288 Peter Kroes
phenomena must be qualifi ed as epistemically emergent. The epistemically emergent
nature of these phenomena is usually, however, of limited duration, since progress in the
knowledge base, mainly driven by a desire for better control, turns them into nonemergent
phenomena. For such phenomena, control may temporarily be an issue, but they rarely are
considered to constitute a principal threat to the control paradigm. So the history of engi-
neering might be seen as a continuous attempt to turn epistemically emergent technical
phenomena into epistemically nonemergent phenomena.^24
A second point to be noted is the association of unexpectedness with emergence. If we
take due account of the distinction between theoretical and inductive (non)predictability,
this association turns out to be unfounded. Let us call the occurrence of a phenomenon
unexpected if it is at the moment of its occurrence theoretically and inductively unpredict-
able. On this interpretation of unexpected, the occurrence of epistemically emergent fea-
tures in technical artifacts is not necessarily unexpected. It may still be possible to predict
their occurrence inductively. So technical systems may exhibit expected or unexpected
emergent features.^25 From the point of view of control, these are very different categories
of emergent features. Unexpected emergent features are the most problematic because they
take us by surprise; there is no way to control them in advance. They differ from expected
emergent features in terms of control. Expected emergent features are theoretically unpre-
dictable and cannot be explained at a given point in time (remember our restriction to
weak forms of epistemic emergence), but that does not imply that they cannot be controlled
to a high degree. In fact the control of expected emergent phenomena is part and parcel
of engineering practice; it is based on inductively established rules of practice and correla-
tions between properties of the emergent feature and properties of components of the
system (which, on the assumption that these properties may be controlled, may be used
as control variables). Of course if it would be possible to come up in these cases with a
functional decomposition of the whole system, which would be tantamount to an explana-
tion of the emergent feature, then that would probably greatly enhance the possibilities of
control of the system.
Whether “inductive control” is also a viable possibility for features that are epistemi-
cally emergent in the strong sense depends on whether the possibility of regularities (cor-
relations) between properties of the emergent feature and properties of the components in
the emergence base is compatible with the in-principle theoretical nonpredictability and
nonexplainability of (properties of) the emergent feature. We leave that issue open, as well
as the issue of whether there are any technical systems that show those kinds of epistemi-
cally emergent features.
Returning to the issues mentioned in the introduction, we observe with regard to the
second, about emergence and functional decomposition, that weak forms of epistemic
emergence pose much less of a threat to the use of techniques such as functional decom-
position than often suggested. Engineers have always had to deal with weakly emergent
phenomena, and their success in turning them into ordinary, nonemergent phenomena