Posts Tagged ‘concepts’

h1

From “ecological theory of concepts”

January 8, 2010

I was reading: Gabora, L., Rosch, E., & Aerts, D. (2008). Toward an ecological theory of concepts. Ecological Psychology, 20 (1), 84-116.

Here is what i find interesting and related with my own ecological thinking.

Traditionally concepts have been viewed as internal structures that represent a class of entities in the world. However, increasingly they are thought to have no fixed representational structure, their structure being dynamically influenced by the contexts in which they arise (Riegler, Peschl & von Stein, 1999).

A concept is defined not just in terms of exemplary states and their features or properties, but also by the relational structures of these properties, and their susceptibility to change under different contexts.

We view concepts not as fixed representations or identifiers, but rather as bridges between mind and world that participate in the generation of meaning.

The approach implies a view of mind in which the union of perception and environment drives conceptualization, forging a web of conceptual relations or ‘ecology of mind’.

Rosch’s theory of graded structure categorization (1973): An extensive program of research has demonstrated that the same form of graded structure applies to categories of the most diverse kinds: perceptual categories such as colours and forms; semantic categories such as FURNITURE, biological categories such as a WOMAN, social categories such as OCCUPATION, political categories such as DEMOCRACY, formal categories that have classical definitions such as ODD NUMBER, and ad hoc goal derived categories such as THINGS TO TAKE OUT OF THE HOUSE IN A FIRE.

Categories form around and/or are mentally represented by salient, information rich, often imageable stimuli that become “prototypes” for the category.

Gärdenfors (2000a, b) has introduced a provocative geometrical approach to concepts. He considers not just binary features or properties, but dimensions (e.g. color, pitch, temperature, weight). He defines domain as a set of integrable dimensions that are separate from all other dimensions. The property ‘red’ is a convex domain in a region defined by the integrable dimensions hue, saturation, and brightness.

State Context Property (SCOP) formalism
Using the SCOP formalism, a description of a concept consists of the five elements:
• A set S = {p, q, …} of states the concept can assume.
• A set M = {e, f, …} of relevant contexts.
• A set L = {a, b, …} of relevant properties or features. (Note that contexts can be concepts, as can features.)
• A function v that describes the applicability or weight of a certain feature given a specific state and context. For example, v(p, e, a) is the weight of feature a for the concept in state p under context e.
A function m that describes the transition probability from one state to another under the influence of a particular context. For example, m(f, q, e, p) is the probability that state p under influence of context e changes to the state q, giving rise to the new context f.

In the SCOP approach to concepts we have introduced the notion ‘state of a concept’. For any concept there exists an infinite number of possible states it can be in. An important notion introduced in the SCOP approach is the ground state of a concept. This is the ‘undisturbed’ state of a concept; the state it is in when it is not being evoked or thought about, not participating in the structuring of a conscious experience.

We say that a context e evokes a change of state in the concept, from a state p to a state q. Borrowing terminology from quantum mechanics we refer to this change of state as collapse. A change of state of a concept that occurs during collapse may in turn change the context.

When a concept interacts with a specific context, it is immediately projected out of the ground state to another state. This means that the ground state is a theoretical construct, such as for example the state of a physical system in empty space. Indeed, one never experiences a concept in its ground state, since always there is
some context present. It is similar to the fact that a physical system is never in empty space. The properties that are actual in the ground state are the characteristic properties of the concept. The influence of context on the state of a concept can be such that even characteristic properties of a concept disappear if the concept is transformed into a new state under the influence of a context.

Each concept (or constellation of concepts) can be considered a context (however unlikely) of another, and highly probable states of one concept can become included as improbable states of another.

Rosch, E. (1973). Natural Categories, Cognitive Psychology, 4, 328-350.
Gärdenfors, P. (2000a). Concept combination: A geometric model. In Blackburn, Braisby, & Shimojima (Eds.) Logic, Language and Computation: Volume 3 (pp. 129-146), CSLI Publications.
Gärdenfors, P. (2000b). Conceptual Spaces: The Geometry of Thought. Cambridge MA: MIT Press.
Riegler, A. Peschl, M. & von Stein, A. (1999). Understanding representation in the cognitive sciences. Dordrecht Holland: Kluwer Academic.

Advertisements
h1

embodied theory of concepts

November 1, 2007

From Claudia Scorolli and Anna M. Borghi Sentence comprehension and action: Effector specific modulation of the motor system

Recently cognitive science and neuroscience claim that cognition is embodied: knowledge is grounded in sensorimotor experiences, and that there is a deep unity among perception, action and cognition (Pecher and Zwaan, 2005).

The theoretical understanding of indirect connection between concepts as symbols of something real, and perceptual experiences we gain with sensory-motor systems, has been put under question.
New, embodied view of concepts considers perceptual symbols as neural representations located in sensory-motor areas in brain. This means concepts are not perceived as arbitary symbols but rather concepts consist of the reactivation of the same neural activation pattern that is present when we perceive the objects or entities they refer to and when we interact with them.

Object attributes are thought to be stored near the same modality-specific neural areas that are active when objects are being experienced (Martin, Ungerleider and Haxby, 2001).

Symbols, according to the embodied view, are not amodal, but multimodal – for example, they refer both to the tactile experience of caressing a dog as well as the auditory experience of hearing a dog bark (Barsalou, 1999; Gallese and Lakoff, 2005).

Concepts make direct use of sensory-motor circuits of the brain (Gallese and Lakoff, 2005).
The same neural areas are involved when forming motor imagery and when activating information on objects, particularly on tools.

Pecher, D., and Zwaan, R.A., 2005, Grounding Cognition: The Role of Perception and Action in Memory, Language, and Thinking, Cambridge University Press.

As a researcher with the background of natural sciences, i wonder what happens if the symbol, what we read, is not the first order symbol, directly related with the objects (real dog – word dog), but if we describe in the scientific text some micro- or macrolevel objects and phenomena (eg. cells, genes, evolution), or if the phenomena we describe are highly of abstract nature (photosynthesis). How do we embody these concepts?

Or, if we read not the scientific narrative but scientific images (graphs etc.) or scientific formulated symbol language (math or physic formulas, chemistry reactions). Presumably we need to process arbitrary concepts somehow and link them to embodied concepts at certain moment in order to grasp the thing?

If we think of scientific narratives (papers, books), we can say that they have reduced action potentialities, or embodied concepts in them, and the difficulty of the reader of such texts is to perform cognitive symbol processing to relate abstract concepts with limited affordances for action-potentialities internally with the embodied concepts that are related with real world perception. And supposedly only then we can understand the text.