Commentary on Marc D. Lewis paper: “Bridging emotion theory and neurobiology through dynamic systems modelling”
Abstract: 45 words
Main Text: 954 words
References: 90 words
Total Text: 1093 words
Tel.: +39 091 6598043
mailto:chella@unipa.it
http://www.csai.unipa.it/chella/
Conceptual space is proposed as an intermediate representation level between the psychological and the neurobiological levels of descriptions of appraisal and emotions. The main advantage of the proposed intermediate representation is that the appraisal and emotions dynamics are described by using the terms of geometry.
Lewis proposes two levels of description of appraisal and emotion dynamics. The higher, psychological level is characterized by perception, attention, evaluation and reflection for the appraisal process, and by arousal, action tendency and feeling tone for the emotion process (Fig.1 of Lewis paper). The lower, neurobiological level is characterised by the interaction among several brain’s parts and circuits.
An intermediate “conceptual” level of representation of appraisal and emotion is proposed and discussed, based on conceptual spaces (Gärdenfors 1997). A conceptual space is a geometric level of concept representation which is intermediate, in the sense of (Jackendoff 1987), between the lower subsymbolic level characterized by descriptions in terms of dynamics of neural networks, as in the Lewis neurobiological level, and the higher level characterized by linguistic descriptions of emotion dynamics as in the psychological level of Lewis.
As sketched below, the conceptual space level of representation has all the capabilities to describe the perception, attention, planning and reflection processes described by Lewis as the basis of appraisal. Moreover, the conceptual space may be easily generalized in order to represent emotions.
The main advantage of this intermediate description is that the appraisal-emotion dynamics described by Lewis may be expressed in terms of geometry, i.e., in terms of vectors, dimensions, geometrics operators, metric functions, etc. Geometric descriptions of cognitive processes are, on the one side, easy to model and to manipulate, as discussed in details in (Gärdenfors 1997), and, on the other side, they may be immediately implemented in an artificial agent by standard geometric programming techniques.
A conceptual space is a metric space whose dimensions are related with the quantities processed by the agent sensors. Examples of dimensions could be colour, pitch, volume, spatial co-ordinates. In any cases, dimensions do not depend on any specific linguistic description: a generic conceptual space comes before any symbolic-propositional characterization of cognitive phenomena.
A knoxel (in analogy with pixel) is a point in the conceptual space and it represents the epistemologically primitive perceptive element at the considered level of analysis. In an implemented robot vision system (Chella et al. 1997), in the case of static scenes, a knoxel corresponds to a geon-like 3D geometric primitive (Biederman 1985). The agent itself is a knoxel in its conceptual space. Therefore, the perceived objects, as the agent itself, other agents, the surrounding obstacles, are all reconstructed by means of geons and they correspond to suitable sets of knoxels in the agent's conceptual space.
Conceptual spaces may represent moving and interacting entities (Chella et al. 2000). Every knoxel now corresponds to a simple motion of a geon, expressed by adding suitable dimensions in the conceptual space that describe the variation in time of the knoxel. For example, considering the knoxel describing a rolling ball, the robot's dynamic conceptual space takes into account not only the shape and position of the ball, but also its speed and acceleration as added dimensions (Marr and Vaina 1982).
The example corresponds to a situation in the sense that the motions in the scene occur simultaneously, i.e., they correspond to a single configuration of knoxels in the conceptual space. To consider a composition of several motions arranged according to a temporal sequence, we introduce the notion of action: an action corresponds to a “scattering” from one situation to another one in the conceptual space. We assume that the situations within an action are separated by instantaneous events. In the transition between two subsequent configurations, a “scattering” of at least one knoxel occurs.
A mechanism of focus of attention may be modelled in the conceptual space by letting the agent to suitably scan the current sets of knoxels in order to select the most relevant aspects of a perceived scene.
The dynamic conceptual space lets the agent to imagine possible future interactions with the objects in the environment: the interaction between the agent and a generic object is represented as a sequence of sets of knoxels that is imagined and simulated in the conceptual space before the interaction really happens in the real world. This loop of imagination, simulation and action is at the basis of the planning capabilities of the agent.
Agent self-consciousness may be generated by a second order conceptual space, in the sense that each second-order knoxel at time t corresponds to the inner perception of the first-order conceptual space by at time t-1, i.e., to the perception at a previous time of the configuration of first-order knoxels representing the agent itself and the other current entities.
To summarize, a conceptual space may represent all the processes at the basis of appraisal. The space may be easily generalized towards an “affective” dynamic space in order to represent the emotion components. A suitable number of dimensions may be added that take into account the “affective” evaluations of the perceived entities. In this new “affective” conceptual space, a knoxel or a group of knoxels is now characterized not only by shape and motion, but also by the associated arousal, action tendency, attentional orientation, and so on.
The appraisal-emotion dynamics described by Lewis in terms of triggers, self-amplifications and self-stabilizations may be modelled in terms of dynamics in the conceptual space: a trigger corresponds to the scattering of knoxels; self-amplifications and self stabilizations may be represented by suitable geometric operators controlling the scattering sequences of knoxels due to the growing up and decaying down of the corresponding affective evaluations.
Therefore, the DS processes described by Lewis and related with the appraisal-emotion processes and their influences of the cognitive capabilities of the agent, may be fully described in terms of geometric operators in an intermediate conceptual space. In this intermediate level, the dynamics described by Lewis at the basis of appraisal and emotions give rise to a sort of “affective” geometry.
Biederman, I. (1985), Human Image Understanding: Recent Research and a Theory, Computer Vision, Graphics and Image Processing 32:29-73.
Chella, A., Frixione, M. & Gaglio, S.
(1997) A cognitive architecture for artificial vision Artificial
Intelligence 89:73-111.
Chella, A., Frixione, M. & Gaglio, S. (2000) Understanding dynamic scenes Artificial
Intelligence 123:89-132.
Gärdenfors, P. (2000) Conceptual Spaces. MIT Press, Cambridge, MA.
Jackendoff, R. (1987) Consciousness and
the Computational Mind. MIT Press, Cambridge, MA.
Marr, D. & Vaina, L. (1982) Representation and recognition of the movements
of shapes Proc. R. Soc. Lond. B 214:501-524.