6. Weapons Research and Weapons Design (8/7/14)
This is a (long) excerpt from Chapter 2 of my book Designed to Kill: The Case Against Weapons Research. I’ve left in the footnotes and made no attempt to edit it, which I should probably do. It is the most comprehensive characterisation of weapons research that I know about, just as well since I wrote it!
To many people, the image of a weapons researcher is a scientist, a man in a white coat, possibly mad, working in some secret laboratory (my own image is that of a physicist in a tweed jacket working on the Manhattan Project). These days the weapons researcher is an applied scientist, or, more exactly, those that do weapon research mostly work in research and development (R&D) establishments, either government or private, though some work elsewhere, in universities for instance. There will also be people trained as engineers and other specialists, as well as those who are properly speaking scientists, working in such places. But what is distinctive about weapons research, and what distinguishes it for instance from weapons manufacture, is that it aims at design: weapons researchers aim to create new and improved kinds or types of weapons - I include here everything that is involved in making weapons work and in using them, such as bullets and bombs, command and control systems, delivery systems, platforms, soldier’s body armour, and so on and so forth.  Weapons manufacture puts the finished design into production and realises the design in the form of hardware and software. The output of the weapons researcher is, on the other hand, not hardware but knowledge, the knowledge of how to make a weapon. The object of all forms of scientific research, pure and applied, is the generation of knowledge, and that is why it is appropriate today to think of the starting point of weapons research as the application of science. This characteristic of weapons research, that it produces knowledge in the form of designs for new and improved weapons, is the point of departure of my discussion. Having stressed that point, in what follows I will take a broad view as to what form the knowledge of how to make a weapon should take and, consequently, as to what counts as a design.
Applied science, and R&D generally, is distinguished from pure science in that the intention is to produce more than just knowledge about the way the world is. Put another way, applied science is directed to producing knowledge that has some practical import. This distinction is a familiar one, and one that I have discussed at length elsewhere. It is consistent with the idea that the things designed and produced, artefacts, are natural objects, albeit ones that do not occur naturally, because pure and applied research employ the same basic concepts and theories, though they employ them in different ways. It is worth illustrating this point with an example. In 1939 Frédéric Joliot-Curie was investigating nuclear fission, which had been discovered the previous year by Otto Hahn and Fritz Strassman. He wanted to know how many neutrons were liberated, on average, in an assembly of uranium; he simply wanted to know this, to find out something about the way the world is. Enrico Fermi did similar things three years later, but with a very different agenda. He wanted to build a nuclear reactor as a prelude to making atomic weapons, but he used the same basic ideas as Joliot-Curie. Naturally-occurring uranium and highly enriched uranium used in the atomic bombs made by the Manhattan Project obey the same laws of physics.
With this is mind, I proposed the following definition of weapons research:
Weapons research is research carried out with the intention of designing new weapons or improving the design of existing weapons or designing or improving the means for carrying out activities associated with the use of weapons.
There are a number of comments that can be made about this proposal. One concerns what might seem to be the inadvertent or unintentional provision of weapons – a point related to the second of the two caveats made in connection with the definition of a weapon given in the previous section. So-called dual-use technologies are of current interest in the context of the biological sciences, in which a technology not intended for military purposes can be turned to such a use with little or no extra work. I will have nothing further to say about dual-use (but see Forge 2010). Other examples can be given with reference to what I have called generic technologies (Forge 2008: 49), technologies that have many different applications. Computing is an obvious example. I should note here that the proposed definition is not supposed to capture every single possible instance of WR and unambiguously and rule out all other kinds of research. It is probably best to think of it as a sufficient condition for WR. I want to engage with those who think of themselves as doing WR, as unequivocally intending to so. I am not concerned with identifying all and only those who have some sort of responsibility for the production of weapons.
A second comment is that this definition, and the preceding remarks about applied science and images of the weapons researcher, seems to have a modern orientation, at odds with the claims of the previous section that the history of WR is comparable in scope to the history of technology. I should make it completely clear that I do not construe WR as only applied science or ‘R&D’. Science in the guise of scientific theory has only informed WR fairly recently, so with reference to this narrow sense of applied science, much that has passed as WR is not applied science. However, the engineers, craftsmen, smiths and inventors who produced new weapons down the centuries can be seen as using some sort of method, insofar as what they did followed some kind of rational process, and, depending on one’s point of view, this could be construed as a scientific method. At the very least, to make some new kind of thing one must be systematic in some sense at some stage, and not just work at random. But nothing whatsoever here depends on whether or not it is correct to refer to these various ways of working as ‘scientific’. It is (of course) necessary to have some definite idea what WR is, and I claim to have given this, in terms of what the people in question intended to do. With a complex and demanding task such as the design of a new or improved weapon, it is obviously necessary to have some disciplined, systematic way of working, or the task could not be accomplished. For want of a better label, I call all such activity “research”, whether or not it is informed by scientific theory or conforms to some acceptable view of scientific method. It is therefore possible, under my definition, for someone to claim that he is doing WR in a hopelessly unsystematic way, with no chance of success. I have no problem with calling this WR.
A third and last comment concerns the fact that class of those who make possible weapons designs, whose contribution is necessary for the design to come into existence, may well include people with no technical expertise, who do not, in any sense, actually work on the design. Typically the latter will be those who finance, commission or otherwise bring into being the knowledge in question by providing various types of support. In the past these have included kings, princes and other rulers of states, such as Dionysios I of Syracuse, Philip of Macedon, The Holy Roman Emperor Charles V, Kaiser Wilhelm I of Germany and President Roosevelt of the United States. These days, many armaments firms are private and sometimes it is private capital that is used. Moreover, the classes of persons comprising the technical and the financial are not disjoint: Hiram Maxim, George Armstrong and Alfred Krupp, to name three, were, at one time, both weapons designers and financiers of their own work. If my overall aims are to show that the discovery of the knowledge of how to make new and improved weapons is wrong and hence requires justification, that such justification is hard to come by and hence that much (if not all) WR should cease, then am I not trying to influence the financiers and commissioners of WR – we might call them ‘weapons research facilitators’ - as well as the designers? A short answer is yes. A longer answer is that my focus of attention in terms of the examples to be given will almost all be on the designers, because even though on many occasions others have been needed to facilitate WR, it is surely the weapons researchers, those who actually invent the designs, that play the central role. Also, it is possible to undertake WR without an external source of finance, although admittedly less likely now than in the past, but it is not possible without someone taking on the role of weapons researcher. Finally, I have argued elsewhere that scientists of all kinds have an obligation not to harm others, and not to provide the means for harming, that stems from their special expertise as scientists (and I will broaden the class of those who have this obligation to include those who might not strictly be called scientists but have the expertise to do WR). It may be that those who finance and commission work have a similar kind of obligation, though I am not aware of the basis of it. Thus my main focus of attention is that who undertake WR in this ‘narrower sense’, and of course if my argument were to prove convincing to that group, then it would be enough to prevent WR.
The proposed definition of WR asserts that it is intention to design weapons that is the characteristic feature of this activity. So what is design, exactly? Designs are said to be blueprints, inventions and plans, all of these terms are often (somewhat misleadingly) taken to be synonyms of “design”. One could then simply say this: a design is a blueprint or plan for an artefact, where the artefact is either a whole, or a part or component of a larger whole. This suggests a criterion for “design”: if one is in possession of the plan, and has the right skills and resources, then one can make the artefact in question. The whole aim or point of design, as it is understood here, is to provide a manufacturing or production unit with precise details for making the artefact in question. Thus the term “blueprint”, now an anachronism because blue drawing paper is no longer used, usually refers to the technical or engineering drawings that provide all the information necessary for the manufacturer. This sense of design, used in the context of technical drawing where the exact technical and quantitative details are recorded, will be different from that used in earlier stages of the process of discovery and invention. In fact for a highly innovative complex discovery, it is to be expected that the design itself will evolve and develop, from an idea through to a set of technical specifications. Such specifications are in essence a set of instructions for making an artefact. This understanding of weapons research, as the quest for design, is deliberately wide, so as to include all the elements and processes that contribute to weapons development and all the personnel involved. And this straightforward view of design and weapons design is evidently along the right lines, but there are some further distinctions that we can usefully make in order to get a clearer view of what design is and how designs are invented or discovered.
In the first place, we may note that in regard to the (seemingly obvious) relationship between design, what it is that is created by human designers, and the thing designed, the artefact, is such that the former comes first. Thus if we say, as it seems we should, that the thing, the artefact, represents the design, then this is a special type of representation. The usual type of representation is one in which human agency takes some extant object as its starting point, but it is the other way around with designs and what they represent. Garbacz calls this ‘poietic representation’ (Garbacz 2009:3)  In the second place, if a design is indeed the ‘knowledge of how to make something’, then that knowledge can find expression on more than one occasion and in more than one way. That is to say, there can be many different realisations of the design, in written, electronic, diagrammatic, etc., forms. Garbacz, who I follow again here, refers to such a concrete instance of a design as an ‘engineering specification’ (Garbacz 2009: 4). This term is perhaps a little grand for some of the older and more traditional instances of designs, but I will make use of it here. I would also note that the artefact itself counts as an instance of its own design, and it is therefore possible that a tractable set of instructions for reproducing it can be obtained just from studying the artefact, provided that what is ‘embodied’ in the artefact can be read off by some process of reverse engineering. These distinctions, between design and artefact and between design and different engineering specifications help, us answer some philosophical questions about design, but also raise further questions. For instance, if we ask what a design really is, the answer is not any engineering specification or other ‘instance’, but what all such things have in common, or express.
These observations raise the issue of just how similar or faithful engineering specifications have to be to the design in order to be said to be an instance of it - and how do we tell? Are there canonical engineering designs, like the old standard platinum metre rod, against which all others are measured? There is a second issue that follows from this which introduces the possibility of a kind of second-order uncertainty: how faithful do artefacts have to be to an engineering specification – how ‘well’ do they have to be made – for them to qualify as a representation of the design? I think it is easy to see how both kinds of uncertainly are possible, namely by human and physical factors, such as human errors and physical (thermodynamic) limits on the precision of translations and production. Thus if the engineering specifications are imperfect and the production process of the artefact lax, then the representation of the design by the artefact will not be optimal. Such lack of precise fit between design and specification and between specification and artefact may be thought of as a kind of mutation that contributes to an evolutionary account of design and artefact development. I will say more about this second matter in a later chapter. Notice also that once a design has been invented and articulated to the extent that there are engineering specifications, and especially when ‘poietic representation’ has been realised, it will be almost impossible to recall or destroy it. Destroying a design would entail destroying all engineering specifications, and all other descriptions of it, all artefacts made on its basis, and all other knowledge thereof. In practice, then, once a design has been invented and realised, it cannot be recalled or ‘uninvented’ – an important observation for the case against WR. Now we need a concrete example to illustrates these ideas about design and I will refer here briefly to the history of the atomic bomb (more on this in Chapter 4).
The idea of using a chain reaction in a fissile substance like uranium (or plutonium) to liberate lots of energy has been attributed to Leo Szilard in 1932.  No one knew at that time whether a device that could do this was possible either in theory or in practice. Physical theory might have shown that certain parameters have values that made the device impossible: for instance if the critical mass of fissile substances, the mass needed to sustain the chain reaction, were too small, there would be no big bang. Thus, once Szilard and others had started to investigate the possibility of the bomb, much basic scientific research needed to be done to determine the values of these parameters. Even if the values were right, there could still have been physical reasons why a bomb was not possible: for instance, it might have been impossible to obtain enough fissile material – the material might simply have not been available, or not available in a usable form. Finally, a bomb might have been practically impossible: that is to say, no one was able to come up with a design. The actual design process that took place at Los Alamos was hard work, with significant changes taking place in connection with the design of the two plutonium bombs. Notice that the kinds of possibility just mentioned are different. If physical theory ruled out the possibility of a bomb, for instance if only one neutron at most is liberated in any fission event, then the chain reaction could not diverge, and that is the end of the story. If the weapons designers at Los Alamos in 1942-1945 were not ingenious enough to come up with a design, however, it does not follow that no one else could. What this brief account shows is that a design is really only a design if it works, or if it works ‘well enough’. Szilard’s original idea was not a design - far from it, as we shall see in more detail later – and much work had to be done for it to be workable. Much of this work involved applying scientific theory to see, in effect, whether a certain kind of artefact could exist. As a final observation here about this design, the Soviets and others invented essentially the same design - one that entailed imploding a subcritical hollow sphere of plutonium to make a supercritical assembly - though with rather different engineering specifications.
Turning now to the design or R&D process, there appear to be two end points: at one end, completion, there is a complete set of technical engineering drawings that give all the information needed to produce the artefact. We might even imagine here that the artefact could be produced in its entirety by programmable machines, once this sort of design is available. At the other extreme, initiation, there is the bare idea with little or no detail: of using a chain reaction in uranium to make a bomb, of having a breech loading rifle, of a laser- based stars wars missile defence that could shoot down nuclear warheads, of a machine gun, and so on. Some such design ideas are fantasy and others are realisable. Some, or indeed as in the case of the atomic bomb a lot, of applied scientific research may be needed to ensure that the idea is not fantasy.  There are other forms that the design process can take. I have said that there is an obvious sense in which an artefact itself is a design. Thus another kind of design process will begin, not with an idea, but with an artefact with the aim of reverse engineering. This suggests that the once designs are invented, then they count as items in the world that are open to (re)discovery. There are other forms still that the design process can take. Firearms have been copied and improved on over the years, with one gunsmith using another’s work as the basis of his own innovation. Here the aim is not to take the artefact and (re)discover the design, but to use the artefact as a basis for coming up with a new design that is an improvement on its predecessor. Moreover, in the age before mass production, where artefacts were produced by craftsmen, the ‘set of instructions’ that constituted the design may simply have been a model artefact or template. Or these instructions may have been the practical or tacit knowledge held by the master craftsman and handed on to his apprentices. Here we can see again how the artefact itself could have evolved, under repeated reproductions by generations of craftsmen.
These concepts of design and design process all fit with the proposed definition of weapons research. If the intention of the scientist, engineer, gunsmith, inventor, or whatever title is appropriate, is to come up with a design for a weapon, whether it is the barest outline of an idea, a project to make an existing weapon more serviceable, a precise technical drawing or a reverse engineered missile, then the agent is engaged in WR. So, for example, a person who dismantles an assault rifle just out of curiosity, to see how it works, is not doing WR, while another person, performing the exact same actions but as a preliminary to using the knowledge to recreate the design, is so engaged.  However, what do we say of a technician, perhaps an expert in technical drawing, whose job it is to produce a set of clear engineering specifications from a less formally exact design specification? Strictly speaking she is not doing research, but yet her work seems necessary to realisation of the artefact. My response here is the same as it was above in regard the weapons research facilitators such as financiers, into whose group we could include expert drafts people and others who work on producing engineering specifications: those who are not in the business of creating or discovering weapons designs are not the main target of the arguments that I want to mount, but if they also apply to, or influence, weapons research facilitators, all well and good. And again, it is possible to state a criterion for membership of the class of weapons researchers, as we have been doing, but the arguments against WR do not depend on how good this criterion is at capturing all and only those who are properly speaking weapons researchers, and it does not count against the case to be made here that is to be developed on the basis of these arguments that they might have purchase on some that are not, strictly speaking, weapons researchers.
Continuing with our discussion of what counts as taking part in the design process, it does not matter if the research activity is successful – it makes perfect sense to say that someone was engaged in WR but failed in her objective. This raises the question as to how we can tell if someone is doing WR, if this cannot be ‘read’ directly off her actions. The first thing to say here is that this may well not matter: unless we specifically want to determine in a particular case whether a given agent is or is not engaged in WR, it will not matter if what she is doing is open to more than one interpretation. The present work is not a history of WR, though it will appeal to that narrative in several places and presupposes that it does exist. Secondly, in many cases it will be obvious that the agent is indeed doing WR. This will certainly be case if she works in a modern day weapons R&D establishment, and it will be the case if she seeks a patent for her work. Patents, in one form or another, have been in existence since the dawn of the modern era and many of the American and European gunsmiths patented their work in the nineteenth century: and John Hall, Johann Dreyse and Mauser, to name three, had whole series of patents for their rifles, as did Maxim for his machine gun. And of course a patent is a design specification.
Some designs are not taken up, not because the artefact cannot be made, but simply because no one seems to want it, or because it was not marketed properly, or for some other reason. What is true of artefacts in general is also true of weapons in particular. There is, then, the general question as to what makes an artefact, and hence a design, successful. And there is a corresponding question: how is it that inventors or designers come up with ideas for new inventions, for new artefacts? These questions are the province of the historian of technology, and there there are various competing viewpoints. Along one axis of competition is the opposition between the view that invention is a discontinuous process, with brilliant inventors putting out radically new designs, for water wheels, spinning machines, steam engines, and atomic bombs, as opposed to the idea that change is the result of continuous small incremental improvements. Something of a compromise between these extremes is the theory, already mooted, which has it that change is evolutionary, that new designs are the product of ‘mutations’ of existing ones - brought about perhaps by the kind of imperfect translation from design to engineering specification and by lax production - which are then ‘selected’ by a variety of ‘environmental’ factors. Clearly, there is an analogy at work here. The evolution view of technology is the one I will adopt, and I will say more about it in later. Its importance for my account is that it sees designs as forming a kind of lineage, in that future artefacts have their origin in those of the past, although not simply as the result of an incremental process, as there is room for fairly rapid change. Thus someone who designs a new and improved projectile weapon therefore also provides the possibility for further improvements. Thus not only do design themselves not die out when they are invented, they have immortal offspring as well.
Some of the recent work on artefacts and design in the philosophy of technology has addressed ontological questions, such as “what is an artefact?” and “what is a design?” I will close this section by stating an answer to the second question, which I find to be a reasonable one. A general account of properties and kinds sees them as universals, which are instantiated in particular things. Universals are therefore the basis of the resemblance between particulars. Philosophers in general, as opposed to philosophers of technology, have usually discussed these matters in relation to naturally occurring objects, but I see no reason for that restriction. So suppose we are confronted with two AK-47 assault rifles. These items will resemble one another in a numbers of ways, have a number of things in common, such as being made of steel, wood, having a certain weight and length. All of these resemblances can be explained by their having the corresponding universals in common. But can we not also posit the universal “being an AK-47” and refer to that as what explains the overall resemblance between the two items, how they resemble one another as instances of the same kind, what they have in common? I do not see why not. Indeed, I do not see why we should not entertain the view that the design of the AK-47 is this universal. Recall that designs are different from particular engineering specifications and they are different from the artefacts that embody them. Would not this view tie together design, engineering specification and artefact, in such a way that we could understand the relations between them? Garbacz' somewhat mysterious ‘poietic representation’ could then be understood in terms of the instantiation relation. Moreover, this view would have the consequence that designs are discovered, not invented, and it would help explain how it is that they cannot be destroyed. I should stress that the case to be mounted against WR does not depend on the correctness of this view of the ontology of design and artefacts, though I believe it is helpful to have some such account in the background.
 So I am using “weapons research” in a wide sense. In what follows most of my examples will be of bona fide weapons – guns, bombs, missiles and the like – and not ancillary elements.
 This is in line with most modern views about the nature of technology: technology is knowledge of technique, of how to do something, in this case, make a weapon.
 See my The Responsible Scientist, Forge 2008. The account given there of pure and applied science, of the realist interpretation of technology and much else about science, technology and morality is the point of departure of the present book. I will give brief summaries of the main positions, where appropriate, in what follows.
Put another way, although artefacts are natural objects, their origins are artificial, and this distinguishes then from ‘natural’ natural objects.
 I first used this definition in 2004, see Forge 2004: 534, and have employed it on a number of occasions since.
Dual-use presents some interesting and difficult issues, some of which are relevant to weapons research, but they do not fall within the scope of the present inquiry because, I assume, that dual-use research is not itself weapons research, and hence those who undertake research that can be classified as dual-use are not bona fide weapons researchers. What I have to say about weapons research may be relevant to dual-use, but the present project is sufficiently large without exploring all of its further implications.
 The main aim of the first part of Forge 2008 was to develop a wide view of backward-looking responsibility, that is to say, giving an account of what is for an agent to be responsible for an outcome. The focus was on the scientist as an agent and the idea of the wide view was to maximise the attribution of responsibility. To that end I maintained that an agent can be responsible for more than what she intends. The same account can be applied here, and so it is possible for an agent to be responsible for doing WR even if this is not what she intends. Again, my purpose here is not to engage is some deep analytic exercise that will give a precise and highly nuanced definition of weapons research; it is with its overall moral implications.
 This is certainly to take a wide, perhaps unreasonably wide, view of the matter, as it would gather all the craft traditions that have been employed in weapons research under the heading of scientific method. But again, there is no special thesis about science and the scientific method and its role in weapons research that I want to uphold here.
 A way of determining who can be responsible for an outcome of some kind which I have developed elsewhere is to determine the full set of what I call causal roles, which are, essentially, the full set of tasks that need to be undertaken to produce the outcome, see Forge 2008: 81.
 This is one of the two main claims of The Responsible Scientist.
Unless we think of those who have, or control, large amounts of finance having some kind of noblesse oblige. To make sure the point here is clear: no one should harm or provide the means to harm, financiers included. However, I’m not aware that this demand weighs more heavily on those with money than on everyone else.
 Artefacts are usually understood as things. There can be designs for processes as well, and indeed WR can certainly aim to design processes. In what follows, I will use “artefact” in a wide (and unconventional) sense to cover both objects of design.
 In this sense, designs are contextual: unless the right skills and resources are available, the design cannot be realised. One might then speak of the specifications counterfactually: were there the available resources, this would have been a set of instructions for this artefact.
 Techné, the journal in which this reference appears is an electronic journal which can be found here http://scholar.lib.vt.edu/ejournals/SPT. I can find no pagination for the issue of the journal as a whole, so I have just listed the page numbers of the article as I have accessed it. I note also that the distinction between design and the thing designed can be thought of as an instance of the type-token distinction.
 This kind of distinction is familiar to philosophers who distinguish statements and propositions, items of knowledge, from the sentences and other forms of linguistic representation that express them. I note that Grabacz is right when he says that the design is not the sum of all the engineering specifications.
The first question is about ‘exemplification’ and the second about ‘optimality’.
 There is a detailed account of the design of the plutonium bomb in Forge 2008 Chapter 2.
 Other relevant parameters concern the cross sections, essentially probabilities, for fission and capture by the nuclei of the fissile material, and the number of neutrons liberated per fission event.
 There is more that could be said about how this works in practice. Thus, prototypes are usually built and tested and their performance might lead to revisions of the design. Moreover, problems or suggestions may arise at the manufacturing stage and feed back into the design process. Thus there is not, in practice, a sharp boundary between R&D and production, with each having their exclusive tasks. Insofar as an activity that take place in a weapons production unit leads to changes in design, I classify that as WR.
 As I have argued elsewhere, what defines this research as applied is not its ‘content’ but the context in which it is carried on. Much of the preliminary research done on the bomb project would have counted as pure or basic research had it been done out of curiosity and not in the context of weapons research.
In a discussion of the idea of design space, Stankiewicz distinguished four ‘design regimes’, which are the craft, engineering, architectural and research regime, respectively. The first is characterised by poor standardisation and high unit cost, by gradual development and little differentiation between design and production activities, and transmission occurring by apprenticeship. By contrast, engineering design is normally accompanied by industrial production techniques distinct from design activities, so there is much greater standardisation. Design is typically represented symbolically. The third category will not interest us here, but the fourth will. In the research regime, new design elements are discovered by scientific research, natural science as distinct from engineering science. For all this see Stankiewicz 2000, especially 237-240. The development of nuclear weapons took place under the research regime.
 According to the theory of responsibility in Forge 2008, if the curious disassembler either foresees that someone else will make use of his work to re-create the rifle’s design, or if he should have foreseen that this will be the case, then in fact he is doing WR. Again it is the context of the work and the intentions of the workers that, in the first place, determines the nature of the activity.
 Different versions of the evolution view have different implications for the rate and nature of change.
 Armstrong 1989 is a good introduction. Armstrong follows Aristotle and holds, against Plato, that there are no uninstantiated universals. A design that was never implemented would have this status. It would still count as an item of knowledge, and something that was the product of intelligent activity, but it would not qualify as a universal until it was instantiated.
 And this view cannot be the whole story. Artefacts are natural objects and hence have the same sorts of properties as naturally-occurring natural objects, but they are also designed by agents and would not exist but for the intervention of agency. This suggests that classifying designs as universals cannot be the whole story, and some reference to agency needs to be included in an ontology of design. And, more can be done by way of elaborating the ontology. For instance, there can be higher-order universals, universals whose instances are themselves universals. We could make use of these to group together designs of the same kind, such as “is a design of a rifle”, which is instantiated by “is a Mauser”, “is a Chassepot”, and so forth. Such higher-order universals could have a role to play in the elaboration of the evolution view of technology as it will be applied to weapons development.