Abstract
A general goal of biologically inspired robotics is to learn lessons from actual biological systems and to find applications in robot design. Neural controllers and adaptive algorithms are major tools to model, at some level of abstraction, functions, structures, and behaviors present in biological systems. This involves, of course, identifying in virtue of what biological systems exhibit the behavioral characteristics we want to explore. One of the biological phenomena of great interest is emotion. Despite the effort of leading researchers to raise the question “whether machines can be intelligent without any emotions” (Minsky., 1988), AI interest in emotional phenomena has increased only in the last decade. An underlying assumption is that many cognitive functions, such as memory, attention, learning, decision making and planning, are at least partly based on emotional mechanisms in biological systems (Damasio, 1995). One of the qualities of emotional behavior is its flexibility (Frijda, 1986), which contrasts with the rigidity of stereotyped behaviors such as reflexes or habits. Hence, it is relevant to investigate what it is that makes emotional behavior flexible. The body, through mostly chemical channels, produces diffuse effects on the neural system, processes at the root of emotional phenomena. Parisi has recently argued that in order “to understand the behavior of organisms more adequately we also need to reproduce in robots the inside of the body of organisms and to study the interactions of the robot’s control system with what is inside the body” (Parisi, 2004), using the term internal robotics to denote the study of the interactions between the (neural) control system and the rest of the body. Mechanisms that control homeostasis, based on hormonal modulation, can motivate appropriate behaviors (Avila-García & Cañamero, 2004; Gadanho & Hallam, 2001). Emergent behaviors from the interaction of a motivational system with the environment may be called emotional. Cañamero’s architecture, for example, consists of “a set of motivations; a repertoire of behaviors that can satisfy those internal needs or motivations as their execution carries a modification in the levels of specific variables; and a set of ‘basic’ emotions.”(Cañamero, 2005). We consider emotional phenomena to emerge from a dynamic interaction between internal states, current perceptions and environmental relations, such that certain neural/physiological states have a close causal link with relational situations. This is, in a nutshell, the embodied appraisal hypothesis (Carlos Herrera, 2002; Prinz, 2004). We use two major concepts from the dynamical systems (DS) approach to cognition (Clark, 1997; Kelso, 1995): collective variables and control parameters. In (Carlos Herrera, 2002) we argue that internal states can be interpreted as collective variables of agent/ environment interaction that allow tracing concern-relevant situations. These variables are “non-specific: they do not prescribe or contain a code for the emerging structure” (Kelso, 1995). They also can be considered control parameters, as activation in the agent’s physiological substrate affects overall action readiness (response, including perceptual and cognitive readiness).
TopBackground
An architecture for the design of emotional appraisal and response in artificial agents must take into account that emotions bear an intrinsic dynamic relationship between internal mechanisms, embodiment and situation (Frijda, 1993; Lazarus, 1991; Lewis, 2005). Emotions are emergent patterns that involve relational behavior as well physiological and psychological processes. In this section we argue that physiological states are essential for understanding emotion appraisal and response: they allow to trace agent-environment relations, and their modification is a mechanism for control of dynamics.
Key Terms in this Chapter
Action Readiness / Tendencies: Physiological states affect the readiness for engagement in certain dynamics of the interaction
Neuro-Robotics: Approach to robot control through the use of neural networks.
Artificial Emotion: The attempt to synthesize in robots or artificial systems some of the functional properties of emotion.
Emotions: Phenomena present in biological systems by which an adaptive agent is capable of appraising the concern-relevance of situations and provide flexible responses through generation of physiological, cognitive and behavioral readiness.
Concerns: The conditions under which a system can continue to function.
Collective Variable / Control Parameter: In dynamical systems theory, collective variables allow tracing global dynamic patterns, control parameters lead the system through such patterns.
Embodied Appraisal: Theory that asserts sensitivity to concern-relevant situations is facilitated by physiological and homeostatic mechanisms in an embodied agent.
Hormonal Modulation: Change in the functionality of neural, sensory and motor systems achieved through changes in hormonal levels.