IAC-10.D1.5.

HOW DO WE FIX SYSTEM ENGINEERING?

Michael D. Griffin
Eminent Scholar and Professor
Mechanical and Aerospace Engineering
University of Alabama in Huntsville
Huntsville, Alabama, USA
[email protected]

61st International Astronautical Congress

The now half-century-old multidisciplinary, interdisciplinary discipline that we call "system engineering" is the
classic half-empty, half-full glass; optimists and pessimists can look at the same thing and draw opposing
conclusions. Optimistically, the maturing of system engineering into a recognized discipline from its roots in large
aerospace and defense programs has been, and will remain, an enabling factor in the ability of societies to deal with
the macroscale problems facing us in energy, environment, and other key areas. Pessimistically, system engineers
have some explaining to do. How is it that we continue to encounter failure of important and complex systems
where everything thought to be necessary in the way of process control was done, and yet despite these efforts the
system failed? Each time this occurs, we as an engineering community vow to redouble our efforts to control the
engineering process, and yet such events continue to occur. The answer cannot lie in continuing to do more of the
same thing while expecting a different outcome. We need to rise above process, to examine the technical, cultural,
and political mix that is "system engineering", and to examine the education and training we are providing to those
who would practice this discipline. This paper will discuss that training from a new perspective, the perspective of
design elegance, how we identify it, and how we can design with elegance as a value.

I. INTRODUCTION
The period from the aftermath of World War II to the Most prominently visible in the aerospace, defense,
present day has been a period of technical innovation and energy sectors, such systems are characterized by
unmatched in human history. Engineers practice the common trait that they must work; failures
today in fields such as integrated circuits, computers, become societal events and national tragedies.
software engineering, information technology, Three-Mile Island, Challenger, Columbia, the power
materials science, nuclear engineering, and of course blackouts of 1965 and 2003 in the northeast United
astronautics, that did not exist in 1950, while States, and the recent Gulf of Mexico oil spill offer
traditional engineering specialties such as civil, just a few of many unfortunate examples. These
mechanical, electrical, chemical and optical sectors are likewise characterized by the necessity to
engineering have been revolutionized by these newer analyze, understand, and predict the in the large
disciplines. To say that, collectively, these things consequences of decisions, actions, and designs
have transformed the practice of engineering is to having ostensibly limited scope, but which actually
trivialize the issue; they have transformed the world. are of profound import, too often seen only in
retrospect.
However, our world today is characterized, even
dominated, not by the singular products which were Thus, success in designing, building, and operating
themselves transformative when they were efficient large scale systems to serve crucial purposes
introduced electrical power, telephony, in both public and private sectors is a defining
automobiles, radio communications, airplanes, requirement for societal advancement and economic
semiconductors, computers but by large scale prosperity in the world of today and in the future. In
systems characterized by complex interactions that world, the elegant integration of technologies is
between and among many separate products, more important, more consequential, than the
elements, and technical disciplines. Advanced development of any particular technology or
societies today are heavily dependent upon systems engineering discipline. Whether widely understood
which cut across the broadest possible spectrum of or not, these facts have given rise to the discipline of
disciplines. system engineering, another field that did not exist in

Page | 1
1950 and which, from a societal perspective, may engineering community to ask whether something is
ultimately be the most significant of those which missing, whether the discipline remains incomplete in
have arisen since that time. ways that are substantive and meaningful rather than
mere matters of detail.
It is beyond the scope of the present discussion to
recount the growth and formal development of the Petroski has offered numerous thoughtful essays2,3 on
field of system engineering and its allied discipline, the role of failure in engineering. While drawn
program management, from its origins in the largely from examples provided by the civil
development of complex aircraft, missile, and space engineering community, the broader principles are
systems in the years immediately following World equally applicable to the system engineering domain.
War II. This subject is well treated by Johnson1 in A key theme of Petroskis work is that a detailed
The Secret of Apollo, a must read by anyone understanding of the manner in which a given design
seeking to understand the history of modern fails in application allows iterative improvement of
aerospace system development methodology. It will successor designs. What, then, are the lessons to be
suffice to note here that after sixty years of learned from failures of complex systems?
development there exists today a substantial body of
methods, processes, and tools which comprise the One observation is that failures of system engineering
discipline of system engineering. Formal academic process have in the past typically resulted in the
degree programs exist to train students in this addition of more, and more detailed, process. In the
methodology, and it is practiced in both government world of 1950, when system engineering was largely
and industry, where ten percent or more of the cost of non-existent indeed, the term had not yet been
a large program may be devoted to the broad coined and the methods, processes, and tools which
category of system engineering. characterize the discipline today had yet to be
invented, this would likely have been the right
II. WHAT IS SYSTEM ENGINEERING? answer in any given case. But in the world of 2010,
System engineering as it is taught and practiced is it is this authors view that the addition of more or
fundamentally concerned with identifying the new system engineering processes is likely not the
separable elements or blocks of a proposed design, right answer in response to any particular failure. It
characterizing the intended relationships between and puts one in mind of the jocular definition of insanity:
among those elements, and verifying that the actual continuing to do the same things over and over, while
configuration is fabricated and operated as intended expecting a different outcome. If system engineers
in its environment. For large complex systems a are to achieve success through failure, the community
modern transport aircraft, a launch vehicle or must be prepared to do something different.
spacecraft, a power plant, a submarine this is no
small feat, and the methods, processes, and tools In the authors opinion, what is needed is a different
which have developed over the last half-century to view of the core system engineering function;
formalize and systematize it as an essential specifically, that this core function is not primarily
engineering discipline are not to be slighted. Yet concerned with characterizing the interactions
failures continue to occur, often of the most glaring between elements and verifying that they are as
and consequential nature, commonly at the intended. In a word, system engineering is not
boundaries or interfaces between elements, often due fundamentally about process, except in cases where
to uncontrolled, unanticipated and unwanted such process is clearly lacking, a state which
interactions between elements, in many cases characterizes very few large scale projects today.
between elements thought to be entirely separate. System engineering is about something more. Let us
consider what this might be.
What is of interest in many of the highly public
failures which have occurred in large scale systems While at its core system engineering is concerned
over the years are not those instances in which with the interfaces between and among separable
something known to have been needed was simply system elements, it should be realized that the more
omitted, or those in which a piece-part simply fails. important understanding concerns the dynamic
While significant, such cases are relatively easy to behavior of the interactions between these elements,
understand and correct. What is of interest are those not the numbers in the associated Interface Control
cases, all too many, in which everything thought to Document (ICD). As Gentry Lee has put it, its
be necessary to success was done and yet, in the end, about the partials, not the values4. Properly
the system did not perform as intended; in a word, it understood, system engineering is concerned with
failed. It is these cases that should cause the system

Page | 2
context over structure, with interactions over technique is a tool. From time to time I am
elements, with the whole over the sum of the parts. briefed on the results of a systems analysis
or systems engineering job in a way that
When this is understood, it becomes immediately prompts me to ask the questions: That's
apparent that the system engineering process bears fine, but is it a good system? Do you like it?
the same relationship to system engineering that Is it harmonious? Is it an elegant solution to
financial accounting does to financial management. a real problem? For an answer I usually get
Careful financial accounting is an essential element a blank stare and a facial expression that
of a good financial management plan; however, suggests I have just said something really
accurate accounting cannot distinguish between a obscene.
good plan and a poor one, and it cannot make a bad
plan better. Similarly, understanding and control of We must bring the sense of art and
system interfaces, development of comprehensive excitement back into engineering. Talent,
test and verification plans, and proper allocation of competence, and enthusiasm are qualities of
requirements are among the things which are crucial people who can use tools; the lack of these
to good system engineering; however, they do not characteristics usually results in people who
help to distinguish a good design from a poor one, cannot even be helped by techniques and
nor can they make a poor design better. tools.

The concerns expressed here are not new. In a With these words, Frosch is holding the system
landmark 1969 paper for the system engineering engineering community to a high standard; i.e., the
community, former NASA Administrator Robert end result of its work is not to be the satisfaction of
Frosch noted that5, possibly complex but ultimately well defined
requirements and processes. A larger, more holistic
I believe that the fundamental difficulty is goal an elegant design is to be sought. It is this
that we have all become so entranced with authors opinion that Frosch is entirely correct.
technique that we think entirely in terms of
procedures, systems, milestone charts, Elegance in engineering design is an ineluctable
PERT diagrams, reliability systems, concept; it is immediately apparent when it exists, yet
configuration management, maintainability it is difficult to define, impossible to quantify and, so
groups and the other minor paper tools of far, apparently incapable of being taught. Yet, no
the "systems engineer" and manager. We aeronautical engineer and no pilot need be taught that
have forgotten that someone must be in the DC-3 was an elegant design, while the Ford Tri-
control and must exercise personal Motor was not. It is offered here that, properly
management, knowledge and understanding understood, system engineering at its core is
to create a system. As a result, we have concerned with attaining elegant designs.
developments that follow all of the rules, but
fail. III. WHAT IS AN ELEGANT DESIGN?
If system engineering is to be fundamentally
I can best describe the spirit of what I have concerned with pursuing elegant designs and
in mind by thinking of a music student who rejecting those that are not, then it is necessary to
writes a concerto by consulting a checklist parse the term more carefully, to define the attributes
of the characteristics of the concerto form, of an elegant design, to learn how to quantify and
being careful to see that all of the canons of rank these attributes, to understand what actually has
the form are observed, but having no flair failed when a complex system fails, and how it
for the subject, as opposed to someone who might be improved in a later iteration. The present
just knows roughly what a concerto is like, work constitutes such an attempt.
but has a real feeling for music. The results
become obvious upon hearing them. The Among the possible attributes that might be ascribed
prescription of technique cannot be a to an elegant design, at least four are germane to the
substitute for talent and capability, but that practice of system engineering.
is precisely how we have tried to use
technique. The first of these is the most basic: does the design
actually work? That is, will the system produce the
We have lost sight of the fact that anticipated behavior, the expected output, over the
engineering is an art, not a technique; expected range of input conditions, control

Page | 3
variations, etc.? This question, relatively simple to
answer for a beam or an amplifier or an airplane But to advocate that engineered systems should not
wing, can be intractable for a complex system. Brute be tightly coupled is too simplistic. In designing
force testing exhaustive input-output measurements systems, it is a truism that we should strive for
across the entire operating range is not the answer. simple interfaces, to make the system easier to
Not only do modern systems with non-linear digital understand and to limit the possibilities for
elements quite likely have more possible state unintended interactions. But equally important is the
configurations than can be examined in any response of the system to unexpected variations in its
reasonable time, it is usually the case that not all input or environment; i.e., faults. Simple interfaces
possible configurations can be reached in the test by themselves do not address the propagation of
environment, or would be safe if they could be faults across and among those interfaces. If a system
reached. is to be robust, what is needed is the equivalent of a
ripstop fabric in the design.
Thus, it must be asked: upon what grounds is it
believed that a proffered system design will work as At present, however, the system engineering
intended? Who understands, in the large, how it community lacks a general theory to deal with these
works, and why it is believed that it will work that issues. It lacks quantitative measures for what
way? If not the chief system engineer, then whom constitutes tight coupling among system elements,
should it be? It would seem that in forty years we it lacks knowledge of what an appropriate threshold
have not gotten beyond Froshs observation that might be if quantitative measures were to become
someone must exercise personal management, available, and in any case it is frequently true that
knowledge, and understanding. It would seem that unintended, unanticipated, and possibly
academic research into the question of how it is that unanticipatible, interactions between system elements
we might know whether a complex system produces which were thought to be isolated are the root cause
the anticipated output for a given input remains of a problem. It is hard to know how a system can be
lacking. designed to be resilient in the face of unknown
unknowns. Finally, in systems where weight and
If a given design is believed to work, it must then volume are at a premium aircraft, spacecraft,
be asked whether it is robust, by which it is meant submarines it is difficult to imagine how tightly
that the system should not produce radical departures coupled system designs can be avoided. Yet, it
from its expected behavior in response to small remains true that some system designs are more
changes to its operating input, internal state, or robust than others, and some system designers know
external environment. It should degrade gradually how to enhance this property while others do not.
and gracefully in response to component failures, Work is needed to understand how this can be done
changes in its operating environment, or when design more purposefully.
loads are exceeded. Briefly, a robust system will not
surprise us. When a tree branch falling on a power Attempts have been made along these lines. While
line can cause a blackout of the power grid in the Perrow does not offer much in the way of praise for
northeastern United States, the system is not robust. safe modes or fail safe design approaches, such
When a piece of insulating foam can cause the loss of concepts along with circuit breakers, dampers, and
a Space Shuttle, the system is not robust. But again, limiters can and do work; they have saved many
while confidence in the robustness of certain systems complex spacecraft. As the theory of adaptive
can be gained through experience and intuition, and control systems continues to mature, it should enable
while fragile systems can be identified after the fact system stability across a wider range of operating
of failure, quantifiable measures of robustness do not conditions, thereby reducing accidents. With these
exist. and other techniques, it may be possible to attain
optimal, tightly-coupled designs that are also robust.
Sociologist Charles Perrow addresses the topic of
system robustness without specifically employing the A third property of an elegant design is that it is
term in his seminal work, Normal Accidents6. Perrow efficient; it produces the desired result for what is
argues that complex, tightly coupled system designs thought to be a lesser expenditure of resources than
he makes much use of the Three Mile Island and competing alternatives. But how is this to be
Space Shuttle Challenger disasters are pre-disposed evaluated? Design studies conducted for the purpose
to exhibit divergent behavior in response to relatively of selecting a preferred system concept from among a
minor problems. The resulting catastrophe is, to set of alternatives are notoriously poor in regard to
Perrow, a normal accident; it is to be expected. their ability to predict cost, schedule, and

Page | 4
performance for a finished system. The system the grass is always greener on the other side of the
engineering communitys ability to differentiate fence.
between and among competing designs, based upon
such cost, schedule, and performance predictions, is With that noted, it remains true that the system
even poorer. As-built prototype designs can of engineering community requires greatly improved
course be compared, but competing designs which means to compare alternative design concepts prior to
were not selected are generally not built. Fly offs selection for development. As Norm Augustine has
are rare in aeronautics and essentially non-existent in famously noted8, ninety percent of the time things
most other arenas where complex systems are will turn out worse than you expected, reflecting the
needed. Yet without direct comparison, the question fact that only ten percent of development programs
of relative efficiency is difficult to answer. An are completed for the advertised cost. Of this
alternative concept not being built will almost always performance, it can only be said that mediocrity
appear preferable to a system which is actually being would be an improvement.
built, and is in the throes of typical engineering
development problems. A fourth attribute of an elegant system design is that
it accomplishes its intended purposes while
Adm. Hyman Rickover noted this aspect of complex minimizing unintended actions, side effects, and
system development programs in his famous consequences. Among these might be wasted energy
discussion of the difference between paper reactors in the form of heat, noise, or vibration,
and real reactors7. When confronted with a situation electromagnetic interference or other undesired
in which a variety of alternative concepts were being interactions with other systems and subsystems, or
advocated in place of the pressurized-water reactor pollution. Control of these effects is a more subtle
design he favored for the nuclear navy, Rickover facet of system engineering, less noted but of
noted that there were two kinds of reactors, paper considerable significance. It will often be observed
reactors; i.e., new reactor concepts, and real that nearly all of a designers attention is focused
reactors. A paper reactor has the following upon producing a design which accomplishes its
characteristics: intended purpose. However, upon further thought it
may be recognized that, once a concept has been
It is simple. selected, most systems do in fact accomplish the
It is small. basic purposes for which they were designed. A key
It is cheap. differentiator between better designs and poorer
It is lightweight. designs is to be found in the unwanted features and
It can be built very quickly. effects produced by poorer designs, and the actions
Very little development is required; it can use which must be taken to compensate for them. Even
off-the-shelf components. designs which work, which are robust in the face of
It is in the study phase; it is not being built varying input and environmental conditions, and
now. which are reasonably efficient in comparison with
competing designs, may be found to produce a
In contrast, a real reactor has the following variety of unintended and unwanted side effects.
characteristics:
As noted above in connection with the system
engineering communitys inability to rank the
It is complicated.
efficiency of competing designs prior to making a
It is large.
commitment to a particular concept, so too the
It is heavy. community lacks the tools to evaluate the unintended
It is behind schedule. consequences produced by a given design. While the
It requires an immense amount of development properties of workability, robustness, and efficiency
on apparently trivial items. can be examined by means of simulation or the
It takes a long time to build because of its fabrication of prototype systems, this is not the case
engineering development problems. where the evaluation of unintended consequences is
It is being built now. concerned. Experimental or prototype systems are
generally not intended to be faithful to the final
All experienced practitioners of engineering system design in regard to precisely the ancillary
development are familiar with this phenomenon, characteristics that may be troublesome for a system
which is as much social and cultural as it is technical: in actual use. Often the troublesome effects are
noticed only in the long term, or are statistically rare

Page | 5
but are found, in retrospect, to be of considerable The first of these necessary advances must occur in
import when they occur. Further, it is a rare the academic community. Most of the research
modeling and simulation tool which provides any required to go beyond rule-of-thumb approaches to
analysis of potential side effects in addition to the the design of elegant complex systems will be
primary purposes for which it is developed, unless performed within the academic community. But at
that analysis is specifically requested by the system the same time, it will be essential to involve the
engineer. Often such questions are not even asked. academic community in a new and different way.

IV. HOW CAN WE ATTAIN ELEGANT The academic community is organized to, and
DESIGNS MORE SYSTEMATICALLY? provides strong incentives for, discipline-oriented
Additional attributes of design elegance may well research. Faculty are hired and rewarded for their
exist; however, the present author contends that any propensity to obtain and conduct sponsored research
design which is believed to possess that elusive in, generally, quite narrow areas of specialization,
property will be thought to possess at least these four which usually are narrowed further as the success and
attributes. A key theme of the present work is that prestige of a researchers career are advanced.
few in the system engineering community recognize Funding is allocated by most research sponsors along
that the responsibility for ensuring that a system discipline lines. As a consequence, multi- and cross-
design possesses these qualities lies with system disciplinary studies are generally discouraged,
engineers. It is advocated here that this should be especially for the creative, young, tenure-seeking
their core concern. faculty which would normally be expected to
contribute the most to the development of a new
A second theme of this work is that, lacking arena. Academic silos, and the rewards for
quantitative means and effective analytical methods remaining within them, are alive and well in the
to deal with the various attributes of design elegance, preponderance of engineering institutions.
the development of successful complex systems is
today largely dependent upon the intuitive skills of But if academic institutions and traditions, and
good system engineers. System engineering today, academics themselves, are to remain relevant, they
with regard to the key attributes of design elegance must address the real problems the world is facing.
identified here, is in a position analogous to that of We in academia must be prepared to devote our
civil engineering prior to the development of the efforts to those problems, rather than others which
theory of strength of materials. Structures were might be more traditional or comfortable, and we
designed and built according to rules of thumb must develop and train the students who can solve
which were derived from prior successes and failures, them. If we are to make critical contributions to the
and passed down from master to apprentice. Good critical challenges which will shape our future, we
structures were those that did not fall down. must understand the interaction between and among
different technologies, organizational management,
System engineers today are in this same position; contract management, public policy, national politics,
good systems are those which do not fail. But today, funding strategies, and the role of these interactions
we have a sophisticated theory of the strength of in the behavior of large scale systems.
materials, and of structural design generally; we
know how to design optimal structures for a wide If we are to remain relevant, if we are to continue to
range of user-specified optimality criteria. In system influence policy and process, we must learn to treat
engineering, we do not have a corresponding theory the interactions between disciplines with the rigor we
for the design of optimal systems. We do not even have, over centuries, learned to apply to the
have an accepted definition of what is meant by an disciplines themselves. We must learn to treat
optimal system. Such things are, by and large, context itself as a discipline. We are not yet doing
matters of intuition and judgment. that, and we certainly are not teaching it. Today, we
are training system engineers to be the engineering
While it is this authors view that the role of human equivalent of financial accountants. We are not
intuition, of engineering judgment, will not soon be teaching them to be Chief Financial Officers.
replaced by algorithms and analytical methods, it
would be well for the system engineering community The second area where improvement must be realized
to develop better and more appropriate theories, tools lies within the community of engineering practice,
and methods to augment that judgment. To do this, which also bears responsibility for the present state of
advances will be required in at least four broad areas. system engineering. In the real world of engineering
development, projects are managed to fruition

Page | 6
through a variety of organizational, legal, and that legal contracts as presently structured do not
contractual tools, many of which are antithetical to constitute the best tool for the management of
the fielding of an elegant design. engineering development projects. Present methods
of system management tend to focus attention on
Experienced practitioners are deeply familiar with what is required, rather than what is important; i.e.,
requirements-based design, the bedrock of doing things right vs. doing the right things. The
engineering and management of large systems. For community of practice must develop management
example, in traditional requirements-based design, it methods that favor the development of elegant
would be customary to specify both the maximum system designs, rather than just legally defensible
allowable cost and mass of a given subsystem. Such designs.
requirements are hard constraints; in theory they
cannot be exceeded, although in practical systems it A third area where advances in system engineering
is often found to be otherwise. Nevertheless, it is in are required concerns the structure and operation of
this manner that the value which is placed on these the engineering design team itself. Whether or not
quantities by the customer is provided, through the the paradigm of the lone inventor is or ever was
system engineer, to the design team. The team reflective of reality, that paradigm is the antithesis of
accepts these constraints, and generally strives to modern system engineering, which is intrinsically a
produce designs which fall within them. team effort. Accordingly, any comprehensive theory
of system engineering must include the study of the
However, it is possible to describe the value human interactions which lead to the final result of
associated with various system properties mass and that effort. Even a cursory study of such interactions
cost in this case by other means. In classical is sufficient to show that the organization and
optimization theory, these properties would be operation of the engineering team itself has a
associated with appropriate weights, selected profound influence on the finished product of that
according to the relative importance assigned to those team.
attributes. In turn, the weighted attributes are
combined analytically to yield an objective One interesting example follows from Arrow's
function; the optimal design is that which produces Theorem10, 11 and the theory of social choice.
the minimum value of the objective function. Note Arrows Theorem shows that, under very general
that there need be no hard constraints in the conditions, decisions involving three or more options
objective function, though these can be included if which are made by means of pairwise comparisons
necessary. More typically, however, the attribute can lead to paradoxical outcomes. Common
weights and the form of the objective function are examples may be found in democratic elections
chosen to penalize more heavily those designs which where there is a significant third party candidate.
utilize too much of one resource or another, yet In the U.S., recent examples include Bush-Clinton-
without imposing hard limits on such use. Perot in 1992, or Bush-Gore-Nader in 2000; in each
case, the actual winner of the election was rejected by
Of interest here is that Collopy has shown9 that a majority of those who voted.
strategies designed to impose firm requirements
which are allocated to individual system elements Such paradoxes in the selection of alternatives were
yields, on average, results inferior to those obtained first studied in the 18th Century by French
using optimization techniques. This result runs Academician J. C. Borda12, who also proposed a
counter to both intuition and established practice. decision rule which avoids them. In Borda counting,
Firm requirements and a logically consistent flow- voting is not pairwise; it is not either-or. If there
down of those requirements, with carefully are, for example, three alternatives, each voter
established parent-child relationships, are at the core expresses a preference for all outcomes, with (2, 1, 0)
of present-day system engineering practice. Yet points assigned respectively to the voters (1st, 2nd,
theory suggests that such practices produce results 3rd) choice. The decision rule is then to select the
inferior to those which would be obtained using alternative which receives the highest point total. In
alternative methods. this scheme, in contrast to that of pairwise voting, no
information is rejected; the ranking of all alternatives
Proponents of requirements-based system design and by all participants is included in the decision rule.
engineering will note that firm, documented However, to the best of this author's knowledge,
requirements are essential to the management of Borda counting is not used as a decision tool
contracts. This may well be so; however, if we are to anywhere in engineering. Indeed, it may be observed
improve the practice of system engineering, it may be in practice that most engineering tradeoff decisions

Page | 7
are made through sequences of pairwise comparisons engineers. To obtain these answers, it will be
-- exactly what Arrow's Theorem shows cannot work. necessary to reexamine our existing research
establishments and paradigms.
This result has significant implications for how we as
system engineers structure our design teams and In the decades since World War II, scientific and
management plans. However, until and unless we technological advancement in the United States and,
begin to delve into the social and cognitive aspects of more broadly, the western democracies has followed
how engineers work together and how system a paradigm in which government sponsorship of
engineering is performed, these and other possible academic research leads to powerful industrial
contributions from fields far apart from engineering applications. This paradigm has served us very well
will continue to go unrecognized. indeed. In the authors opinion, it has worked as well
as it has in part because the relevant questions in
Once this is understood, other questions arise; e.g., most individual engineering disciplines are
what is the role of personality type? It is readily understood, and understandable, by all parties.
observed that many technical professionals excel at, Academic researchers in a given discipline
and take great satisfaction from, the acquisition of understand what their counterparts in government
great expertise in particular disciplines. The value of and industry do in that discipline.
such expertise in designing and developing any
complex system does not require further discussion. The same is not true of system engineers and system
Similarly, it will be observed that some whose early engineering. Academic researchers and research
work is established in a given discipline will later teams are rarely, if ever, exposed to the actual
pursue technical and career breadth rather than ever- practice of system engineering as it occurs on a major
increasing depth. This sort of expertise is rarely development program. Similarly, few if any
recognized, particularly in academic circules, yet it is successful practicing system engineers have either
from this latter population that system engineers are the inclination or, frankly, the classically required
generally drawn. academic credentials necessary to make a transition
to academia. Finally, it is rare or non-existent to find
It may be observed that, today, it is easier to identify graduate students the lifeblood of academic
good system engineers than it is to define good research involved in a significant way in the
system engineering. The question then naturally development of large, complex systems. Because
arises: can system engineering, with the more these necessary conditions have not been present in
holistic view required to pursue design elegance as the development of system engineering as they have
considered here, actually be taught? Can system in other engineering disciplines, the rate of
engineers be trained? Or must they be identified? Is advancement has been correspondingly less.
there a useful middle ground?
If, as claimed here, our success as a society is
If system engineering is at its core a holistic increasingly dependent upon the mastery of complex
discipline, if it is to be fundamentally concerned with system design, and if we believe there is progress to
design elegance, and if breadth of viewpoint is be made in system engineering through the use of the
important to attaining that goal, then what is the role same government-academia-industry coupling that
and value of social, cultural, age, and gender has served us so well in other areas, then it will be
diversity on the large design engineering teams which necessary to find ways to expose academicians and
are required for the successful development of their students to the real problems confronting those
complex systems? Quite apart from the overtly who actually practice system engineering, and to find
political way in which the importance of diversity ways to expose practitioners to the useful results
is generally couched, is there real, quantifiable, which will arise from properly focused academic
engineering value to be had by purposefully research.
designing the design team with the inclusion of
diverse views as a goal? V. CONCLUSION
Properly understood, the core purpose of the
This brings us to the fourth area where advances are discipline of system engineering, and the primary
required if the understanding and practice of system responsibility of the system engineer, is the fielding
engineering is to be improved. The questions raised of an elegant design. As discussed here, an elegant
here are questions of social science and cognitive design is one which produces the intended result, is
psychology as well as of engineering, and the both robust and efficient, and generates a minimum
answers will matter to society at large as well as to of unintended consequences. Considerable work

Page | 8
remains ahead to define and quantify these attributes 4. Lee, Gentry, Personal Communication, Jet
in ways which are meaningful to the engineering Propulsion Laboratory, Pasadena, CA.
community, and to develop methods and tools by
which these goals may be attained. 5. Frosch, Robert A., A Classic Look at Systems
Engineering, Speech to IEEE Group on
To accomplish these results, it will be necessary to Aerospace and Electronic Systems, IEEE
enlarge the view of academia to include the International Convention, New York, 26 March
interdisciplinary and multidisciplinary studies 1969. Reprinted in Readings in System
required for both research and education in system Engineering, NASA SP-6012.
engineering. It will likewise be necessary for the
community of practice to develop methods to 6. Perrow, Charles, Normal Accidents: Living with
incorporate the results of improved theories of system High-Risk Technologies, Princeton University
engineering in the management of actual projects and Press, Princeton, New Jersey, 1999.
programs. The study of human interactions,
cognitive psychology, social choice theory, and other
7. Rockwell, Theodore, The Rickover Effect, John
disciplines must be included in the development of
Wiley & Sons, New York, 1995.
effective theories of system engineering. Finally,
there must be closer collaboration between those who
practice system engineering as it occurs in the real 8. Augustine, Norman, Augustines Laws, 6th ed.,
world, and those who will teach and study it. American Institute of Aeronautics and
Astronautics, 1997.
If we can do these things, we may ultimately be able
to produce designs which meet Froschs standard of 9. Collopy, Paul, Adverse Impact of Extensive
elegance in a more systematic manner. Attribute Requirements on the Design of
Complex Systems, AIAA 2007-7820, AIAA
Aviation Technology, Integration, and
REFERENCES Operations Conference, Belfast, Northern
Ireland, 18-20 Sep 2007.
1. Johnson, Stephen, The Secret of Apollo, Johns
Hopkins University Press, Baltimore, Maryland, 10. Arrow, Kenneth J., Social Choice and
2002. Individual Values, New York: John Wiley and
Sons, 1953, 1961.
2. Petroski, Henry, To Engineer is Human: The
Role of Failure in Successful Design, Vintage 11. Saari, Donald G, Disposing Dictators,
Books, New York, 1992. Demystifying Voting Paradoxes: Social Choice
Analysis, Cambridge University Press, New
3. Petroski, Henry, Success Through Failure: The York, 2008.
Paradox of Design, Princeton University Press,
Princeton, New Jersey, 2006. 12. Borda, J. C., Memoire sur les elections au
scrutiny, Histoire de lAcademie Royale des
Sciences, Paris, 1781.

Page | 9