Cliff Hooker is Professor of Philosophy at the University of Newcastle. He is Director of the Complex Adaptive Systems Research Group, researching foundations of self-organisation, bio-cognitive organisation - both organismic and scientific evolution-development, and sustainable development, and author/editor of 20+ books and 100+ papers across these areas plus foundations of physics. He is Director of Assessing sustainability dimensions and impacts, The Cooperative Research Centre for Coal in Sustainable Development, aiming to re-focus sustainable development around resiliency, and houses and supervises 'Sustainability Options for Australia's Future' for the Joint Academies' Committee on Sustainability, National Academies' Forum. He teaches these interdisciplinary ideas to engineering, psychology, medical and business students as well as philosophy students.

Why integrated policy is inherently hard
by Cliff Hooker

Session 1: Questions/discussion

Since there was a bushwalking picture of me on the program, I brought along a serious bushwalking picture of me standing at the Bibbulman Track, Western Australia, which the serious bushwalkers will know. I thought that was entirely appropriate.

(Click on image for a larger version)

My background has been interdisciplinary from the time I started in university. I started as a fusion physicist, I went to philosophy and I now spend more than half my time teaching biocognitive organisation and sustainability to engineers and cognitive scientists. What I decided I would try and contribute here, amongst the huge range of possible topics, is something fairly technical but I think crucial: underneath all of the difficulties of communication, of diverse interests, of political power and so on, there lie a set of very hard intellectual problems whose resolution full multi-disciplinary integration requires, and I have taken as my task the attempt to sketch those.

And since – following Dr Monk's talk – I was brought up from a callow youth on diagramming arguments, which is what one tradition of philosophy has done for at least 150 years, I diagrammed my overall point in a quick summary of the issues:

(Click on image for a larger version)

So the claim is the following. On your left in the slide is the sequence of stages through which every standard integration problem has to go – population problem, environmental problem, you name it. (Step 0) You start with a problem that you want to address.

(Step 1) In science you build a model of the world (or the relevant aspect of the world, since you don't want to model more than you have to), that tells you how behaviour will occur. Sometimes, of course, you will have to put uncertainties and risks and so on into the model; sometimes there will be big gaps called ignorance. But you aim to describe how the dynamics of the world works.

(Step 2) You then want some process for evaluating the states of that model, according to whether the outcomes – for example, the final states – are good or bad, but also the states along the way. For instance, if you have to go through Hell along the way, it had better be because the final state is more valuable.

(Step 3) You then want to do policy analysis. You want to say to yourself, 'Well, given the kind of possible trajectories available to me according to the model, and given my valuation of the worth of those trajectories, how shall I set policy so that I start from where we are now and finish up somewhere valuable?'

(Step 4) And finally, of course, you want to put policy into practice. It must capture the commitment, or at least the acquiescence, of governments, bureaucrats, businesses, diverse affected public groups and so on if it is to be effectively implemented.

What I want to say, basically, is that at each stage along here there is an integration task, and that integration task is not only culturally and procedurally difficult, it is inherently difficult theoretically. And (unfortunately) I think this conference would not be well served, were we to imagine that all of our problems were merely 'external' and socio-cultural.

(Click on image for a larger version)

So here is the thesis. Each of the component tasks is inherently hard. The first one is: how do you achieve an integrated model? The second one is: how do you achieve integrated evaluative assessment of the states of that model? The third is: how do you achieve integrated policy analysis – that is, assessment of strategies? And there is a fourth task: how do you integrate those policy proposals into the larger policy context, the larger governance context in which you are operating?

In this talk, I am going to concentrate on just the first three integration tasks, the 'internal' ones. Every slide hereafter will tell you what the integration task is, tell you why it is hard, make one or two other comments and pass on.

(Click on image for a larger version)

The integration of models is the bringing together, in a single set of interactions, all of the significant relationships that affect the reality that you are studying.

For example, there is a nice little study of the dike structure at Nyngan by Barry Newell and Robert Wasson that integrates the hydrodynamics and climate models that produce flooding with the engineering structures – the structure of the dikes – and with the reaction of the community to those dikes, and which explains why, over the course of 100 years, the dikes have gotten higher and the problem of flooding has gotten increasingly severe, despite the fact that at every stage it has been claimed that the problem has been solved. The model that explains those things by bringing together the interaction of human factors in the community with the engineering factors and with the hydrological factors, so at least three – and, in fact, about five – disciplines have to talk to each other to build that single simple interaction model. 

So we get partial models because we have got partial expertise, partial interests, partial access – we don't have all the data. The result is that very often we cannot inter-connect them, even with a good will, let alone with insufficient attention to communication. Secondly, multidimensionality brings its own complexity of analysis, one aspect of which is illustrated in mapping multi-factor arguments. But anyone who has done computerised mathematical modelling will know how complex multidimensional modelling quickly becomes practically (and sometimes in principle) intractable. The problems increase exponentially, as they say.

(Click on image for a larger version)

However, in addition to that, there are three inherently difficult problems such models can throw up. First, typically different disciplines or sometimes subspecialties will produce models on quite different spatial, and/or temporal and/or compositional scales. So, for example, the town of Nyngan typically makes developmental decisions on a less than 10-year scale. The floods occur on a 20- to 50-year scale. So the hydrologists and the sociologists just don't see eye to eye on spatial scale, for starters. Nyngan forgets what happens in 50 years, which is why the problem has been getting worse and worse. Integration across different scales may require significant alteration of model dynamics, hence also of generation and test methods and so on, and will thus often prove inherently difficult.

Secondly, we have developed different kinds of models for human affairs and for natural affairs. Physicists like me like differential equations and other such models, sometimes finite difference equations. But in human affairs we use reasoning and decision theoretic models, for example, which have a completely different logical structure – and I do mean that technically; the structural differences are quite profound. (For instance, logic has no inherent dynamics, indeed no inherent timescale, is inherently local and monotonic, hence not self-transformative, and so on.)

Thirdly, of course, often in human affairs we use directly evaluative models, whereas the scientist is stuck with empirical models. The people of Nyngan primarily approve or not of flood damage, they don't simply form empirical beliefs about it. They wish to place sharp constraints on it, for instance an ethical constraint that it not endanger human life, not follow its continuous variation with dike height. We have to bring together those processes and understand how they interrelate.

(Click on image for a larger version)

Suppose we have got our integrated model, like the little model of Nyngan. Now we have to figure out how to evaluate a state of the model – for example, a flow of the river, the dikes of a certain height, the township having built down onto the floodplain and, inevitably, a flood coming in somewhere between 20 and 30 years that will knock out a lot of the town structure. How do we evaluate those states?

Clearly, we have to compare apples and oranges here. We have to compare, for example, something as simple as, 'Is it a good idea to put off the floods that really hurt you but have them hurt you more when they do arrive? Or would we be better off having lots of small floods that we can tolerate but preserving our basic infrastructure?' There is here not only a comparison of differing welfare (economist's sense) states but also a deep question about how to compare two different kinds of risk values.

Well, the way we normally do it is using economic tools. The market is our primary way of forcing apples/oranges comparison by simply saying, 'See how many buy apples, see how many buy oranges and we will know how much the community values one compared with the other,' even though the economist says, with a smirk, 'We will never directly compare, between individuals, their values' – no interpersonal comparison of utilities.

The slide provides several other difficulties. Ethics doesn't make price-wise trade-offs. How do we incorporate ethical constraints? We have to make certain very complex judgments in economics that are usually withheld from us publicly, like defining social welfare, choosing surrogates. These are crucial – for instance, I can manipulate any cost-benefit analysis to achieve nearly any outcome you wish by just changing two surrogates and/or the discount rate. So these choices are clearly value-laden decisions – how are they properly made? And so on. The slide summarises the difficult data and decisions required to apply net present value economic methods.

(Click on images for larger versions)

The third task perhaps you won't appreciate. I will make a single point about it.

Where there is substantial uncertainty, e.g. affecting future options (as with technological uncertainty), you cannot optimise over strategies because you don't know them. So you cannot simply choose the policy that has the best predicted outcome. You must therefore abandon the economist's optimisation in favour of satisficing – that is, doing well enough across as wide a range of future contingencies as practicable. And how you do well enough across contingencies is to build adaptive resources so that when you learn what the future is like, you can respond to it, because you have the resources. Governments and businesses understandably tend not to like that, they prefer to get it right to avoid any appearance of inefficiency, but nature invests heavily in adaptive resources despite initial costs (you are a splendid example) and, better, it often finds ways to so build them into its ecological designs that it synergistically expands system capacity. More's the pity we scarcely know how to design similarly.

The idea of a Robust Adaptive Strategy is a theory of decision making that is only just emerging now. Indeed, I think it is at the heart of sustainability. [For more, see the Industry Discussion Papers at www.ccsd.biz/research/project1.2.cfm.]