A model is not what it represents but captures one or more aspects of an object just as a toy model car captures the form of a real car. A model is an interpretive description or representation of an object and is itself an object apprehended by a thinking object. In effect a model is our attempt at making sense of the world around us. In our book Doing it Differently we have used a ‘sense-making’ model called Cynefin proposed by Kurtz and Snowden.

It follows that models are inherently uncertain. A model may deliberately disregard certain aspects of the reality of an object (i.e. may be variously abstract in order to simplify or to demonstrate similarities), or there may be aspects missing from the model that are unknown. A model is therefore a necessarily simpler, partial and incomplete version of the object it represents and may rely on idealisations, simplifications and analogies.

Testable models are those which can be used to make predictions that are testable. Testable predictions are testable statements. Testable statements are those the truth of which may be established to a degree through tests that may vary from ‘untested’ to ‘very highly tested’. Tests are performed by criticism and by experiment. Criticism is the studying, analysing, evaluating, discussing and judging of the quality of a model. An experiment is a set of contrived actions on the natural physical world which have been purposely designed to assess ingeniously the quality of a model. They are purposeful actions to discover outcomes that result from processes controlled by Nature.

All models are uncertain to a degree. Where the results of a test are true then the model attains some degree of confirmation. The degree of confirmation is a quality of a model which rises with the number and ingenuity of tests that are true.

Two important qualities of the degree to which a model is confirmed is the degree to which it is true or is factual – see uncertainty. A model is empirically true (lower case t) if a statement based on a model agrees with the results of an experiment to an acceptable tolerance within a context and in accordance with the purpose of the model. Such an agreement is a necessary but not sufficient condition for empirical truth. Facts are true models. Absolute Universal Truth (AUT) is truth in all contexts – known and unknown. The number of possible contexts is unknown and possibly infinite and infinite regress is inevitable, therefore AUT can never be completely established. The falsity of a statement or prediction based on a model is testable since one demonstration that it is false, outwith an acceptable tolerance in accordance with the purpose of the model, is sufficient.

We believe that a good explanation H of evidence E in a context C is a) when H states what caused E in C and b) E can be inferred from H to a high degree in C and c) H is dependable in C (i.e. likely to be true). Inference to the best explanation is the drawing of a good explanation, which is considered as the best, from a set of specific premises. It is abduction where probable is replaced by best. Abduction is a deduction in which the propositions are uncertain.

Rigour is the strict, scrupulous adherence to rules. Logical rigour is strict adherence to the rules of logic. Mathematical rigour is strict adherence to the rules contained in stated axioms. Scientific rigour is strict adherence to the testing of scientific predictions. Practical rigour is strict adherence to the use of rational judgement to design appropriate systems that successfully fulfil a stated set of purposes.

Design is a model making process of creatively deciding the form and structure of a system to satisfy a set of purposes within a specific context. Design requires a duty of care. Architectural design is deciding the form of a system to satisfy human need. Engineering design is deciding the form and structure of a system to satisfy both the demands of Nature and the requirements of the architectural design.

A good design solution H at time t, is one which is fit for purpose and consists of a set of decisions taken with a duty of care which is demonstrated by explicit supporting evidence E within a context C, when a) E is the best explanation of the causal processes in H available at time t; b) H can be inferred from E to a high degree; and c) E is dependable in C (i.e. likely to be true in context). Inference to the best design solution is the drawing of a good design solution, which is considered as the best, from a set of specific premises.