64 Françoise Longy
amateur tests arrive at the conclusion that these are indeed so badly conceived that they
cannot provide any useful information as far as cars are concerned. They provide informa-
tion about what happens in conditions very different from the ones met when driving a
car in a real-life situation. If the experts are right, then a selection of airbag-trigger models
made on the basis of such tests will not be relative to car-airbag-trigger effects and so will
not be able to ground any car-airbag-trigger function.
The previous thought experiment could apparently yield the following conclusion: the
amateur case reveals the very nature of prototype testing. The argument would be that
prototype testing concerns real effects, but not real-life effects. For this reason, it could
be argued, prototype testing can ground no function relative to real-life effects; it can only
justify a rational expectation about real-life effects by considering somewhat related
effects. This is neither a sound argument nor a sound conclusion. The difference between
real-life situations and simulations can in fact be very small, all the more so if the accuracy
of the simulation conditions is itself continually being improved, as is certainly the case
with the best labs for car-crash testing. The testing of new medicines also gives a good
example of the continuity between artifi cial and real-life conditions. The testing of medi-
cines is usually organized in different stages starting from various chemical lab tests, going
on to testing on animals, and fi nishing with single-blind or double-blind tests supposedly
carried out in real-life conditions. Once again, the lesson learned from our investigation
is that no clear-cut separation is to be found. There is a continuous line going from real-
life situations to artifi cial test conditions. At one end there is selection in real life, a little
further on there are tests done in “controlled (real-life) situations” (tests made in hospitals,
supervised fi eld trials, etc.), the next stage involves the tests made in good labs of big
fi rms or big research teams, until fi nally we arrive at the other end of the spectrum with
the poor tests made by amateurs or in unsatisfactory conditions. This spectrum shows that
there really is no ground for separating good prototype testing from real-life selection. So
if the test made is a good one, there is nothing against categorizing the effects relative to
which the prototypes are tested as they would have been in the corresponding real-life
situations. To conclude, there is no serious objection that could be brought against the
statement that M12 has been tested in relation to car-airbag-triggering effects if the tests
conducted were good ones.
Inventing often involves tinkering and this provides another argument against separating
the invention period from the “normal life” period when artifacts are reproduced but often
also modifi ed. There is often no clear answer to the question “Is X a new device
with a new function or is X a new specialized version of an already existing device?”
For instance, a car airbag trigger may result from modifying a trigger used in planes.
Suppose that an airplane trigger has been modifi ed for trains, then for trucks, and then
for motorcycles, and that these modifi cations have been successful every time (the
modifi ed triggers have demonstrated their ability to work in real life as expected). Suppose
now that the same plane trigger is modifi ed to be introduced in cars. How should