Formal Requirements for Using Network Metrics in Representing Cognitive Structures
|
|
Walter
Bien - DJI Formal Requirements for Using Network Metrics in Representing Cognitive Structures |
The representation of a network with paths build by a combination of links with an additive metric needs some formal requirements on the used data. Such requirements are e.g. that data are reflexive, symmetric, transitive, complete and for some content reasons balanced.
We proved this requirements for cognitive representations of different relations between persons acting in small groups. The result show highly structured data in relation to the proved requirements, but also an extremely high number of violations against the requirements.
|
|
Stephen
P. Borgatti - Jeffrey C. Johnson - Boston College Some Issues in Comparing Aggregate Cognitive Networks |
Given a set of cognitive networks such as collected by Krackhardt or Kumbasar, a natural analysis to attempt is a comparison of the aggregate networks obtained from two groups of respondents, such as men and women. A typical approach to this problem uses the non-parametric, permutation-based QAP technique to correlate the two aggregate sociomatrices. However, the null distribution that QAP is based on may not be appropriate for comparing subsamples of a dataset. QAP compares the correlations between subsamples with correlations among a set of matrices that include many that are quite different from those could be obtained by aggregating subsamples. Consequently, the QAP approach reaches significance too often in this context. An alternative method is proposed here that compares the observed correlation with correlations among matrices that are of the same kind as the observed - they are obtained by aggregating data from random samples of the dataset. The new method yields results that accord better with researchers' intuition. In addition, the new method can be applied to a variety of data types beyond those appropriate for QAP.
|
|
Tiziana Casciaro - Harvard
Business School Dreadful Ties: The Structure and Perception of Negative Interpersonal Networks |
Social network researchers tend to focus on the positive side of interpersonal relationships (e.g., friendship, advice, and trust). Negative relationships are acknowledged at times, but rarely measured because study participants are reluctant to answer questions about the disagreeable or malevolent side of their social lives. For this reason, we know little of how networks of positive and negative relationships differ in their structure, if at all. Nor can we say how transparent, or concealed, this structure is to the observer's eye. This study uses cognitive social structure data on the patterns of like and dislike among 53 students at an American university to demonstrate asymmetries between the way positive and negative relationships are built and perceived. Implications for the significance of positive and negative affect in the study of social networks are drawn.
|
|
Peter Erdi - Dept. Biophysics,
KFKI Res. Inst for Particle and Nuclear Physics of the Hung.
Acad. Sci. Cognitive and Social Systems as Complex Interactive Networks |
Real world systems in many cases can be represented by networks.
Networks can be seen everywhere (neural networks of the brain,
food webs and ecosystems, electric power networks, system of social
connections, global
financial network, the world-wide web).
Since the famous social psychological experiment of Stanley
Milgram, it is known that from certian point of view we live in
a 'small world'. However, the relationships between the structure
of large networks and their dynamical properties generally are
not well known. The performance of many biological, ecological,
economical, sociological, communication and other
networks can be illuminated by using new approaches coming from
graph theory, statistical physics and nonlinear dynamics.
The relationship between the topological structures of neural and mental structures are explained. Furhter examples will be given to illustrate the power of the new approaches in the understanding of the organization of cognitive and social structures.
|
|
Joel Marti Olive - Department
of Sociology, Universitat Autonoma de Barcelona An Argumentative Approach to Discouse: "Topos", Discursive Structures and Social Network Analysis |
This paper pays special attention on methodological and analytic issues related to argumentation and discourse, and on Social Network Analysis as a way for its study.
From a theoretical point of view, argumentation is an area of inquiry which can be approached from a variety of perspectives and techniques. Anscombre and Ducrot (L'argumentation dans la langue, 1983) define an argument as a relation between a premiss "p" and a conclusion "c", relation which can be standarized in a linguistic-cognitive rule called "topos".
From this point of view, an analytic proposal is developed. The procedure is based on the successive identification of argumentative relations in the text. As a result, discourses are not segmented into preestablished themes or categories, but globally represented in networks of "topos" which keep its unity. These cognitive networks are analyzed and represented by social network analysis.
|
|
Joseph Perczel - International
Council of Psychologists Inc. The Dienes-Phenomenon. A Cognitive Switchboard between Inter-scholarly Networks |
The Dienes phenomenon (DPH) has been described the first time at the Second International Conference on Work Values (Prague, 1992, Proceedings, Pp. 35-42), as a radical, innovative approach to the interpretation of experimental data. The application of the method turned out to become an able instrument of interdisciplinary transformations, a Gestalt-switchboard using cognitive network operations - so to say. The basic innovation of DPH is the point that the way of presentation of data already determines their immanent disciplinary character and changes in presentations are operating switches between scholarly networks. Accordingly data have two different relations among themselves: permanent identities, differences (différence, külön-bség) and potential ones (différance, külömb-ség) varying according to disciplinary lines.
|
|
Roel Popping - Groningen University,
Department of Sociology Knowledge Graphs and Network Text Analysis |
Sociological knowledge is available at many different places.
It is desirable that empirically tested knowledge is available
at one place. The many studies containing this knowledge contribute
a little to theorising in a specific field, but do not hold a
complete overview. Network structures can be used to get such
a complete overview.
Networks used for this representation are called knowledge graphs.
This paper presents the state of art with respect to this kind
of graphs and next concentrates on two issues. The first one is
the determination of concepts (points) to be used in the graphs.
Usually they are too broad; they are abstract concepts build up
of several other concepts. The other issue concerns the use of
conditional statements. Some findings hold only for a specific
group, say only boys or girls. This has implications for the representation
of the theory. The examples that are used come from the labour
market theory.
|
|
Keith Rennolls - Universtiy
of Greenwich Attempting to Analyse Biased Samples of Subjective Views |
In a social group each member of the Group may have a view of any other member of the group. This view may be expressed by a value on an appropriate scale which may depend on the "true underlying measures" of both the viewed and viewer. The way in which the view-value and the true-value differ will depend on how the viewer responds to the relative and absolute perceived positions of ego and other. Some actors will be "attractors" in that they give other actors a view-score closer to their own peceived s
Data may be collected on the views of a sample (possibly 100%) of the individuals in the group. The first question to be considered is the esimation of the "true-measure parameters", for different sample structures and for different models of behaviour structure in the group.
It is then possible consider asking a particular actor (their
second order view of) how they would expect the other actors to
view each other. Does this second order view contain new information
which may be used to improve the efficiency in the estimation
of the group parameters? If so then it may be possible to select
just a few strategically placed (i.e.biased) viewers for first
an second order view measurement in order to characterise the
underlying parameters of group members, as well as their attit
Simple models and fitting algorithms for this "relativistic"
model of group attitudinal structure will be presented on test
datasets.
|
|
Joan Miquel Verd - Universitat
Autonoma of Barcelona, Department of Sociology A Proposal of Three-dimensional Coding of Texts Using Cognitive Networks. An Application to Narrative Interviews |
Nowadays cognitive networks are applied to very different fields.
Besides the uses in Linguistics, Psychology or Artificial Intelligence
as a system of knowledge representation it is possible to use
them as a valuable help in qualitative text analysis. This paper
presents an analysis procedure that combines cognitive networks
and causal networks in order to obtain an objective, explicit
and formalized structure of discourse. The objective is to reconstruct
the discourse of individuals by attaching codes to segments of
text and connecting them. This structure develops in two different
planes. In the first one, a hierarchical relation between codes
is stablished, going from codes with a high level of abstraction
to those with a lower level of abstraction. This procedure is
applied to semistructured interviews in which the relation between
training and employment is posed. The representation of the networks
by means of graphs and the simplification based on the hierarchies
of concepts prove to be very useful in understanding and explaining
the training and labour trajectories described by individuals,
as well as to compare the description of different trajectories.
Introduction
Since the Uppsala and Innovation models were developed, numerous scholars have advanced various criticisms. Some have said the models are too deterministic. It is likely that many firms rely instead on careful strategy making, which considers a number of more ore less abstract success factors. By using these factors the decision to go international may be carefully mapped out in a planning process. Thus more emphasis should be given to the causal maps managers develop.
Methodological Foundations of the Study
The basic elements of cognitive maps are elicited concepts and their relations. Research, which applies cognitive mapping, utilizes a number of distinct elicitation methods. Critical design issues of mapping procedures have to cover saliency, comparability, validity, and atomism vs. holism, reliability, and practicability.
Sampling and analysis
The sampling procedure consists of a two step process. The students (n=162) of three university post-experience courses for managers in export and international marketing took part. A standardized elicitation process did the elicitation to every student. In a second step general information on the respondent s firm has been collected.
The structural analysis of the cognitive maps is carried out by the means of network analysis for managerial decision making. Theoretical as well as practical conclusions are drawn from the results.
