Agent-based modeling and simulation - Modeling and Simulation of Complex Systems - page 20

Tropistic agents are characterized by the fact that they do not
possess a model of the world they are acting in. Therefore, behavior
of tropistic agents is at any point in time completely defined by
their current surroundings and actual sensor inputs [39, p. 448].
In contrast to tropistic agents, hysteretic agents make use of an
internal state, i.e., they keep track of external states and their
sensor inputs and can use this knowledge later on for their choice
of an action [39, p. 454].
This distinction is in good agreement with the classification by Slo-
man [123] who added but one more class:
Reactive agents abstract of internal states and react directly to
their sensor inputs.
Deliberative agents make use of an explicit mental representation of
their perceptions and follow (more or less) sophisticated planning
procedures to choose their next action.
Reflective agents are additionally capable of reflecting previous
choices and are learning from the outcomes of previous actions.
Besides these clear-cut classes hybrid and layered agent architectures
are wide-spread, aiming to combine multiple mechanisms to utilize
the particular benefits [149], [135, p. 335ff.].
2.2.3 Relationships with other modeling paradigms
Characterizing modeling paradigms, the clearest distinction can be
made between macroscopic modeling approaches (e.g., using differ-
ential equations or System Dynamics) and microscopic modeling
approaches like agent-based modeling and simulation [68, 22, 27].
Macroscopic models abstract from individual system components and
use high-level representations of the system under investigation. In
contrast, microscopic models use a very fine-grained representation of
a system and explicitly take into account individual characteristics
and behavior of system components.