ble to pass through a trial and to know a little more about it. Experi-
ments, as any researcher worthy of the name knows quite well, are dif-
ficult, uncertain, risky, and never allow recourse to reliable witnesses*
who would be available from a catalog, as it were. They can fail; they
are difficult to reproduce; they depend on instruments. A bad experi-
ment is not one that fails, but one from which the researcher has
drawn no lesson that will help prepare the next experiment. A good
experiment is not one that offers some definitive knowledge, but one
that has allowed the researcher to trace thecritical pathalong which it
will be necessary to pass so that the following iteration will not be car-
ried out in vain.
To use the notions of experimentation and learning curve advisedly,
we must of course take them out of laboratories and share them with
the whole set of those beings, humans and nonhumans, who turn out
to be involved in them. Up to now, under the modernist regime, exper-
iments were undertaken, but among scientists alone; all the others, of-
ten in spite of themselves, became participants in an enterprise that
they lacked the means to judge. We shall say, then, that the collective
as a whole is defined from now on ascollective experimentation.Experi-
mentation on what? On the attachments and detachments that are go-
ing to allow it, at a given moment, to identify the candidates for com-
mon existence, and to decide whether those candidates can be situated
within the collective or whether they must, according to due process,
become provisional enemies. The entire collective has to ask itself
whether it can cohabit with so-and-so, and at what price; the entire
collective has to inquire into the trials that will allow it to decide
whether it is right or wrong to carry out that addition or subtraction.
The deliberations of the collective must no longer be suspended or
short-circuited by some definitive knowledge, since nature no longer
gives any right that would be contrary to the exercise of public life.
The collective does not claim to know, but it has to experiment in such
a way that it can learn in the course of the trial. Its entire normative
capacity depends henceforth on the difference that it is going to be
able to register between t 0 andt+1 while entrusting its fate to the
small transcendence of external realities.
We shall be told that the norm at stake here is a very fragile one, and
that the entire characterization of history cannot be entrusted to such
a weak difference, to a meredeltaof learning. But with respect to what
POLITICS OF NATURE
196