Engineering as social experimentation

Engineering should be viewed as an experimental process. It is not, of

course, an experiment conducted solely in a laboratory under controlled
conditions. Rather, it is an experiment on a social scale involving human
subjects. in April 1912, the Titanic was proclaimed the greatest engineering
achievement ever. Not merely was it the largest ship the world had seen, having
a length of almost three football fields; it was also the most glamorous of ocean
liners, and it was touted as the first fully safe ship. Because the worst collision
envisaged was at the juncture of two of its sixteen watertight compartments, and
as it could float with any four compartments flooded, the Titanic was believed to
be virtually unsinkable. With such confidence, the captain allowed the ship to
sail full speed at night in an area frequented by icebergs, one of
which tore a large gap in the ship’s side, flooding five compartments. Time
remained to evacuate the ship, but there were not enough lifeboats. Because
British regulations then in effect did not foresee vessels of this size, only 825
places were required in lifeboats, sufficient for a mere one-quarter of the Titanic’s
capacity of 3,547 passengers and crew. No extra precautions had seemed
necessary for an unsinkable ship. The result: 1,522 dead (drowned or frozen) out
of the 2,227 on board for the Titanic’s first trip.
The Titanic remains a haunting image of technological complacency. So many
products of technology present potential dangers that engineering should be
regarded as an inherently risky activity.

Engineering as Experimentation
Experimentation is commonly recognized as playing an essential role in the
design process. Preliminary tests or simulations are conducted from the time it
is decided to convert a new engineering concept into its first rough design.
Materials and processes are tried out, usually employing formal experimental
techniques. Such tests serve as the basis for more detailed designs, which in
turn are tested. At the production stage further tests are run, until a finished
product evolves. The normal design process is thus iterative, carried out on trial
designs with modifications being made on the basis of feedback information
acquired from tests. Beyond those specific tests and experiments, however, each
engineering project taken as a whole may be viewed as an
experiment.

Similarities to Standard Experiments

Several features of virtually every kind of engineering practice combine to make

it appropriate to view engineering projects as experiments. First, any project is
carried out in partial ignorance. There are uncertainties in the abstract model
used for the design calculations; there are uncertainties in the precise
characteristics of the materials purchased; there are uncertainties in the
precision of materials processing and fabrication; there are uncertainties about
the nature of the stresses the finished product will encounter. Engineers do not
have the luxury of waiting until all the relevant facts are in before commencing
work. At some point, theoretical exploration and laboratory testing must
be bypassed for the sake of moving ahead on a project. Indeed, one talent crucial
to an engineer’s success lies precisely in the ability to accomplish tasks safely
with only a partial knowledge of scientific laws about nature and society.
Second, the final outcomes of engineering projects, like those of experiments, are
generally uncertain. Often in engineering it is not even known what the possible
outcomes may be, and great risks may attend even seemingly benign projects. A
reservoir may do damage to a region’s social fabric or to its ecosystem. It may
not even serve its intended purpose if the dam leaks or
breaks. An aqueduct may bring about a population explosion in a region where
it is the only source of water, creating dependency and vulnerability without
adequate safeguards. A jumbo airplane may bankrupt the small airline that
bought it as a status symbol. A special-purpose fingerprint reader may find its
main application in the identification and surveillance of dissidents by
totalitarian regimes. A nuclear reactor, the scaled-up version of a successful
smaller model, may exhibit unexpected problems that endanger the surrounding
population, leading to its untimely shutdown at great cost to owner and
consumers alike. In the past, a hair dryer may have exposed the user to lung
damage from the asbestos insulation in its barrel. Third, effective engineering
relies on knowledge gained about products both before and after they leave the
factory—knowledge needed for improving current products and creating better
ones. That is, ongoing success in engineering depends on gaining new
knowledge, as does ongoing success in experimentation. Monitoring is thus as
essential to engineering as it is to experimentation
in general. To monitor is to make periodic observations and tests to check for
both successful performance and unintended side effects. But as the ultimate
test of a product’s efficiency, safety, cost-effectiveness, environmental impact,
and aesthetic value lies in how well that product functions within society,
monitoring cannot be restricted to the in-house development or testing phases
of an engineering venture. It also extends to the stage of client use. Just as in
experimentation, both the intermediate and final results of an engineering
project deserve analysis if the correct lessons are to be learned from it.

Learning from the Past

Usually engineers learn from their own earlier design and operating results, as
well as from those of other engineers, but unfortunately that is not always the
case. Lack of established channels of communication, misplaced pride in not
asking for information, embarrassment at failure or fear of litigation, and
plain neglect often impede the flow of such information and lead to many
repetitions of past mistakes. Here are a few examples:
1. The Titanic lacked a sufficient number of lifeboats decades after most of
the passengers and crew on the steamship Arctic had perished because
of the same problem.
2. “Complete lack of protection against impact by shipping caused
Sweden’s worst ever bridge collapse on Friday as a result of which eight people
were killed.” Thus reported the New Civil Engineer on January 24, 1980.
Engineers now recommend the use of floating concrete bumpers that can
deflect ships, but that recommendation is rarely heeded as seen by the 1993
collapse of the Bayou Canot bridge that cost 43 passengers of the Sunset
Limited their lives.
3. Valves are notorious for being among the least reliable components of
hydraulic systems. It was a pressure relief valve, and a lack of definitive
information regarding its open or shut state, which contributed to the nuclear
reactor accident at Three Mile Island on March 28, 1979. Similar malfunctions
had occurred with identical valves on nuclear reactors at other locations.
The required reports had been filed with Babcock and Wilcox, the reactor’s
manufacturer, but no attention had been given to them.

These examples illustrate why it is not enough for engineers

to rely on handbooks and computer programs without knowing
the limits of the tables and algorithms underlying their favorite
tools. They do well to visit shop floors and construction sites to
learn from workers and testers how well the customers’ wishes
were met. The art of back-of-the-envelope calculations to obtain
ballpark values with which to quickly check lengthy and complicated
computational procedures must not be lost. Engineering
demands practitioners who remain alert and well informed at
every stage of a project’s history and who exchange ideas freely
with colleagues in related departments.