Concepts Acquisition and Knowledge Organisation
Fátima Regina Brito Uhr, Vânia F. C. de Sá Henriques e Adriana Benevides Soares
Universidade Gama Filho – Programa de Pós-Graduação em Psicologia
Rua Manoel Vitorino, 533 – 20740-280 – Rio de Janeiro - RJ
The present paper is aimed at studying the forming of concepts, as it is also an attempt to understand how these concepts are organized in our memory. In the early works on forming of concepts, psychology was based on research in which, for many years and in various laboratory settings, the subjects, both animals from different species or human beings, had to learn or to identify artificial concepts. The reason for the use of this kind of task was to avoid the interference of the individual’s previous training that would come out of the control of the experimenter, along with a greater simplicity of the artificial concepts compared with the natural ones; this had the effect of facilitating the study of processes by which were formed the artificial concepts.
The concepts studied by Frege (1982, quoted by Pozo, 1998), are defined from the distinction between the “reference” and the “sense” of the concept. Reference is related to worldly facts and objects that indicate it, meanwhile its sense is given by the relationship to other concepts. Recently, Miller and Johnson-Laird (1976, quoted by Pozo, 1998), recovered this distinction by differentiating the “process of identifying” a concept and its core: the process would be linked to attributes of the concept, and the core would be determined by the web of other concepts in which it would be inserted.
There are other definitions referring to these two ways of defining a concept. To our interest, even if there exists any difference between such distinctions, it is of our interest that all coincidentally differentiate between a definition “from toe to head”, derived from its attributes, and a definition “from head to toe”, depending on its relation with the other concepts that make up a semantic network or a “theory”. Another aspect in which coincide most of these distinctions (Frege, 1982; Miller and Johnson-Laird, 1976; Wickelgren, 1979) (quoted by Pozo, 1998), is that the core or the sense of the concept is where are defined their relevant references or attributes. What make the concept of “invertebrate” significant are its differences regarding the concept of “vertebrate”. Most of scientific concepts have a deeply marked relational nature. We cannot understand “volume” without knowing what are “weight” or “size” (Carretero, 1984, quoted by Pozo, 1998), just as the concept of “proletariat” is not understood, unless one also understands the concept of “bourgeoisie” (Pozo, 1985). Overall, we can say that the location of a concept in an individual's cognitive structure will determine the concrete attributes that will define such concept.
Besides occupying a place in the wider structure or theory, concepts themselves have an internal structure and two conceptions: one classical, the other probabilistic.
The classical conception brings out that the concept is made of various necessary and sufficient attributes, in such a way that all examples of concept contain common attributes and no example of concept has such attributes. For the probabilistic conception, concepts have a diffuse structure, there are no attributes, whether necessary or sufficient, to define them, generally viewed as everyday concepts, for example, “pen”, “chair”, etc.
Theories on Concepts Aquisition
According to classical conductists, learning concepts is fundamentally the acquisition of excitatory and inhibitory potentials. Each stimulus element has its own potential, in consequence of its association with a reinforcement. An individual’s response to a stimulus (i. e., its attribution or not to a defined category) is dependent on the total sum of excitatory and inhibitory potentials of the elements that make it. The theory assumes that, at any time, the animal is associating all the stimuli and their components to answers and reinforcements received. Only those elements or stimuli systematically associated with a reinforcement will have an excitatory potential (such as, e. g., “bark” with “dog”), the rest of the stimuli will have a null potential (e. g., “brown colour” with “dog”) or an inhibitory (e. g., “to fly” with “dog”). Similarly, the theory assumes the acquisition of concepts as being gradual or continuous.
For the methodological conductists, most of the operative responses are produced in the presence of certain environment keys or stimuli. When an operative is sent in these conditions and reinforced, all present stimuli elements gain control of the response emission. Control can be extended to other settings containing common stimuli. Acquisition of concepts, however, would be the moulding of behaviour by contingencies so that the stimuli that have the propriety would remind the response while other stimuli would not. In induction (or generalisation), organism activity is an expression defining the fact that acquired control by a stimulus is shared by other stimuli with common properties, or the control is shared with all stimulus properties taken separately. Abstract responses will only be acquired when a reinforcement agent agrees, for in natural instances it seldom stays under isolated property control.
For the mediational conductist theoreticians, the significance of concepts would not be based, generally, in stimulating elements common to the instances of concepts; it would remind of a mediator response, common, rather of verbal nature.
Mediational conductist theories try to explain the origin of significations, but they do so from associationist precepts of conductism.
Mediator responses make up true internal representations of the stimulus, being close to cognitive postures.
These responses continue, however, faithful to the associationist belief in the principle of correspondence and isomorphism: both present the same characteristics found in external responses and its acquisition may be explained by the same words, through the same concepts of acquisition of manifest behaviours. So, these theoreticians go on imagining the concept learning as a process of discrimination and generalisation, where different stimuli remain associated with a response. However, different from the classical conductist conception, concepts are not made of elements of stimuli, but from verbal mediations.
For the constructivist theoreticians, the early versions of concept learning arise within the conductism itself in the form of “selective processes”.
This theory rests on many empirical findings, both related to animal learning and human being forming of concepts.
Howard (1983, quoted by Pozo, 1998) brings out that, in spite of their great diversity, all these theories coincide in that:
I. There is a “bank” of potential hypothesis at the individual’s disposal in the beginning of the problem resolution process;
II. At each attempt, the individual chooses one or more hypothesis among those available and responds from this base;
III. If the chosen hypothesis leads to a correct stimulus classification, it is retained, but, if rejected, it is replace by others of the whole.
Bruner, Goodnow and Austin (1956, quoted by Pozo, 1998), studied the Focus and Examination Strategies, where they pointed out that, at least in the case of artificial concepts, the most efficient strategy is the one that, from an initial concept model or example, may modify the features defining concept progressively.
These authors also refer to the possibility of identifying concepts from two stimuli with similar structures, one abstract, and the other thematic. While the results of the abstract tasks analysis are precise and predictive, that is not the case in the thematic task, as it is not only a list of attributes, references, but it has a meaning, a sense, therefore, its acquisition cannot be reduced solely to formal proceedings.
There has been much restriction over the forming of artificial concepts. The idea that we acquire all concepts by means of a hypothesis verification process is matter of debate, because such theory is based on the assumption that the individual, specifically the human being, extracts his knowledge of reality by applying logic rules.
The idea that the human being is directed by exclusive criteria of logic rationality is seriously disavowed in psychology of thought. There is an abounding literature showing that an individual’s reasoning is influenced by many other variables beside the task’s logic form.
The main theories on acquisition of natural concepts attempt at explaining deficiencies through the view of attribute definition: for example, the effects of typicality and the lack of concept distinction.
According to this theory, the prototype is represented by characteristic attributes; that is, there are no definition attributes, but only attributes of different weights or different degrees of importance within the concept.
The fact that a concept belongs to a category is defined by the similarity of attributes of an object to this category prototype, whether the prototype is represented by characteristic attributes or by category species.
With the prototype theory, Rosch et al. (1976) suggested that the conceptual hierarchies had a structure defined in three levels:
- over-ordinate (e. g., weapons, furniture), basic (gun, chair) and subordinate level for specific concepts (e. g., revolvers, riffles, stool, armchair).
Basic level is where concepts have its greater number of “Distinctive Attributes” and cognitively the most economic level, that is, it is the level in which concept attributes are not shared with concepts of the same level.
The theory of prototype aroused some criticism: (1) not all concepts have prototype characteristics (e. g., abstract concepts like science, crime, etc.); (2) it seems there is an infinite flexibility in belonging or not to these categories, as opposed to concrete categories; (3) theoreticians were more concerned in asking how do we group objects within a category instead of another, instead of explaining why are the categories coherent. In simpler terms, things are grouped in categories because they have certain attributes in common.
The works on acquiring concepts were useful to influence those trying to understand in which way is knowledge organised. In this manner, from these collected data begins a process of representing knowledge organisation.
Many models are offered and criticised, joining to current models that explain, still with some frailties, how we code, store and retrieve information.
Knowledge Organisation
Collins & Quillian, 1969, state that research on acquisition of concepts helped promote an economic model cognitively, dividing the environment in classes of things so as to decrease the volume of information we need to learn, perceive, remember and recognize.
Most of research on knowledge organisation had to find out how do we organise instances in concepts and relate these concepts together as its goal.
In 1969/1970, Collins and Quillian developed a theory and a computational model that were a vision from the definition attribute viewpoint. By attempting to verify this proposal, they represented concepts as hierarchical nets. The basic viewpoint on concepts assumed by the theory was essentially the definition attribute viewpoint.
Concepts are represented as node hierarchy of interconnected concepts (e. g., animal, bird, canary). Some nodes (e. g., bird is an over-ordinate of canary, and animal an over-ordinate of bird) are therefore subordinate of others (e. g., canary is a subordinate of bird).
Collins and Quillian assert that when someone verifies a concept, he is trying to go from a node to another, leading to the prediction that, the greater the distance between nodes, the greater the length of time to verify the phrase, like: Is a canary an animal? These predictions were confirmed and, in the instance/attribute case, the attribute location in the hierarchy related to the mentioned instance served to predict the necessary time to verify the phrase.
This theory received two main criticisms: 1) it could not grasp significant features of conceptual behaviour, and 2) the central conception of this viewpoint – that concepts lie on a conjunction of essential characteristics – is simply wrong.
Eysenck and Keane (1994) think these results do not annihilate the attribute definition viewpoint. They are still coherent with the proposition that the concept involvement rests on a conjunction of necessary attributes.
On the other hand, some researchers like Fodor, Garret, Walker and Parkes, 1980, Wittgenstein, 1998, argue that every attempt made to separate concepts into essential and self-sufficient is fundamentally misconceived.
Collins and Quillian’s work had an influence on researches made to consider relational concepts as well as object concepts. In 1975, Collins and Loftus proposed a modified web model to tackle the insufficiencies of the previous model. From this perspective of relational concepts, this model was important because it introduced a variety of tagged interconnections among web nodes.
However, independent lines of research on relational concepts had already been established by researchers in linguistics and computational science, and these were ideas that brought new understandings for the psychological work.
Indeed, human knowledge is formed by more than information similar to attributes concerned with concepts or individual relationships. It is plausible to assume that this knowledge is organised in a much more complex form than hierarchical groupings or similar to concept classes.
The outline is a construct normally used to explain the complex organisation of knowledge. The outlines are structured groupings of concepts and generally contain generic knowledge on the most diverse events. They were used to explain our ability in producing inferences in complex events, work out patterned assumptions on unmentioned events aspects and originate predictions on what will possibly happen in the future.
The outline variables may be the script and the frame. The script theory and its successor, the dynamic memory theory, proposed characterising the knowledge people have of the sequences of trivial events, like “going to the supermarket and doing some shopping”. The individual already brings his script complete in the head (get in, take a cart or a basket, look for what is needed, go to the check-out, pay and leave the supermarket).
Rumelhart and Norman (1980, 1981) defended the existence of three basic ways in which learning may happen within a system based on outlines:
1. Incorporation – we remember the instance existing already in our knowledge;
2.Tuning – we develop and refine concepts of an outline by means of experience;
3. Re-structuring – involves the creation of a new outline or by means of analogy or an outline induction.
Schank and Abelson (1977) still defended that scripts are formed only from direct personal experience; thus, on the contrary, few of us would be able to understand an event like a bank hold-up once few of us experienced such a situation. However, we can clearly understand this scenery. It means we have to have a group of more abstract structures that allows us to superimpose ourselves to rigid structures of scripts and understand others’ actions and goals in situations we never experienced personally.
Schank (1982) thus re-organises the script theory, putting on concepts like Memory Organisation Pack (MOPs) – made of generalised conglomerations of events called scenes. However, instead of having an assembly of components organised in a script, MOPs organised the assemblies of scenes and added up specific contextual information.
Final Considerations
Theories on knowledge organisation have been, in the course of time, object of study for many researchers. Recent works, in the last 30 years, have emphasised theories based on concepts like definition, prototype, constructs (outlines) and dynamic memory attributes to explain knowledge organisation.
Cognitive science has used empirical tests, formal tests and computational models to test and contest already existing theories or confirm them.
In the text, the authors express the opinion in which, up to now, not one of the represented theories is sufficiently extended to embrace cognitive processes of perceiving, learning, remembering and recognising.
Even so, the important is that progress has been made, especially because researchers do not go astray in fruitless debate on basic philosophic positions.
Yet the authors suggest the need of deeper works, capable of joining two distinct traditions in research: knowledge complex organisations and simple organisations, explainable, because researches on simple organisation seem to point out that simple concepts may depend on complex structures and knowledge.
Finally, Eysenck and Keane (1994) point in the direction of some unified mechanism that could explain the ways people form categories, hierarchies and other forms of outline organisation.
These two trends of research should be integrated, but in other hand a theory of knowledge cannot escape from the rational/emotional confrontation inherent to the human being. It is under this logic and affective condition that cognitive processes happen.
BIBLIOGRAPHY
EYSENCK, M. W. & KEANE, M. T. (1994) Psicologia Cognitiva; Porto Alegre, Artes Médicas.
POZO, J. I. (1998) Teorias Cognitivas da Aprendizagem, 3d ed., Porto Alegre, Artes Médicas.