The Antiparallel Structures
of Science and Engineering

by Eric Drexler on 2009/06/22

Part of a diagram of contrasting information flows in scientific inquiry and engineering design
The heart of the problem
(Full image below)

Science and engineering are inseparable domains of thought and action, linked by a shared language of mass and energy, molecules and thermodynamics, physical systems and physical law. This shared language makes communication deceptively easy — easy, because scientists and engineers can see every detail in the same way; deceptive, because they see these details in different contexts, forming different patterns and presenting different problems. In a fundamental sense, science and engineering are antiparallel, facing in opposite directions. The resulting gaps in understanding can open a chasm wide enough to trip a manager, or to swallow a project.

This places a premium on minds that encompass both, whether they work as “scientists” or as “engineers”. The modes of thought are fundamentally different, but needn’t clash.

As I discussed in a recent post, scientific inquiry and engineering design are often intimately interleaved (in projects, in activities, in creative minds), and to such an extent that (perilously!) they may seem the same. Here, I will focus on the differences that thread through a complex relationship.

A familiar pattern difficulties at the science/engineering interface often impedes corporate research, and I’m working with a former R&D manager to develop a presentation package that addresses this. In the broader technical community, however, similar difficulties impede progress in understanding what science can and can’t tell us about the future potential of technology, thereby impeding the development of reality-based policies. In both instances, the costs include delay, friction, waste, risk, and missed opportunity, and in both instances, understanding the structural basis of the problem can help to resolve it.

Antiparallel Structures

Both scientific inquiry and engineering design can be dissected into three levels: physical systems, concrete descriptions of physical systems, and general patterns that apply to an indefinitely large number of systems. Here’s a diagram that captures some key relationships:


A diagram of contrasting information flows in scientific inquiry and engineering design

The information flows that link these levels are antiparallel: In scientific inquiry, physical systems shape their descriptions through measurement, and the results constrain and shape general, abstract models (theories) by testing them. In engineering design, by contrast, descriptions (specifications) shape physical systems through fabrication, and general, abstract models (system concepts) shape descriptions through design.

The universal and the particular

These information relationships differ in another dimension: the universal vs. the particular, equality vs. constraint, “For all x, a = b, c = d,…” vs. “There exists x such that ab, cd,…”. On one side is an ideal of generality and perfection that can only be approximated; on the other side is a target that is routinely achieved.

While science aims (ideally) to produce exact descriptions of all parameters of all members of a general class of physical systems, engineering aims to manufacture instances of a single kind of system, making choices to ensure that itsfunctional parameters will equal or exceed those specified by a design description.

Likewise, while science aims to formulate a single theory that exactly fits all parameters of every description, engineering aims to design at least one description of a system having functional parameters that equal or exceed those required by one of a potential multiplicity of system concepts.

In this connection, is a proliferation of possible ways of satisfying a constraint good, or bad? In science, finding more possibilities creates greater uncertainty; in engineering, finding more possibilities provides greater freedom of design. This is a basic question with opposite answers — and there are many more.

Pervasive consequences

Science and engineering share a language of physical systems and physical law, but they ask different questions, seek different knowledge, and serve different ends. The ramifications range from different views of the non-linear system dynamics to differences in working relationships and institutions. The consequences are pervasive and deep, familiar and surprising, and extend far beyond what I have sketched here.

It is hard to quantify or predict the value of modest improvements in mutual understanding among scientists, engineers, and research managers, but the potential value is surely enormous. Modest improvements in understanding and communication can speed progress, reduce risks, and occasionally uncover a transformative strategic opportunity.

I think that a creating a deeper and more widely shared understanding of the contrasting faces of science and engineering can help to produce those modest improvements.


See also:

{ 6 comments… read them below or add one }

stephen June 23, 2009 at 1:21 am UTC

Thinking about the ‘ideal’ of engineering design was an interesting mental exercise. There seems to be some uncertainty about what an ideal even represents in the realm of design. Probably along the lines that the ‘instance’ has negligible cost, both in time and materials, to create, maintain, and reuse/destroy, exists indefinitely and the constituent materials can be reclaimed for unlimited iterations of use. Yet none of these ideal properties exist anywhere in the known universe (thanks, Science). It seems that in most real-world situations, tradeoffs exist between competing and possibly mutually exclusive desirable properties of a given design, e.g. the weight vs. strength of a material. Probably the most relevant to any general engineering project would be the cost of design itself and determining whether or not a given set of requirements is achievable within the context/constraints of the project.
It’s also interesting to note how many, if not all, of these ‘ideal’ properties promise to be improved dramatically by molecular manufacturing.

Eric Drexler June 24, 2009 at 1:05 pm UTC

Stephen, regarding trade-offs, I’d strengthen “most” to “almost always” — a design that is optimal by a metric A is seldom optimal by a different metric B, at least in most areas of engineering. The exceptions tend to arise where the design variables are markedly discrete (rather than continuous or fine-grained), and where the system being designed is relatively simple. Otherwise, there’s almost ways some way to tweak a design to improve it in response to a change in purpose.

In practice, economies of scale and various kinds of friction (including the cost of offering or even noticing an alternative) often make designs stable across many years and many applications.

Regarding idealization, it’s important to distinguish between (1) idealization of the design process, in the sense of a design as a pure activity (separate from inquiry), vs. (2) an ideal situation for design, in the sense of having access to excellent materials, low production cost, great software, etc.

Icy June 30, 2009 at 4:11 pm UTC

A very clear way of thinking about the differences in the two fields. Thanks.

shyama debbarma March 17, 2010 at 1:55 pm UTC

i am an engineering student i really feel to have the concepts of enginerring theoris and its application please may it be suggested for easy understanding

Jeffrey Soreff December 30, 2010 at 2:26 am UTC
farhad February 12, 2013 at 3:04 pm UTC

name of institute in new delhi

{ 25 trackbacks }

Leave a Comment

Previous post:

Next post: