When I was studying physics, I was introduced to the idea of ‘system of interest’: defining at the outset that ‘thing’ you were immediately interested in learning about. There are three ideas to be developed from this starting point. First, it is important simply to list all the components of the system, what Graham Chapman called ‘entitation’; secondly, to simplify as far as possible by excluding all possible extraneous elements from this list; and thirdly, taking us beyond what the physicists were thinking at the time, to recognise the idea of a ‘system’ as made up of related elements, and so it was as important to understand these relationships as it was to enumerate the components. all three ideas are important. Identifying the components will also lead us into counting them; simplification is the application of Occam’s Razor to the problem; and the relationships take us into the realms of interdependence. Around the time that I was learning about the physicists’ ideas of a ‘system’, something more general, ‘systems theory’ was in the air though I didn’t become aware of it until much later.

So, always start a piece of research with a ‘system of interest’. Defining the **system** then raises other questions which have to be decided at the outset – albeit open to later modification: questions of scale[1]. There are three dimensions to this. First, what is the granularity at which you view your system components? Population by age: how many age groups? Or age as a continuous variable? Usually too difficult. Second, how do you treat space? Continuous with Cartesian coordinates? Or as discrete zones? If the latter, what size and shape? Thirdly, since we will always be interested in system evolution and change, how do we treat time? Continuous? Or as a series of discrete steps? If the latter, one minute, one hour, one year, ten years, or what? These choices have to be made in a way that is appropriate to the problem and, often, in relation to the data which will be needed. (Data collectors have already made these scale decisions.)

Once the system is defined, we have to ask a question like: how does it work? Our understanding, or ‘explanation’ is represented by a **theory**. There may be an existing theory which may be partially or fully worked out; or there may be very little theory. Part of the research problem is then to develop the theory, possibly stated as hypotheses to be tested.

Then there is usually a third step relating to questions like: how do we represent our theory? How do we do this in such a way that it can be tested? What **methods** are available for doing this?

In summary, a starting point is to define a system of interest, to articulate a theory about how it works, and to find methods that enable us to represent, explore and test the theory. We can call this the **STM approach**.

This, though the reference to theory (or hypothesis) formulation and testing, establishes the **science** base of research. Suppose now that we want to apply our science to real-world problems or challenges. We need to take a further step which in part is an extension of the science in that it is still problem-solving in relation to a particular system of interest but has added dimensions beyond what is usually described as ‘blue-skies’ science. The additions are: articulating objectives and inventing possible solutions. Take a simple example: how to reduce car-generated congestion in a city. the science offers us a mathematical-computer model of transport flows. Our objective is to reduce congestion. The possible solutions range from building new roads in particular places to improving a public transport system to divert people from cars. Each possible solution can be thought of as a **plan** and the whole activity is a form of **planning**. In some cases, computer algorithms can invent plans but more usually it is a human activity. For any plan, the new flows can be calculated using the model along with indicators of, for example, improved traffic speeds and consumers’ surplus. A cost-benefit analysis can be carried out and the plan chosen that has the greatest rate-of-return or the greatest benefit-to-cost ratio. In reality, it is never as neat as this, of course.

How can we summarise this process? I learned from my friend and collaborator. Britton Harris, many years ago that this can be thought of as **policy, design and analysis**: a PDA framework to complement the STM. His insight was that each of the three elements involved different kinds of thinking – and that it was rare to find these in one person, or applied systematically to real problems. There is a further insight to be gained: the pda framework can be applied to a problem in ‘pure’ science. the objective, the policy, may be simpler – to articulate a theory – but it is still important to recognise the ‘design’ element – the invention that is required at the heart of the scientific process. And this applies very directly to engineering of course: engineers have both a science problem and a policy and planning problem and both STM and PDA apply.

There is an important corollary of adopting a systems perspective at the outset of a piece of research: it forces interdisciplinarity. This can be coupled with an idea which will be developed more fully later: requisite knowledge. This is simply the knowledge that is required as the basis for a piece of research. When this question is asked about the system of interest, it will almost always demand elements from more than one discipline; and these elements combine into something new – moiré than the sum of the parts. There is a fundamental lesson here about effective research: it has to be interdisciplinary at the outset.

This provides a framework for approaching a subject, but we still have to choose! These decisions can be informed but are ultimately subjective. A starting point is that they should be **interesting** to the researcher but also **important **so some wider community. The topic should be ambitious but also feasible – a very difficult balance to strike.

Alan Wilson, March 2015

[1] Scale questions define disciplines and subdisciplines: quantum to cosmology; ethnography and psychology to social policy.