Brains working towards machine consciousness

The principle of Object Oriented (OO) programming is that a system is analysed and modeled in terms of the objects within it. Each object encapsulates the data and methods which the object needs to interact with others, and this both simplifies design and improves the goodness-of-fit of the OO design to the system.
Generally, I find, matching the architecture of a produced system to that of the environmental system with which it interacts makes it better able to provide services to that system. So, taking the classic example of a video recorder, the OO program to control it will tend to provide better control of the player than a monolithic program, not because of any inherent mystical similarity magic, but because of the simplicity of matching software module to hardware module.
Learners in a connected world need some support - obviously they have the infrastructure in place, the internet, tools for accessing people and information. But in situations where there are hundreds, or thousands, of other voices involved in a dialogue, there are very few people who can keep track of what is being said, of where the useful nuggets of information are, without some sort of tool to support them.

Many of the models of mind which are current suggest the mind is made up of multiple 'modules' (for a discussion of some of them, I suggest reading The Philosophy of Social Cognition by Benoit Hardy-Vallee, which I had the pleasure of contributing to in a small way as editor [I note there are still a number of small typographical errors I missed, however!]). Although the models have fairly disparate stand points on what these modules are, and how they are organised, the modular nature is a common theme. And, if a good fit between model and system helps with the effectiveness of the model and the maintenance of it, this should come as no surprise - one of the key things our minds have to do is keep track of multiple agents in the real world - being modular allows the mind to better match its own architecture with that of the world around it.

Modular minds have come about to support our activities within communities (and to help us avoid predators, and catch prey) and with the advent of the new ways of working brought to us by the internet and the web, we are increasingly able to engage in communal learning. It makes sense, therefore, to design systems to support us in that learning (and in other collaborative tasks) which have an architecture similar to the mind, but also to the communities in which we work, play and learn.

Fortunately, nature has managed to make the architectures of mind and community very similar. Multiple 'modules' able to interact and communicate with one another, taking on different roles as necessary to support the activity of the system as a whole - this model is one which, in computing terms, we think of as an agent based architecture.

Dies it matter which model of the mind we choose to use as a baseline for an agent based support system? Not really - if the model is not obviously wrong in some respect, it has some value. It may be that one model of the mind is better than the others, or, indeed, may even be an accurate model - but pragmatically, if the system we design to help people work in massively multi-person environments provides them with real benefits, it does not really matter whether this is the one we based the system on.