Empirical Modelling (EM) offers both a broad foundational perspective on computing and a novel practical approach to modelling. Central to the EM perspective is an emphasis on the power of the computer to represent state - in particular, state which is easily interpretable. Adopting a broad concept of what constitutes a computer - any reliable, interpretable, state-changing device - leads to a view in which the context for computing (e.g. other devices and users) is as significant as the computer itself. The notion of programming, in a narrow 'algorithms + data' sense, then likewise needs to become broader - something more like 'configuring a system'. In order to address a new application such 'configuring' requires thorough analysis and exploration to understand sufficiently the components of the application and how they work with each other. It is this analysis and exploration for the sake of understanding which is the goal of our modelling methods. We call it 'empirical' simply because the methods and principles involved are based on observation and experiment.
We begin an analysis of a domain or application by identifying the fundamental, pre-articulate, structure of our personal experience of the domain. Our present framework for representing this structure includes identifying:
Such an analysis leads relatively directly to a computer representation of the domain in which a script of definitions corresponds to a particular state of the model. This is achieved with the aid of tools such as the tkeden interpreter and notations for visualisation and animation. Any changes the modeller makes in the definitions will propagate automatically throughout the script (hence also through the model) to maintain the dependencies - very much in the style of a spreadsheet. When the modeller is confident that certain sequences of changes give experimental results correlating closely with similar real world experiments these sequences can be automated, and in this way a full behavioural model can be gradually developed.
The typical activity in EM is therefore the use of the computer for building artefacts (models) which are distinguished by:
It is characteristic of EM that the focus is upon the construction of environments for interaction and experiment in a domain that is not yet well-understood. This modelling activity complements and supplements the conventional development of software from a relatively well-understood (albeit often changing) set of requirements.