The emerging language of diplomacy

Authors: Jovan Kurbalija (content analysis) and Goran Milovanovic (statistical analysis)

Here, you can find an early analysis of the emerging language of internet governance from 2004. In 2023, as the UN Digital Global Compact is negotiated, this analysis can help us detect continuity and changes in digital governance’s language, framing, and narratives.

Introduction

The preparatory activities for the World Summit on Information Society, together with other multilateral initiatives in the ICT field (e.g. UN ICT Task Force, Dot Force, Global Knowledge Partnership), initiated the process of development of a specialised language of ICT diplomacy. This language has been emerging through the interplay among different institutional and professional cultures (traditional diplomacy, information and telecommunication sector, civil society, business circles, etc.). The main aim of our research project is to: (a) identify the main features of the ICT diplomatic language (b) identify linguistic patterns (c) analyse communication among different professional cultures. We are currently focusing on comparative analysis of the five reports from the WSIS regional preparatory conferences (Africa, Europe, Asia-Pacific, Americas and Western Asia).

Through this analysis we try to identify emergining language patterns and identify cultural differences among various world regions approaching the same issue. Beside various quantitative analysis declarations are analysed through the use of our ICT methodology which consists of five (governance & standardisation basket, legal basket, development basket, commercial basket, socio- cultural basket).

Here are the links towards complete texts of five analyzed declarations:

• The Bucharest Pan-European Conference in Preparation of the World Summit on the Information Society: Towards An Information Society: Principles, Strategy And Priorities for Action (9. November 2002)

• Bavaro Declaration – ECLAC Economic Commission for Latin America and the Caribbean Regional Preparatory Ministerial Conference of Latin America and the Caribbean for the World Summit on the Information Society – Bavaro, Punta Cana, Dominican Republic (29 – 31 January 2003)

• Report of the Africa Regional Conference for WSIS – Bamako (28 – 30 May 2002)

• Resolution of the Bishkek-Moscow Conference on the Information Society (Second stage) – Moscow (23-24 October 2002)

• Asia-Pacifik Regional Conference – the Tokyo Declaration – The Asia-Pacifc Perspective to WSIS

Document length and word/paragraph ratio

In this section, we present the results of the statistical analysis of the basic features of five declarations on Information Society development. We are interested in the length of the documents, number of words and paragraphs , etc. These kind of analysis may seem obvious and non-interestening at the first sight: to the contrary, we will argue that even the basic features of diplomatic texts formulated in different world regions uncover some significant differences in the way that the concept of Information Society is understood. Language is a complex system with many levels of analysis to think of, and in this section we present what could be termed as a “surface scratch” of the language analyses. Later, we will go deeper in the structure of the declarations.

First of all, we present quantitative information on the length of these different documents on I-Society development.

	Africa	Europe	Asia	Americas	W. Asia
Words	2675	1993	2993	3517	3810
Paragraphs	94	27	70	46	47
Characters	17485	13916	10764	24532	23148

Among many different aspects of the texts of these four declarations, the first we are about to present is the ratio between the number of words and number of paragraphs per declaration. This ratio tells as about the average number of words per paragraph and is interesting because of the difference which appears between the declaration texts formulated in different cultures.

	Words	Paragraphs	Words/Paragraphs
Africa	2675	94	28.46
Europe	1993	27	73.81
Asia	2993	70	42.76
Americas	3517	46	76.46
W. Asia	3810	47	81.06

Probably, this result has its roots in culturally determined linguistic habits in formal document writing. We can show that there is no systematic relationship between the length of the document and the number of words per paragraph.

Obviously, there is no systematic relationship between the two; the value of the Pearson’ s R correlation coefficient is 0.33 and not statistically significant.

At the moment, It is hard to guess what particular factors influence the words per paragraphs ratio. It is clear from the depicted relationships that both of the declarations formulated in the “western world” (Europe and America) have a large number of words per paragraph, but the Western Asian declaration also do. On the other hand, the African and Asian declaration show significantly lower words per paragraph value. Along with the development of the WSIS preparatory activities, the sample of declarations and documents will grow, and more elaborate analyses will be possible. Since the words per paragraph ratio is not systematically related to the document length, we state again our assumption that it is related with cultural factors in the first place.

Key concepts

Now we will examine the content of the declarations. We counted the frequency of the occurrence of the key concepts and keywords in I-Society development in all five documents. First, we will present this data, and then present the multivariate analyses based on the observed correlations of occurrence of key concepts and keywords in the documents, followed by the appropriate interpretation.

Seven key concepts in I-Society development where chosen for the analysis: Information Society, Civil Society, Digital Divide, Human Rights, Capacity Building, Sustainable Development and Private Sector.

We present the frequency distributions of the key concepts occurrences in all five documents.

	Africa	Europe	Asia	Americas	W. Asia
Information Society	22	20	16	37	13
Civil Society	29	3	3	4	4
Digital Divide	10	2	1	4	2
Human Rights	0	0	0	4	1
Capacity Building	0	0	1	2	0
Sustainable Development	0	2	0	0	0
Private Sector	12	9	5	3	9

The frequency of key concepts occurences in five analyzed documents is now used as the central information in the description of their content. The graph above represents these frequency distributions by using a different colour for each document. The coloured profiles at the graph can be understood as a representations of each document according to the frequency distributions of key concepts in their content.

We will now present the results obtained from multivariate analyses of the correlation data based on the occurrences of the key concepts in these five different documents. We describe the procedure in details:

• Two correlation matrices were calculated. The first correlation matrix contains the correlations between the frequencies of key concepts occurrences in the same declaration (the correlations between columns in Table 3.). The second correlation matrix is based on the same data from Table 3, but contains the correlations between the frequencies of key concepts occurrences across different documents (the correlations between rows in Table 3.).
• These correlation matrices will present the input for statistical analyses by means of multivariate statistical methods. These correlations can be thought of as a meassure of similarity among the objects of the analyses. Take for an example the declarations correlation matrix (Matrix 1, shown bellow). What does the numbers tell? The exact interpreation is the following: the more similar two key concepts distributions are, the higher the apsolute value of the correlation in the declarations matrix. The statistic which is used to express the correlation among variables is Pearson’s R coefficient of linear correlation (Pearson’s R can not be directly interpreted as a measure of similarity and in some cases must be proprely transformed prior to multivariate analyses).
• Exactly the same logic can be applied if instead of calculating the correlations among the distributions of key concepts per declaration one decides to caculate the correlations among the distributions of declarations per key concept. These correlations are presented in Matrix 2. The objects of analysis are now different: in Matrix 1, the objects of the analysis are documents (declarations), while in Matrix 2 the objects are key concepts.
• Hereby we present both correlation matrices:

	Africa	Europe	Asia	Americas	W. Asia
Africa	1	0.56	0.60	0.49	0.65
Europe	0.56	1	0.98	0.90	0.96
Asia	0.60	0.98	1	0.95	0.93
Americas	0.49	0.90	0.95	1	0.80
W. Asia	0.65	0.96	0.93	0.80	1

	I-Society	Civil Society	Digital Divide	Human Rights	Capacity Building	Sustainable Development	Private Sector
I-Society	1	0.04	0.29	0.82	0.76	-0.10	-0.52
Civil Society	0.04	1	0.96	-0.29	-0.36	-0.27	0.68
Digital Divide	0.29	0.96	1	-0.04	-0.18	-0.28	0.55
Human Rights	0.82	-0.29	-0.04	1	0.81	-0.32	-0.69
Capacity Building	0.76	-0.36	-0.18	0.80	1	-0.38	-0.92
Sustainable Development	-0.10	-0.27	-0.28	-0.32	-0.38	1	0.22
Private Sector	-0.52	0.68	0.55	-0.68	-0.92	0.22	1

• The multivariate methods of data analysis – multidimensional scaling and cluster analysis – were performed in order to determine the latent data structures underlying these correlation matrices. The results of both analyses will enable us to properly classify the objects of the analyses – e.g. declarations and concepts. Multidimensional scaling and cluster analysis were chosen for two reasons: (a) clear and understandable data representations can be obtained from these analyses, and (b) the sample was too small to perform a valid principal components analysis of the correlation matrices.

Results of analysis

First we will present the result obtained from multivariate analyses of the correlation matrices described above, than the interpretation of these results. Hereby we present a 3D conceptual space obtained from the multidimensional scaling of the properly (1-r) transformed declarations correlation matrix (Matrix 1).

Now we present a joining-tree diagram obtained from cluster analysis of the properly (1-r) transformed declarations correlation matrix (Matrix 1).

Interpretation: The cluster analysis of the “declarations matrix” produced the joining-tree diagram shown above. The groupings of the declarations at the diagram can be compared to the distances among them in the conceptual space generated by the procedure of multidimensional scaling. The African declaration is isolated both at the joining-tree and in the conceptual space, while the most similar among the five documents are the European and the Asian document. From the profiles of key concepts distributions shown above, it is clear that the African document is idiosyncrtaic to some extent. It’s main distinctive characteristic is the frequent usage of the concepts of civil society, digital divide and private sector. The Americas declaration is also separeted in the conceptual space. In the joining-tree diagram, it is sub-categorized under the same branch as West Asian, Asian and European documents, but at the higher level of linking. When we take a look at the distirbution of key concepts in this declaration we found that it is essentially similar to those declarations, but uses the concept of Information Society more frequently than any other document in the analysis.

The same analyses were performed on the properly transformed correlations from the concepts matrix (Matrix 2). We present the 3D conceptual map produced by the multidimensional scaling procedure, and a joining-tree diagram obtained from cluster analysis.

Interpretation: Both the inspection of the conceptual space and of the joining-tree diagram reveal two obvious groupings: (a) the concepts of I-Society, Human Rights and Capacity Building, on one side, and (b) the concepts of Civil Society, Digital Divide, Private Sector nad Sustainable Development, on the other. Sustainable Development seems to be less connected to the concepts of Civil Society, Digital Divide and Private Sector than these three concepts appear to be linked among each other. The concept of Sustainable Development is isolated because it appears in only one document – the European declaration. The concepts of Civil Society, Digital Divide and Private Sector are those concepts for which we already know that are frequntly used in the African declaration. The frequency profiles of the concepts of I-Society, Human Rights and Capacity Building seem to be systematically related accross these documents.

Comments regarding multivariate analyses

The interpretation of the 3D conceptual spaces generated by the multidimensional scaling procedure is dependent on the meaning of the data entered in the analysis. Let’s us remind that our data are the correlations between the frequencies of key concepts occurrences in the texts of five declarations on the I-Society development. The closer two points representing the declarations in the space are, the stronger the tendency that the similar distribution of frequencies of the key concepts exists in both documents.

In the case of concepts matrix, the conceptual space contains the I-Society development concepts, not declarations. The objects of the analysis are different, and the interpretation changes accordingly: the closer two points representing the concepts stand, the stronger the tendencies that the concepts they represent appear together in the same documents.

The goal of the cluster analysis is to “search” the representational space of some objects of analysis in order to determine the underlying data structure relying on the distances between the objects in the search space. One can think of the cluster analysis as a mean for the optimal “slicing” of conceptual spaces that we obtain from procedures such is multidimensional scaling. However, cluster analysis performs its search for structure in higher-dimensional spaces than those which result from the multidimensional scaling procedures (the later analysis also starts in a higher-dimensional space, but its goal is to reduce it to a lower-dimensional one, which is then treated as a resulting solution and is prone to interpretation). The interpretation of the joining-tree diagrams is straightforward: they depict the optimal categorization of the analyzed objects (declarations or concepts). It is also directly related to the multidimensional scaling in a following manner: the closer the points are in a conceptual space, the more probably will the objects represented by them be found under the same branch of the hierarchical joining-tree.

Keywords analysis

Beside the key concepts there are certain terms that make considerable language impact. We selected mixture of terms with different purposes: term “internet” was chosen to indicate link between information society and the key dynamical element of this society – Internet. Terms hardware, infrastructure and software were chosen to show level of technical language in five declarations. We also included three prefixes/adjectives which are usually used to describe modern ICT-developments: “e”, cyber and virtual. The analysis of these five documents show that prefix “e-“ has become dominant after its introduction in the Bucharest declaration. “Cyber” and “virtual” are almost non-existing. It is interesting to notice that term “cyber” is used in only international treaty dedicated to the ICT issue – Council of Europe Convention on Cybercrime.

Hereby we present the frequency distributions of keywords occurrences:

	Africa	Europe	Asia	Americas	W. Asia
Internet	1	0	3	1	14
should	37	22	14	12	49
hardware	1	0	1	0	0
infrastructure	8	3	7	5	9
software	4	0	2	4	1
e-	1	18	11	14	32
cyber	1	2	3	3	0
virtual	0	0	0	0	1

As in the previous analysis of key concepts frequency distributions, we see that the profile of the African declaration is somewhat different than the other declarations’ profiles. These difference will be noted following the multivariate analysis. Again, two correlation matrices were calculated, the declarations matrix and the keywords matrix:

	Africa	Europe	Asia	Americas	W. Asia
Africa	1	0.71	0.74	0.55	0.77
Europe	0.71	1	0.95	0.94	0.95
Asia	0.74	0.95	1	0.93	0.94
Americas	0.55	0.94	0.94	1	0.86
W. Asia	0.77	0.95	0.94	0.86	1

	Internet	Should	Hardware	Infrastructure	Software	e-	Cyber	Virtual
Internet	1	0.73	-0.28	0.69	-0.33	0.79	-0.70	0.98
Should	0.73	1	-0.07	0.68	-0.19	0.40	-0.99	0.78
Hardware	-0.28	-0.07	1	0.42	0.41	-0.74	0.14	-0.41
Infrastructure	0.69	0.68	0.42	1	0.27	0.10	-0.61	0.60
Software	-0.33	-0.19	0.41	0.27	1	-0.66	0.24	-0.38
e-	0.79	0.40	-0.74	0.10	-0.66	1	-0.42	0.83
Cyber	-0.70	-0.99	0.14	-0.61	0.24	-0.42	1	-0.77
Virtual	0.98	0.78	-0.41	0.60	-0.38	0.83	-0.77	1

Results of analysis

Hereby we present a 3D conceptual space obtained from the multidimensional scaling of the properly (1-r) transformed declarations correlation matrix (Matrix 3).

Now we present a joining-tree diagram obtained from cluster analysis of the properly (1-r) transformed declarations correlation matrix (Matrix 3).

Interpretation: the results of the multivariate analyses of the keywords declarations matrix are consisent with the prior analysis performed on key concepts. The African declaration is isolated from others since it is characterized with the very low frequency of the “e-” prefix when compared to other declarations.

The same analyses were performed on the properly transformed correlations from the words matrix (Matrix 4). We present the 3D conceptual map produced by the multidimensional scaling procedure, and a joining-tree diagram obtained from cluster analysis.

Interpretation: Two major groupings of key words occure following the multivariate analysis of the keywords matrix: (a) a group containing the keywords Cyber, Hardware and Software, and (b) a group containg other keywords. The keywords Internet and Virtual are highly connected since the only document containing the keyword Virtual is at the same time the one with the most frequent usage of the keyword Internet (W. Asian declaration).

Content analysis: Analysis of semantic patterns

The content analysis we present here is specific in a way. We used a text-analysis software tool, designed to analyze texts and represent its content through five master variables (semantic factors, so to say): Activity, Optimism, Certainty, Realism and Commonality. These master variables are calculated as linear combinations of a larger number of variables, each calculated on a lexical basis of typical words occurrences. This analysis is performed in order to identify prevailing rhetoric of the declarations. Research in the field of diplomatic language indicates high correlation between existence of certain semantic patterns and effectiveness of the document. For example Vienna Convention on Diplomatic Relations (1961) which is considered to be one of the most applied international treaty shows high level of realism and commonality. In the early stage of the development of the international regime, as it is the case of the WSIS-process, it is expectable to have the current level of distribution of semantic patterns (more optimism – less realism).

The text analysis software used to perform this analysis utilizes a sort of a rather typical discourse processing, based on frequency counts of typical words and subsequent categorization based on previously established semantic criteria. This method utilizes a number of different dictionaries generated from a 20,000 texts sample as a normative basis for its content analysis. The main idea of this sort of discourse processing is to analyze the occurrence of words semantically related to the wider, descriptive categories.

Categories	Africa	Europe	Asia-Pacific	Americas	W. Asia
Activity (language indicating resoluteness, inflexibility and completeness)	51.83	46.2	47.97	48.31	49.68
Optimism (language endorsing some concept, person, group or event and highlighting their positive entailments)	51.25	54.77	51.74	50.88	52.9
Certainty (language indicating resoluteness, inflexibility and completeness)	46.04	50.54	49.56	42.59	51.71
Realism (language describing tangible, immediate, recognizable matters that affect everyday life)	44.7	41.91	42.29	42.66	42.67
Commonality (language highlighting the agreed-upon values of a group and rejecting idiosyncratic modes of engagement)	52.06	49.54	50.79	51.43	50.12