Sensory Augmentation - a pathway to new Human Computer Interactions?

Sensory substitution is the principle by which one sense is used to receive information normally received by another. This is accomplished through use of a sensory substitution device, a system consisting of three stages: an input sensor system which collects data from the natural world, an output system which portrays this information to the human body, and an algorithm which maps the inputs to the outputs. These stages require careful consideration and their design can make a large impact to the effectiveness of the device built. My idea is to evaluate and de-convolute the impact of interface, algorithm and training on the usability of sensory substitution devices, in order to then contribute to the research area through development of a device or interface which follows the research guidance.
A thorough evaluation of these principles would allow more effective and focused design of new devices. Current design of sensory substitution devices results in very varied solutions with little common ground in their design principles. There is a distinct lack of collaboration between engineers and neuroscientists on the subject and many of the devices built suffer from design paradigm issues which limit their effectiveness and uptake.
In order to actually use a device, a user must be trained. The sensory substitution device can therefore be broken down into three design considerations, which until now have not been de-convoluted and experimented with independently: The interface, the algorithm, the training.
There have been a few reviews of sensory substitution devices however little formal breakdown of these design considerations. Isolation of the limitations that the three aspects place on devices would prove very useful to the design community.
When it comes to training, most sensory substitution devices focus on a combination of active training for very specific tasks, and passive training through use of the device in its intended context. I question whether it's possible to train a user to excel at a specific task through passive training alone, if provided with the right environment in which to learn. Ideally if a user could slowly gain the advantages of the sensory substitution device whilst performing a task they would otherwise be performing anyway, then uptake of the device for its advantages would be significant. I hypothesise that the limitations of many devices can be lifted by ensuring the data mapping process is performed in a liquid and analogue way and if there is sufficient existing sensory overlap. This approach has been successfully demonstrated to some degree, and is subtly hinted to in much new research, but has not been effectively tested and defined to the degree where it is established as a principle for sensory substitution design.
If the algorithms and principles of encoding that surround SS devices can be optimised such that "externalisation" experience can be established through passive use in task-specific environments, this would open a new pathway to data interpretation through sensory augmentation devices.