Implicit vs. Explicit Anticipation
Implicit (Behavioral) vs. Explicit (Representational) Anticipation
Implicit Anticipatory Behavior
It is not true that only an anticipatory system can show anticipatory behavior; some animal conducts (e.g. reacting to a noise by escaping) can be evolved without any mechanism for anticipating future stimuli (e.g. the future danger). Pezzulo and Castelfranchi (2007) argued that those are still implicit anticipatory behaviors (i.e. anticipatory conducts without an anticipatory representation), since their origin is functionally anticipatory. The animal, in fact, apparently reacts to the current stimulus (the noise) but this mechanism was evolved because of its function of dealing with a future event (the danger); thus, functionally the animal acts in view of the future even if a representation of the future is not implied.
Even a dynamical systems approach focused on evolution can account for this kind of phenomena. Almost any evolved structure is implicitly anticipatory, in the sense that it manages the coordination of proximal events to realize distal effects. It is thus the evolved structure that implicitly encodes anticipation and permits to be ready to coordinate with effects occurring in the future, without a need for representing them. However, we argue that not all anticipatory behavior works in that way: the successive functional steps involve explicit representations of anticipated future states (see also Pezzulo and Castelfranchi 2007). Implicit anticipatory behavior can be the case of habits for animals, or behaviors related to homeostatic variables (even if represented as set points or attractors in dynamical systems). But if the right structure is not already in place it is impossible to use it for coordinating with a distal effect. This is the case of an agent realizing a novel goal, i.e. a situation which is never experienced. Even if structures permitting implicit anticipatory behavior may achieve a good generalization over novel situations, the range of anticipatory behavioral capabilities they can express is not comparable with that realized in natural cognition.
Explicit Anticipatory Behavior
Butz et al. (2003) proposed to distinguish two forms of anticipation implying explicit representations, perceptual and state-based anticipation. Many cognitive functions are not possible without a state-based anticipation such as planning via mental simulations, and conceiving novel goals. In the case of goals, when a novel state of affairs is desired and has to be achieved by an agent, the right structure to deal with future stimuli has to be shaped on-the-fly. In order to do so, the agent has to anticipate (at least to a certain degree of abstraction) the state of affairs to achieve in order to prepare its resources; once this is done, the goal state acts as a regulatory mechanism for the system.
The distinction between implicit and explicit anticipatory behavior serves to draw a distinction between different kinds of anticipatory phenomena all existing in nature, and to analyze how they are modeled in artificial cognition. In brief, the whole spectrum of anticipatory capabilities occurring in nature is, up to the moment, only possible by means of building up and maintaining (partially) internal models of external reality and manipulating them in ways which are disengaged from situated action: this is the case of explicit anticipatory behavior.