar that guides the user toward its target. The study provides preliminary evidence that AuRAs prove to be a better design and interface paradigm by showing that on average, subjects take less time and travel less distance using the virtual avatar. Finally, the authors introduce the AFAR toolkit for developing AuRAs to facilitate rapid prototyping of AR agents based on the belief-desire-intention (BDI) model for agents.

In conclusion, this paper is easily accessible to readers who are not familiar with this topic and provides a fairly comprehensive review of prior work. However, the motivation of using AuRAs based on a generic agent taxonomy [1] seems rather verbose. Also, conducting the user study in a simulated AR setup and the lack of definitive evidence based on the preliminary results lessen the impact of the authors’ claims. Finally, it would have been useful to describe existing use cases (if any) of the AFAR toolkit to justify it as a standard choice for developing AuRAs.