The Bridges Final Report contains an account of the research (accessible through www.mcrit.com/bridges) - what we did and how: the chemical composition of the drug and the methods used to synthesise it from a range of different plants and herbs. The main language used in Bridges research was neither plain English nor plain any other, but cryptic C++ coloured with tinges of Visual Basic and a lot of transport modelling jargon. This cocktail has meant that the final reporting stage was a rare literary exercise in removing unnecessary sophistication and going back to simple English words.
In fact, the paramount goal of developing Bridges software technology can be stated in simple terms: "Empowering policy-makers -in particular those responsible for European transport policies- with friendly and productive access to advanced decision-making tools, such as transport models and harmonised databases". In order to do this, we developed a number of highly specialited tools that missing in the industry (data formats, communication protocols, component routines, expert systems, user-interfaces, etc.) required to build up open and scalable multi-software systems useful supporting transport decision-making. Before looking at these tools in detail in the Final Report (exploring the what's and how's) it would be a good idea to "catch the bull by the horns" for a few moments and explore a little the question the scientists often try to escape: the why's.

In a way, it is not surprising at all that most potential beneficiaries of support systems integrated with Bridges research outputs -European policy-makers and experts- tend to consider the support system beverage too heavy, even if it could presented in a nice, attractive bottle, a user-friendly bottle, so to speak (or precisely because of the contrast between the promise of this beautiful bottle and its actual contents: a gurgling potion of rare changing colours). Our prophecy: "all drinkers are going to be empowered to interact with advanced knowledge-tools by themselves" seems, if anything, a dangerous scenario because it is a threat to the current status quo. Potential users of advanced support systems have the right intuition when they consider innovative software systems, such as those we are interested in, as dangerous as powerful drugs; like drugs, software systems induce consumers to accept new ways of thinking and may lead to organisational changes in institutions; software support tools induce a permanent addiction, which is anything but soft.

And Bridges outcomes are not simple tools for doing the same old things easier or faster; Bridges tools make it possible to do new things (e.g. users can put their hands on advanced forecast models and desacralise them, users can manage large databases with friendly GIS support and realise how little we know about anything...), and because of this capacity, most information and communication technologies represent a personal challenge and a cultural threat (e.g. removing low-added value "intermediaries" between policy-makers and scientific assessment tools).

So let us be crystal clear: Bridges tools are not envisaged as simplifying our working life, they are planned to make analysis and decision-making a more scientific-oriented process. A genuine willingness to open the door to the improvement of present decision-making processes, is necessary to implement with success scientific-support tools such as consistent evaluation models, advanced forecast models and acurate databases. And a genuine willigness of policy-makers and policy-analysts to understand the complexity of the problems they are dealing by with is required to make useful any software systems interfacing them with the scientific-tools. There is the hope that new generations of scientific-support tools (both advanced and friendly) and the increasing demand from society (requiring "convincing" reasons to get the approval of any new policy action) may induce policy-makers and policy-analysts to accept the challenge of learn and use scientific-support as much as they can.

At the end, the fundamental discussions with regard to decision-support systems is not about computers and software, but about people and how they are organised within institutions to take decisions.

We have to start looking for an answer many centuries ago, when kings asked prophets to guess the future: "what will happen if I declare war on my enemies?" was in those remote days the FAQ -most frequently asked question- among policy-makers. In those conflictive times, all prophets in the business pretended to have a direct link to God, the only who knew the "book of the future" because he wrote it at the beginning of history. So, the only way prophets could guess the future was by "divine revelation". Instead of computers, mathematical formulations, theories of nature, empirical validations, prophets used the most effective method: "asking God", the one who knows everything. And prophets had the monopoly on this special relationship.

The desire to know the future (the consequences of our present actions) is deep inside any human being, since the arrow of time never goes back. Divine revelation is, needless to say, a very strong prediction method because, for believers, a revelation is a self-fulfilling prophecy. The best predictions must be self-fulfilling: because it was predicted, it will happen. All in all, the paramount merit of a good oracle is his own reputation, which is obviously based on a fiction (a direct relationship with God). Intelligent kings promoted this fiction when they could dictate that the oracle's predictions be in favour of their decisions. Stupid kings supported the fiction as well, since they were unable to realise that God was, in fact, invented by the prophet, and that there was no "book of the future" at all. But, who is brave enough to tell the King he is wearing no clothes? Certainly not the prophets themselves: the best prophets were like today's lawyers, not so crazy as to give conclusive results, to say "yes, you are totally naked" or "no, you are not naked", they rather prefered cryptic sentences with double meanings, confusing metaphors and paradoxes, such as the Delphic oracle's splendid "the Greeks will win" prediction for a battle between Athens and Sparta, two Greek cities. Losing the "book of the future" was a sad moment: anything could happen, and even worse, at random. There is no destiny. When Nietzsche announced the "death of God" human beings were then completely alone, immersed in the "unbearable lightness of being", in the words of Milan Kundera.

By the Middle Ages, more professional prophets begin to develop more sophisticated "guessing methods", models to simulate what would happen if..., inventing new "protoscientific gods", so to speak, and improving their prediction capabilities over time by simply checking the accuracy of their results and trying to provide global rational explanations for particular facts. Of course, the new methods were also kept secret, in monasteries high up in the mountains or deep inside dense forests: kings are always more influenced by the revelations of a fictitious god (which do not require rational explanation for anything) than by those of protoscientific gods (which require all-too-much explanation for the slightest little detail). And knowledge progressed slowly: what professional prophet worth his salt is willing to hand over the secrets of his success to his competitors?.

Descartes was responsible for emphasising the merit of using rational "methods", Galileo and Newton were the first to develop objective methods, empirically tested, for predicting the easiest part of nature: physics, the world of dead bodies in the range of ordinary speeds. A few centuries later Darwin provided apparently solid rational explanations for evolutionary biology (but no a method to predict the evolution of species over the next millennium, and still nowadays the so-called "Creationists" win legal battles against "Evolution" because without prediction a scientific theory cannot be totally proved).

In broad terms, a scientific model is just "an abstraction of some slice of reality that the model claims to represent". Most of theoretical science relies upon so-called "mathematical models" to mirror the observable quantities in the real world. The scientific mathematical model "encodes" or "compresses" real-world objects and phenomena into abstract mathematical objects. There are two quite distinct categories of models: the empirical and the explanatory.

The empirical models involve examining past trends and attempting to create a predictive scheme on the basis of statistical methods. The explanatory approach constructs models not so much based on statistics as on the laws of nature (hydrodynamics, thermodynamics, astronomy, and all the principles of physics and chemistry, etc.). Interestingly enough, a better explanation does not necessarily produce better predictions, and vice-versa. While evolutionary biology is able to explain the evolution of life but not to forecast, for instance, the emergence of new species, quantum mechanics provides amazingly accurate predictions but little explanation of, for instance, what "quarks" actually are. In short, a hypothesis or theory is not necessarily more certain when its predictions are confirmed.

How much are scientific models able to explain and predict?. J.L Casti has proposed a cross-table where most scientific domains are located. Celestial mechanics, weather, chemistry, all provide relatively accurate predictions and solid explanations; Quantum mechanics good predictions but virtually no convincing explanations, just the opposite to Evolutionary biology. Needless to say, social sciences such as economics provide both bad predictions and explanations.

There is a growth of complexity, from physical systems (such as astronomy) to natural systems (biology) and social systems (economics). Complex systems tend to be organised as "networks", where its elements have a high volume of dynamic interrelation with others. Understanding a Complex System requires analysing it simultaneously on different scales, since the relations between elements and scales are far more important that the single elements themselves. Conventional reductionists, element-per-element, analysis does not work for an understanding of the emergence of stable, periodic, complex or chaotic behaviours of a whole system. There is an inherent unpredictability in chaotic processes: their evolution in time could be "computationally irreducible".

In relation to human behaviour, scientific methods are still far from even convincing explanations. (e.g., mainstream economics is still based on assuming ideal people, perfectly informed, taking rational decisions to optimise their welfare, when we are not like that at all: real people are never well-informed, take intuitive decisions to satisfy minimum constraints and adapt these decisions over time based on their perception).

In the second part of the 20th century, the computerisation of methods (converting methods and mathematical models into "computer models") gave scientists, even social scientists, a bonus of often undeserved prestige. Based on non-linear dynamics models, programmed on huge mainframe computers at M.I.F., Prof. Forrester predicted the "end of world" in the early sixties. Gosplan officers planned the Soviet Union's economy based on the largest Linear Programmes known, able to produce "optimum" quantities for all goods to be produced by each factory and consumed by each group of consumers in the economy. This gave a wonderful sense of perfection to Soviet planners: they had a superior ethic -give to everybody according to their needs, ask of everybody according to their capacity- and perfect computerised models to calculate quantities with full precision. Since the beginning of the century, thinkers and writers have attempted to envisage such an ideal platonic governor, a computer filled with all the data of the world running under self-learning intelligent software, like a Laplacian demon.

But computers represented a major breakthrough in science, and the results they produced after long computations were often surprising. Before computers, non-lineal components had to be neglected to derive partial solutions from large systems of non-lineal differential equations. Nobody really knew "how much" it meant to lose these non-lineal components, until Lorenz discovered from looking at computer printouts his very famous "butterfly", a strange attractor with a butterfly-like shape: very sophisticated models can produce stable and periodic results as well as simple models can produce complex and chaotic ones. For instance, an iteration on a logistic equation produces chaotic behaviour depending on the value of the parameter: Small changes in the parameter can produce any behaviour.

Automatised computation helped us to see the impact of minor changes in initial local conditions on the solutions over time. This was solved analytically by Lyapunov, but only thanks to computers could it actually be tested. Complex Systems are unpredictable because their sensitivity to initial conditions is often higher than our capacity to measure them properly. Therefore, "simulation" is needed to check the scope of this uncertainty. All considered, science is changing from "looking at regularities, stability and periodicity" to "looking at complexity and chaos". This implies a cultural breakdown, a much less optimistic approach towards "prediction" but a richer and wiser interest in "understanding".

But even this is not new. Amazing studies by Gödel had already demonstrated the impossibility of a perfect and formalised mathematics. There are numbers so complex that no computer can generate them. Computers are not perfect, infallible machines. However, in 1976, Kenneth Appel and Wolfgang Haken produced the first mathematical proof using computer simulation for the so-called "Four Color Cojunture": The computer checked one by one the 1,936 special configurations into which Appel and Haken demonstrated (analytically) that the problem could be simplified. But a Cray supercomputer has a ratio of one undetected error per thousand hours of operation. Then, "Is a mathematical proof a proof no one can check?".

We now know that computer models are not so powerful as divine models, but they are the best imitation available. And they have produced a major cultural breakdown in human history, according to to Gregory Bateson, since this is the first time there is a way to encapsulate "human knowledge" outsite a human mind: a computer model is therefore a priviliged intermediare between humans thoughts, a neutral "inteligent" device able to encapsulate human rationaliy free from any emotion or intuition. Outsite a human mind, only computers and gods claim to have such a rational attributes. We say that God creted as "similar to God", and we created computers "similars us". Modellers and computer specialists have become like prophets, the only ones able to talk to computers in the same way the prophets were the only ones with access to God: "the computer has spoken". Still many people tend to see computers as sacred oracles: it is hard to realise that these small gods are in fact just electronic devices, until now able to manage data and even information by themselves, but hardly able to generate new knowledge without interaction with human beings; even if computers can learn without human supervision, they always would lack the moral and ethical component, the emotion and compassion, needed to take fair and right decisions. Computers may represent perfect and objective rational investigation, but there is a need to balance this cold scientific vision with subjective human feelings and emotional inspiration, and to accept a set of arbitrary moral commands (ideally stated by divine revelation), to finally take right decisions.

It is a fact that during the late 20th century many decision-makers and citizens still feel intimidated by computers in the same way the ancient kings feared the wrath of the gods. But at the same time, many others feel strong disappointment with computer models, since they tend to become more and more specialised, and then put aside crucial questions, use overly complex formulations given the poor data available and reach obvious or well-known answers. The ancient prophets pretended to know everything, they told kings how easy things would be if they followed their advice, but now scientists claim not to know things even in their own narrow field of expertise. The more a scientist researches, the less he says he knows. "I only know that I know nothing", says Socrates. All together, the present gap between scientists and policy-makers is an updated version of a very old story of cultural misunderstandings between the "owners of the knowledge" and those "responsible for using it".

Despite their limitations, in contemporary so-called "mature democratic societies" rational and objective methods (therefore scientific methods) play an increasing important . Policy-analysts cannot escape from using computer models, the only objective and transparent analytical method they have. The alternative to science is clearly much worse: policy actions based just on principles and subjective intuitions. Because decisions have to be accepted by different people with contradictory goals and interests, they cannot any longer be imposed, and only rational explanations provided by objective methods will serve as a convincing argument at a negotiation table. Even if the rational explanations are partially wrong, the predictions rather inaccurate -but still realistic-, objective methods is needed to help the definition and implementation of political decisions. If mature societies are to be democratic, ordinary people have to be empowered to move from protective positions (often related to ignorance and fear of the future) to a common synergies and interests. Only the well-informed and kwnoledgble policy-makers are able to move from positions to interests and succeed in complex negotiations.

The more secular (or to put it differently, post-religious) 21st century, will in all likelihood be an amazing time for professional scientists, since neither decision-makers nor their constituencies will be as naďve (either in what they reject or what they accept from scientific models) as they were in the 20th.. Even computers are losing their divine character, becoming simple communication and learning devices, a form of interactive TV screens connected to telephonic lines. New generations of people are interacting much more naturally with computers and models, and are more willing to put their hands on computers and use them for their own purposes -when they are useful, and not vice-versa.

Today, and in the coming years, empowering the capacity of decision-makers will require more than reliable "predictions" of the impact of their decisions: Decision-makers have to establish an interactive dialogue with computer models (and the experts developing the models) in order to increase their personal knowledge, enrich their personal intuitions and communicate decisions efficiently (not to mention, improving the models!). It is by aggregating data into meaningful indicators, by changing the model parameters, representing alternative policies and visualising the impacts in time, that the policy-makers and policy-analysts can actually understand the complexity of the current transport system, its sensitivity to apparently minor policy changes, its strong inertia forces and interdependencies, often leading towards counterintuitive behaviour patterns.

All considered, what really matters today is helping decision-makers to adopt "scientific ways of thinking": decision-makers have to be encouraged to reject prophecies by "black-box" models, by any cryptic model demanding a "religious-like" acceptance based solely on the "prestige" of the modeller. More than the results obtained, it is the method of knowledge applied that matters. On the other hand, it is always important to remember that models in social sciences -in particular in transport- are very weak in terms of both prediction and explanation in relation to other scientific fields (e.g. weather forecast or evolutionary biology). This leads to two main conclusions: it is expected that further research on transport models will result in more sophisticated formulations, with higher software and computer requirements (so increasing the gap between models and policy-analysts) and no single model or scientific theory can nowadays claim to be "good enough" at modelling a transport system (so simultaneous access to many models, comparison of results, and ex-post assessment is vital). Not only is there a gap between modellers and decision-makers, there are gaps between modellers themselves: Scientific publications and world conferences are full of theoretical papers providing little help (or none at all) for an expert reader trying to understand in depth and apply in practice the proposed modelling innovations. Simply by reading or listening, an advanced computer model cannot be properly understood or duplicated.

More friendly access to many increasingly sophisticated computer models cannot be gained simply by developing user-friendly interfaces, implementing new algorithms capable of providing faster results or by deconstructing large models into many complementary and meaningful submodels capable of taking full advantage of more recent innovations in software and computer technologies. Even if these strategies are in any case necessary optimisations, more friendly access to decision-support models must be achieved by building up integrated and open systems of reliable and consistent models driven from many personalised user-friendly interfaces, and capable of supporting policy-makers' decisions. The process of building such a decision support system requires excellent understanding of the policy problems to be assessed, deep knowledge of modelling techniques, database management expertise and a high level of software and computer tool skills. Fields which are closely interlinked.

After almost fifty years of research and experience, computers today have still not been able to generate anything close to human intelligence. But they are indispensable, among many other reasons, to help people to understand and deal with the growing complexity of this world: they are excellent at "computing" (making fast calculations with data) and sorting and ordering (classifying huge volumes of data). They are becoming useful in the field of communication as well, being able to manage most kinds of data formats (moving from figures and text to graphics, maps, video and voice, creating complete virtual worlds and so on). If General DeGaulle had had even a simple EXCEL Spreadsheet and an email, he probably would not have found 256 varieties of cheese to be so many ("How can I govern a country that has 256 varieties of cheese?" he once complained).

We hope this long answer -which raises a lot of additional questions-, beginning with prophets and oracles, leading to software tools for modelling social sciences and empowering decision-making, helps to understand why ordinary people, like us, derived so much enjoyment from being involved in such abstract and instrumental research as Bridges, as well as why we believe ordinary people can enjoy using our research outcomes for their own purposes. Yes, the aim of the research is the development of tools, but tools that must be used by people. All considered, we had good reasons to work very hard on this project and, we believe, EC/DG TREN also had good reasons to support us financially and scientifically in doing so.

Looking back, it is fair to say that is was a truly exciting undertaking.