The social construction of shared concepts

empirical study of a distributed cognitive process.

1. Abstract: • Recently, different disciplines have come to regard cognition as a

social phenomenon,distributed over a group of individuals. Sociologists have noted long ago that knowledge is a social construction (Berger & Luckman, 1967), however without proposing an explicit model of this process.

• In this project we intend to investigate empirically different socio-cognitive processes through which shared concepts are constructed out of individual concepts. We shall focus in particular on the factors that influence this processes and on the manner in which the resulting "consensual" concepts differ from the initial individual concepts. That, we hope, will give us a better insight into the mechanisms guiding distributed knowledge development, which in turn will allow us to increase the efficiency and reliability of that process.

2. Introduction

• Constructivism: reality of constructs:reality is independent of human thought but the meaning or knowledge is always a human construction.– Giovanni Battista Vico (XVII)

“Verum est ipse factum”: the true is precisely what is made.– Lev Vygotski, Jean Piaget

social constructivism, learning theory, ZPD.

– Heinz von Foerster, Ernst von Glaserfeldradical constructivism

– Paul Watzlawick“How do we know what we believe we know?”

– George Kellygroup expectancies as validators for personal constructs

– Herbert Simon, Ed Hutchinssciences of the artificial: cybernetics, cognitivism, disrtibuted cognition

– ...

3. Aim

empirical measurement and comparison of: different factors of social interaction (interface)

-online-live discussion-face to face carrousel

different concepts-fruit & vetgetables-sport-happiness

via different measures-responsvariability in groups-conditional entropy (consensus/opposition)-interraterreliability (consistency of

responsepatterns between individuals in the group, Cronbach’s alpha, correlation measures)

4. Research hypotheses

• Operationalisation of concept: the distributed,“external” approach (Heylighen, 1999 Van Overwalle et al. ,2003),i.e. “a process of categorisation, whereby phenomena are classified as instances of the concept to a greater or lesser degree”. A concept can thus be represented as a vector of which the components correspond to the earlier mentioned categorization strenghts.

• Concepts are abstractions of recurrent aspects of reality. Each individual experiences the world from its own perspective, therefore those abstractions will differ also, in the sense that the categories will not completely overlap.

4. Research hypotheses

• In order to communicate effectively, different individuals must use the same categories.

• An interacting group will undergo a proces of self-organisation (Steels, 1998, Bonabeau et al.,1998), whereby out of local interactions a global, more coherent pattern emerges. This implies that the divergence in categorisation among individuals will diminisch, ideally leading to a single concept.

• This shared concept will be “fitter” than the initial concepts, as it will integrate the diversity of subjective experiences in a broader, intersubjective whole.

5. Operationalisation

• For each subject, the individual concept is represented by a vector. The comparison of the different vectors gives us an objective measure for the spread or diversity in the viewpoints. The average of the vectors defines the “collective” concept for the group (Heylighen, 1999).

• After the subjects have interacted, individual and averaged concepts are measured again.

• Expectancies:1) The spread among the subjects will diminisch.2) The collective concept will be consolidated:

agreed vector components strenghtened, disagreed vector components weakened. (consensus)3) Possible polarisation in case of strong divergence, splitting the vectors in different clusters that could be seen as alternative interpretations of the concept.

6. Empirical approach• In our basic set-up, a small group (about 10) of

experimental subjects are requested to discuss a given concept, with the objective of achieving a shared understanding. The concept is chosen such that everyone has some experience with it, but there remains sufficient vagueness or ambiguity to allow different interpretations.

• To minimize the risk for emotional arguments or political games, the concepts were selected to be as neutral as possible (“fruit”, “sport”, “happiness”), and the participants are told explicitly that there won’t be any “winners” or “losers”.

6. Empirical approach

• The subjects are informed about the subject before the experiment, so that they can prepare their thoughts without mutually influencing each other. They are asked in particular to suggest a number of examples, counterexamples and intermediate cases of the category. We select the most representative ones of those, and submit the resulting list of twenty items to all subjects. We ask them to score each one on a 7-point scale, indicating the degree to which they consider it to belong to the category. This produces the initial concept vectors for all participants

7. Written version• In the written version of the experiment, the interaction

takes place asynchronously,using an electronic discussion system developed by F. Heylighen. (Each participant starts with a short description of what the concept means for him or her, and then is allowed to reply to the interpretations of others, using examples, arguments and counterarguments. After a period long enough to allow each subject to intervene several times, the discussion is stopped, and the concept vectors are measured again. The statistical comparison of initial and final vectors provides us with a quantitative analysis of the evolution of the concept. A textual analysis of the different interventions provides us with a more qualitative picture of the arguments and factors that have influenced the outcome. The possible reasons why a particular participant has or has not changed positions are explored by focused interviews.

8. Oral version

• The oral version of the experiment is similar, except that the group of participants now discuss face-to-face during a two hour session without facilitation. Concept vectors are again measured before and after the session. The discussion is recorded on videotape, and afterwards analysed for specific factors that appear to have influenced the outcome. Immediately after the session, selected participants are interviewed in order to explore their unstated reasons for changing their perspective.

• The results of the ‘happiness’-goup are correlated with an averaged expert-score before and after the interaction. A significantly increase of thecorrelation, confirms the hypothesis (fitter concept)

9. Results & Discussion

• In our two forum groups (written version) and the two live- interaction groups about the ‘sport’-concept, we don’t see any of our hypotheses confirmed. Only in the carrousel-group wee see a slight significant (sign.088) diminishment of the spread. In the forum-groups the spread did not change significantly. In the live discussion group we measured an even significant raise of the spread, but our qualitative analysis revealed an orally negociated way of quotation, what led to a certain polarisation. We also could measure this trend as a diminishing entropy in this interaction. If we take a closer look nevertheless, we encounter a very high initial value for Cronbach’s alpha (>.9) for all the four groups, which suggests a firm consistency from the beginning. The concept was already gridded.

9. Results & Discussion

• In our Happiness-groups on the contrary, all our expectations were confirmed. There was a significant fall of the spread in the two cases(.09) for the discussion and (sign .341) for the face-to-face carrousel.

• The comparison of the correlations between the scores of de students and the experts revealed a raise for either of the conditions, what confirms our hypothesis.