In this paper we provide some epistemological and historical remarks that concern systems engineering and modeling. In this paper we provide some epistemological and historical remarks that concern systems engineering and modeling. This is consistent with the idea that science and coscience (i.e. philosophy of science) can cooperate fruitfully. However, we think that it is confusing to mix them vaguely and that we have to separate real applications and the consideration about their philosophical implications. Thus, using a terminology that philosophers love, this work belongs to the meta-level. The match between philosophy and engineering is quite unusual and merits further clarifications. The basic issue in the philosophy of science can be introduced as follows: scientists study the world, philosophers of science study how they do that (and sometimes they also study scientists themselves). Borrowing from Lipton [1], “ I am a philosopher of science: what do I do? Here is the short version: astronomers study the galaxies; I study the Astronomers.” There has been a certain disregard by philosophers of science towards technology, which they consider a straightforward application of pure sciences.
of technology and engineering within science and human knowledge. To continue the
analogy introduced above, engineers study how to design systems, philosophers of
engineering study how they do that. The crucial difference between engineers and scientists
is that the former decide how to manufacture or produce working artifacts and systems,
while the latter analyze nature and formulate theories to explain how natural systems
work. Decision-making naturally brings engineers to think about objectives much more
than natural scientists need to do. This profiles a different kind of rationality. On the
one hand, we might believe that models are simplified versions of reality that exists
independently from our ideas about it, and that the task of science is to describe this
reality. On the other hand we might think that models do not describe reality, but they
actually create it (indeed most of the modern epistemology tells us that all observation
is theory-laden, for example see Hanson [
This debate fails to have a clear outcome within epistemology, but if the target system is artificial rather than natural, then it must have a goal, and the issue becomes clearer. What we might call the “pure problem” of scientists is “is it true?”, while that of engineers might either be “does it work?”, or, perhaps more appropriately, “does it do what the stakeholders want?” It is clear that there is a relationship between being able to verify a statement and making a choice. However, decision making and systems design have some features that make them a special case from an epistemological point of view.
“In engineering the ultimate purpose of modeling is to realize reliable artifacts
or technical processes. This contrasts substantially with the natural sciences
where, conceptually at least, the aim underlying the modeling activities is to
gain knowledge for knowledge’s sake .” [
Epistemologists, in the first half of the ’900, usually made reference to natural sciences as chemistry, biology and, most of all, physics (probably due to its resounding success). In this case, the observer is in front of a system that is given and he/she has to describe and understand it. However, in engineering the system is actually built by the observer (or one of his/her fellow humans). Thus, the demarcation criterion of natural science may be not perfectly suitable. Systems designers still have to do verifications and observations (as natural scientists) but most of all they have to make choices. They are interested in the truth of statements as much as in the effectiveness of choices. From the point of view of systems design, good models are the ones that help to split properly the domain of possible choices in good and bad ones. Some epistemologists underline the problem-solving aspect of science, for example Laudan.
“Science is essentially a problem-solving activity. [. . . ] The approach taken
here is not meant to imply that science is “nothing but” a problem-solving
activity. Science has a wide variety of aims [. . . ] My approach, however,
contends that a view of science as a problem-solving system holds out more hope
of capturing what is most characteristic about science than any alternative
framework has .” [
“For Kuhn, science is problem-solving rather than truth-seeking activity . . . .
And what would be a more striking example of problem-solving than OR! . . . As
a problem-solving activity OR is oriented towards practice: it tries to use the
methods of science to find optimal solutions to problems concerned with
alternative courses of actions. As the solutions are its primary aim, it is clear in
which sense OR is not a truth-seeking activity: it is not a knowledge-seeking
enterprise .” [
the same perspective. It has been said that “Philosophy of science is about as useful to scientists as ornithology is to birds” 1, namely that it is not very useful in practice, but we try to show that some epistemological issues arise anyhow. In our opinion, they require a consideration. At least, epistemology is useful for an external analysis of the scientific method. A scientific analysis of the scientic method would be self-referential.
“A philosopher, therefore, who knew nothing except philosophy would be a
knife without blade and handle. Nowadays a professor of philosophy very often
is a man who is not able to make anything clearer, that means he does not really
philosophize at all, he just talks about philosophy or writes a book about it.
This will be impossible in the future. The result of philosophizing will be that
no more book will be written about philosophy, but all books will be written in
a philosophical manner.” [
is relevant in epistemology. He proposes that a value is epistemic if it helps to “ promote
the truth-like character of science”. Otherwise, it is non-epistemic. Dorato [
However, for engineering and technology disciplines the role of non-epistemic values appears to be less clear. Safety, equity and economical sustainability are examples of non-epistemic values (since they do not produce knowledge) that have an important role in the decision making process concerning real systems. Engineers, who have to choose between two or more alternative models, in some cases, have to consider nonepistemic values, and integrate them in their models. It is the case of the systems we have considered in this work.
rely upon pure epistemic values. In fact, a first possibility is the experimental approach.
methodology. Another one deals with ethical values. This is not totally common. Thus, in the next sections, we present the tradition that is behind each one of these approaches.
3
Models
The root of the term model can be traced back to the Latin term modus which in turn
would derive from the Indo-European root “med-”. Its meaning is measure [
“Be sure to have a complete Model of the Whole, by which examine every
minute Part of your future Structure eight, nine, ten Times over, and again,
after different Intermissions of Times” . [
and in a more general sense as “kind of behavior” and Bacon indicates with modulus a mental copy of the real world, which is quite close to the modern use. Nowadays these terms are intensively diffused. For example, during the decade 1990-1999 there have been 17,000 publications including them in the title.
The remarkable point is that along centuries there is an interesting feature which characterizes models: they appear to be tools which help to design artifacts. Models are visions of a target system constructed respecting constraints drawn from its environment, which help the system designer/architect to conceive it. In engineering disciplines, modeling is first of all an activity that is close to design. The designing of systems and services requires both analytical and synthetic processes, because designers invent and create new artificial systems to fulfill a need. This is different from describing and understanding a given natural system. From this point of view modeling assumes a meaning which is much more practical with reference to other scientific disciplines as natural sciences, formal logic and mathematics. Modeling is a set of activities, tools, heuristics (in the broad sense of the term), capabilities which lead a designer to build system-answer to a problem-question which he/she is confronted to. 3.1
“The mathematical models that are used in OR are representations of the
system under study. These models may be imperfect and idealized, but still the
quality of the solutions that they yield crucially depends upon their closeness
to reality in the relevant respectes .” [
“ Simply put, the Product Verification Process answers the critical question
Was the end product realized right? The Product Validation Process addresses
the equally critical question - Was the right end product realized?” [
“ - What is a valid model? - has been one of the least discussed topics in the
OR literature. [. . . ] Thinking about model construction and model validation
is basically to raise the issue of different ways of producing knowledge and
deciding about the acceptability of the knowledge thus produced” .[
“Whether Operational Researchers are aware of it or not does not make any
difference: to take an option in the debate on model validation in OR is,
explicitly or not, to actualize epistemological choices”. [
Experimental approach to model validation
“The principle of science, the definition, almost, is the following: The test of
all knowledge is experiment. Experiment is the sole judge of scientific truth”
[
that emerged in contemporary epistemology show that the role of experimentation is (sometimes) considered as troublesome. There are a bright and a dark side of the coin.
strong arguments to trust a hypothesis. Secondly they can favor the discovery of new theories showing new unknown phenomena which call for an explication. As representatives of these two uses of experiment, we can cite, among others, G. Galileo and F.
both of them supported a new vision of knowledge based on observations that had to be performed without prejudice or preconception. However, we consider Galileo to exhibit an example of the use of experimentation to confirm a theory and Bacon as an example of use of experimentation to favour the discovery of new theories.
the Ptolemaic theory that fixed it at the center of the universe. In this case, facts obtained trough experimental work (repeated observation) confirm a theory (Copernican system).
Observations can foster new, general ideas, as explained by Bacon. In fact, Bacon was a convinced inductivist. His Novum Organum (1620) can be considered as the first modern work on inductive logic. In particular, it analyses the methods that can be used to produce theoretical inductive inferences, namely from particular to general, which had been relegated to a minor role during the previous centuries.
“The syllogism consists of propositions, propositions consist of words, and words are tokens for notions. Hence if the notions themselves (this is the basis of the matter) are confused and abstracted from things without care, there is nothing sound in what is built on them. The only hope is true induction .”
piricists of Vienna Circle2: who stated, in their Manifesto, that true knowledge is totally empirical because the scientific enterprise is characterized “essentially by two features. First it is empiricist and positivist: there is knowledge only from experience [...] Second, the scientific world-conception is marked by the application of a certain method, namely logical analysis .”
a proposition consists in its method of verification, and a proposition which cannot be verified is meaningless. Thus, the role of experimental verification is even stronger than in the vision of Galileo and Bacon, since it is at the basis of meaning.
poses that it is not possible to test experimentally a single hypothesis because complex theories includes many hypotheses and it is really hard to establish which statements are contradicted by a test (systems engineers would call this a traceability problem).
by Copernicus and the one proposed by Tycho Brahe. This position is known, nowadays, as Duhem-Thesis3.
A second difficulty concerns the trustworthiness of what we are used to considerobjective facts. Starting from the platonic allegory of the cave up to now, several philosophers have warned about the possibility that facts could be illusory. Many times in the history of philosophy evidence has been called into question. However, in this case, the target is not knowledge in general, it is the exactly the scientific method which is questioned. In the context of modern science a common reference, from this point of view, is the work of Hanson, as mentioned above. Hanson believes that there is not unconditioned observation of facts and, moreover, there is not a neutral language to
in 1922. Among its members there were Moritz Schlick, Rudolf Carnap, Richard von Mises, Otto Neurath, Herbert Feigl.
but, in reality they refer to quite different thesis. express them. Observational terms are “full of theory”. Thus the idea that theories are confronted to pure facts is wrong, in his opinion.
“There is a sense, then, in which seeing is a ‘theory-laden’ undertaking.
Observation of x is shaped by prior knowledge of x. Another influence on
observations rests in the language or notation used to express what we know, and
without which there would be little we could recognize as knowledge .” [
opinions, our background knowledge and, in general, our theory.
confirmation, which he explains through the example of the ravens. We normally admit that the observation of a black raven confirms the hypothesis that “all ravens are black”. On the other hand, a white raven is a clear counterexample. However if we also admit (and in general we do) the equivalence condition, then we get strange results. The equivalence condition states that if two hypothesis are logically equivalent, then certain evidence that confirms the first one confirms also the second (equivalent) one. A logical equivalent of “all ravens are black” is “all non-black objects are non-ravens”. This last is confirmed by a non-black non-raven, e.g. a white tie. It follows that a white tie also confirms “all ravens are black”. This is logically correct, but it sounds strange.
We know that Popper proposes a fundamental improvement to the verification principle of Vienna Circle. He believes that inductive inferences have no justification, since no matter how many singular facts you have observed, you are never sure that a different singular phenomenon could occur, making your general conclusion wrong. Thus verification is, in practice, not feasible. He introduces a different criterion to defend the possibility of empirical justification of a theory. A theory has to divide the world into two distinct classes of phenomena: the ones that are compatible with it and the ones that contradict it. Thus, we should not look for facts that confirm a theory, but for the ones that could make it false. The longest a theory resists to these assaults, the better. It is trusted, or, using his terminology, corroborated. This is a considerable progress with reference to the positions of Vienna Circle. Problems caused by induction are reduced.
Nevertheless, according to his opponents, the falsification method proposed by
Popper does not escape to the issues of theories underdetermination. During the sixties,
authors like Kuhn and Lakatos promoted the idea that science progresses through many
different ways, making our comprehension of its method more encompassing. Their
focus was no more on one single theory against facts. Scientific research started to be
considered as a complex system that comprehends many heterogeneous elements. The
terms paradigm proposed by Kuhn and research program proposed by Lakatos gained
a remarkable success and entered the terminology of philosophy of science,
becoming quite common. In particular (following [
ers (sometime) have to make choices that can not be based on experimental evidence.
5
Collaborative approach to model validation
the capacity of science of catching the ultimate truth about reality. For example, instrumentalism.
“Instrumentalism can be formulated as the thesis that scientific theories, the
theories of the so-called “pure” sciences, are nothing but computational rules
(or inference rules)”. [
to be described by a conclusive explanation, are totally left out. Instrumentalism does not focus on the distinction between truthfulness and falseness of scientific theories.
sentatives of this approach are, among others, E. Mach, H. Poincare`, P. Duhem, E. Le Roy. For example, Poincare`proposes that we can consider the axioms of the geometry as simple conventions. Similarily, Le Roy thinks that science has a pure instrumental value and that scientific laws are only convenient synthesis of sets of facts. The position of Duhem is more variegated, but not very different.
“A physical theory is not an explanation. It is a system of mathematical
propositions which can be derived from a small number of principles that serve to
precisely depict a coherent group of experimental laws in a both simple and
complete way”. [
language is not viable. No theory can provide general rules that are valid in all cases. On the contrary, we can establish only local norms since human language is elaborated in local contexts. He thinks that these norms emerge from behaviors and cultures based on what he calls language games, i.e. specific sets of linguistic rules. A perfect language does not exist and in particular there is not a perfect scientific language. Moreover, in his opinion, this reflects the absence of a common underlying structure, namely the
sitions are represented most of all by the Tractatus logico-philosophicus. absence of a common logic. We should drop the idea that there is one single “Logic” at the basis of human rationality and accept the fact that we act and think according to particular practices which are functional to particular aims and can not be generalized.
Instrumentalism, conventionalism and the “second” Wittgenstein open the door to the entrance in the field of philosophy of science of elements that, in the first decades of the 20th century, had been kept out. Social components are introduced as a fundamental part of scientific knowledge. The separation between external and internal components of scientific enterprise starts weakening, so that context and content begin running into one and knowledge is no more justified true belief , but, more weakly, locally accepted belief. Physics loses its supremacy as model of all scientific disciplines, and the nineteenth-century idea, renewed by the project of unity of science of Vienna Circle, that all branches of science could be reduced to mathematical explanation, is replaced by a more encompassing approach that admits final causes, interpretations, narrative explications. From the point of view of these authors, the study of nature is similar to the study of social institutions, myths, political groups. In other words, these epistemologists think that knowledge is only a social construction, namely that truth does not exist in itself and it is only agreed consensus (often, of experts). This current of thought suggests that what we consider true is composed by simple beliefs that someone, who has the power, prestige or status to do it, has legitimated.
Bloor and Barnes and other researchers of the University Edinburgh funded in the
’60 the Strong program of sociology of knowledge (Strong Program, for short)
endorsing these ideas. This stream of research fits in with the tradition of sociology of science
of Merton (cf. [
We remark that, among others, Popper was absolutely opposed to this approach and
he believed that sociology and psychology cannot be used to ground science.
“. . . to me the idea of turning for enlightenment concerning the aims of science,
and its possible progress, to sociology or to psychology . . . is surprising and
disappointing. In fact, compared with physics, sociology and psychology are
riddled with fashions and uncontrolled dogmas . . . This is why I regard the
idea of turning to sociology or psychology as surprising .” [
is an approach to scientific knowledge that tells us that a decision can be legitimately supported by a deal stipulated by all the people in charge of the choice.
decision making has its own tradition and, thus, indirectly, a kind of legitimation. We do not believe that this is the best method, neither that this is the only method, as strong program sociologists tell us. Nevertheless, in practice, when no other options are available, or empirical evidence is missing, decisions are taken by means of stakeholders’ agreement. We concur that this is not inadmissible. In practice, it happens, quite often.
6
Ethical approach to model validation
and management sciences in particular).
Churchman [
“Decisions should be made by consensus of all who are directly affected by the
decisions, the stakeholders .” [
Wallace’s edited book, Ethics in Modeling [
Mingers [
“ scientific legitimacy of OR models (ethics outside OR models) and the
integration of ethics within models (ethics within OR models)” [
previously. They identify three possible attitudes towards the relationships between OR and ethics. The first one corresponds to a sharp separation between them. It ensures objectivity of OR, but, in their opinion, is incomplete. The second one integrates ethics in OR. This approach is more complete, but has the flaw of accepting a certain amount of subjectivity. The third approach is based on a distinction between OR model and
“We present three methodological approaches to combine ethics with Operational Research. The first one is ethics outside OR models [. . . ] The second approach is ethics within OR models [. . . ] The third approach is ethics beyond OR models” 7
Teleological approach to model validation
tems engineering. In particular, we dare a possible (audacious) link. The concepts of goal and requirement, used in systems design, have their conceptual “ancestors” in the
For empiricists, the concept itself of teleological explanation of phenomena, namely the existence of purposes and objectives in nature for the sake of which things are done, is unadmissible. This would confer to nature something like a “free will”, which is incompatible with the idea of nature as mechanism. However, Aristotle advanced aims as one of his famous four causes: material, formal, efficient and final .
“Aristotle was deeply committed to investigating and explaining natural
phenomena, which is reflected all through the surviving treatises on natural
philosophy [. . . ] What unites the questions explored in these natural treatises,[. . . ]
is that they are predominantly questions asking for the purpose of things, or, as
Aristotle puts it, questions asking for - that for the sake of which -. According to
Aristotle’s understanding of scientific knowledge, the answers to these specific
why questions constitute teleological explanations [. . . ]” [
“Whereas in a typical causal explanation the earlier-in-time cause explains the
later-in-time effect, in teleological explanations, as traditionally understood,
the later-in-time effect (that is, the aim or purpose for which something
happened) explains the earlier-in-time cause (that is, why something happened).
The typical locution of a teleological explanation is: this happened in order
that that should occur.” [
“Bacon. . . quotes with approval the Aristotelien maxim - Vere scire est per
causes scire - and the Aristotelien distinction of four causes, Materia, Forma,
Efficiens, et Finis [but proposes . . . ] his famous condemnation of final causes
[. . . ] He blames their use in Physics; he approves their use in Metaphysics.”
[
“ Leibniz did admit teleological explanations alongside mechanical ones. Apart
from the need of teleological explanations (in terms of God’s purposes) in
metaphysics, he argued that physical phenomena can be explained by
mechanical as well as teleological principles. . . . Indeed, Leibniz wholeheartedly
accepted the Aristotelian final causes alongside efficient causes”. [