Animal-computer interaction : designing specialised technology with canine workers

This thesis reports on research underpinning the design of a canine-centred dog-to-human communication technology, specifically an emergency alarm system that enables trained assistance dogs to call for help on behalf of their owners. Thousands of vulnerable people worldwide living with conditions such as epilepsy, diabetes or limited mobility, rely on assistance dogs to help them in their daily lives. When, for various reason, the human becomes incapacitated, such as when they are experiencing an epileptic seizure, have fallen, or have gone into a hypoglycaemic coma, it is down to their dog to take action to resolve the situation. Interactive technology can provide an assistance dog with the means to raise the alarm and summon help, but in order to enable them to independently and successfully engage with an alarm, it is critical that they are able to make sense of when and how to use the device to increase their chances of successful interaction. Thus, the research presented here aimed to understand the factors that might influence the dog’s ability to successfully interact with the system we undertook to design. Our initial design was informed by various biological, cognitive, and ergonomic considerations of dogs. We then elicited specific requirements for a canine emergency communication system by observing training practices to learn how trainers communicate with the dogs; interviewing human-dog partnerships to understand their needs; and engaged in rapid prototyping sessions with the dogs to identify their preferences. Using these requirements, we developed several high-fidelity prototypes, which we tested with assistance dog users and their handlers, to identify which design features might best facilitate the dog’s interaction with the device, and in turn enable the design of the training process through which the dogs learn to use the device as independent agents. This led to the practical observation that for many assistance dogs, using an interface that allows them to bite an attachment with their mouth and tug it until it detaches was easy for them to learn to use. We found that when designing technology for assistance dogs, researchers need to consider to what extent the dogs might be expected to drive the interaction and that researchers need to design not only to support the interaction itself but also to facilitate the training process that will eventually lead to the dogs being able to interact with the technology.