542 Tim Lewens
ics. This was the time when artificial systems were first constructed that could
monitor their surroundings and make responsive adjustments to their behaviours:
‘Some machines are intrinsically purposeful. A torpedo with a target-seeking mech-
anism is an example. The term servomechanisms has been coined precisely to
designate machines with intrinsic purposeful behavior’ [Rosenbluethet al., 1943,
19].
Cybernetic accounts of goal directedness drew on the successes of these en-
gineering endeavours, and attempted to characterise goal-directedness in terms
of feedback loops. Critics were quick to pounce on problems that derived from
Rosenbluethet al.’s assertion that ‘All purposeful behavior may be considered to
require negative feed-back. If a goal is to be attained, some signals from the goal
are necessary at some time to direct the behavior’ [Rosenbluethet al., 1943, 19].
This requirement makes room for the failure of a system to attain its goal (a mis-
sile may explode before it reaches its target), but it rules out the possibility of a
system having a non-existent state as a goal. A missile cannot receive signals from
an object that does not exist, and an insect cannot be said to be seeking food if
there is no food in its local environment. So the feedback requirement has the
awkward consequence that organisms are only seeking food when food is present.
Note, also, that signals of the relevant type cannot be transmitted after they are
received. This rules out using a simple feedback analysis to make sense of the ap-
parent goal-directedness of development. An adult state capable of reproduction
cannot send signals to the developing embryo, because the adult state, assuming
it comes to exist at all, post-dates the developing embryo.
Perhaps we could resolve this problem by looking for some internal representa-
tion, within the goal-directed system, of the goal in question. Feedback loops could
then monitor whether the represented goal state is drawing nearer and direct be-
haviour accordingly. The main problem here, of course, is to give some sense to the
idea of an internal representation of a goal state. The agent model of the organism
that invites development to be characterised in terms of goal-directedness does not
offer much immediate help here. On the face of it, what makes it the case that
an agent has some particular goal is a plan that guides the agent’s actions. And,
of course, plans can be unsuccessful in their execution, or they can be directed at
goals that do not exist, as when one sets out to photograph the Loch Ness Monster.
Now maybe there are some animals — chimpanzees, other primates, dogs, some
birds — that have intentional states, and that can form plans as a result. When
we describe these animals in goal-directed terms (“The dog is trying to get the
bone”), the plan-directed account of goal directedness might work, but we could
not understand the unfolding of animal or plant development as guided by a plan
in this intentional sense.
One possible response here is to fall back onto the selected effects account (or
some other artefact-based account) as a means of making sense of developmental
plans. On this view, we should think of the developmental plan for a plant as
similar to the plan for a building — the plan is directed towards a particular
outcome in virtue of the design process that has given rise to it. In the case