Visualizing change in political influence networks to support journalism

(1)

Visualizing change in political inﬂuence

networks to support journalism

SUBMITTED IN PARTIAL FULLFILLMENT FOR THE DEGREE OF MASTER OF SCIENCE

Marieke van Kouwen

10822763

MASTER INFORMATION STUDIES

Human Centered Multimedia

FACULTY OF SCIENCE

UNIVERSITY OF AMSTERDAM

July 30, 2016

1st_Supervisor Dhr. Prof. Dr. M. Worring University of Amsterdam 2nd_Supervisor Dhr. Dr. M.J. Marx University of Amsterdam

(2)

1. INTRODUCTION

Journalists have a supervisory role in modern society, often referred to as ’The Fourth Branch of Government’. Their task is to provide citizens with a glimpse into the political decision making process, which is often influenced by organizations and associated lobbyists. More transparent governmental information can potentially build trust, strengthen responsibility, and increase citizen satisfaction [13]. Open data offers journalists a new, unique opportunity to obtain and substantiate stories in order to provide citizens with up-to-date information on political influence networks. The aim of this project is to develop a research tool that supports journalists in visually exploring open data files describing the network of politicians, organizations, and the relations between them. The analysis and exploration of networks in general is usually done with the aid of node-link diagrams (graphs). These traditional network visualizations consist of nodes, linked by edges. Generally, data visualizations form a powerful tool to assist us in summarizing and reasoning

about data. They amplify our cognition by enhancing our working memory, help us easily recognize patterns and irregularities, allow data reading and exploration, reduce search time, guide statistical analysis; check validity, and help in formulating new hypotheses [23]. When depicting political influence networks we encounter two major problems using the traditional node-link diagrams. Firstly, due to clutter of the network elements, it becomes a challenge to discover insights in these extensive complex networks. Secondly, the evolution of the network is essential for journalists to interpret historic and current political dynamics, though invisible in the node-link diagram. These reasons lead us to conclude that other visualization techniques than the traditional node-link diagram are needed to depict political influence networks. To find a solution better suitable to the needs of the data-journalist, we need to address the two challenges described above: complexity and temporal context. Firstly, our visualization needs to deal with the problem of the size and complexity of the network, which causes clutter of network elements. The common solution addressing this problem is filtering and aggregation: the clustering of nodes. While these solutions reduce clutter, they obscure information as well [20]. In the case of researching political influence networks, even the smallest amount of information can be of significant

Visualizing change in political inﬂuence

networks to support journalism

Marieke van Kouwen

University of Amsterdam

Abstract — Open data offers journalists a new possibility to investigate governmental information. However, even with this data available, information is hard to find, explore, and interpret. To support journalists in researching the changes in political influence networks, we developed Clique: an interactive tool, that uses three data visualizations to show the evolution of political egocentric networks. The first visualization allows the user to select a period during a politicians career in an interactive timeline, which makes use of the Triangular Model. Subsequently, the egocentric network changes during the selected period are displayed in a second visualization using the difference map technique, clustering them into five types of changes. By selecting an actor in the difference map, a third visualization is generated which explores the shared relationship between the two selected actors in the network. We found that Clique is especially considered useful during the initial phase of journalistic research, and supports journalists in making unexpected connections while exploring the network.

Index Terms — Data visualization, Dynamic networks, Open data, Big data, Egocentric abstraction

First reader: M. Worring, University of Amsterdam Second reader: M. Marx, University of Amsterdam In cooperation with Waag Society, Amsterdam Manuscript received July 30, 2016

(3)

2. RELATED WORK

Most examples of dynamic network visualization concerning political connections disregard network evolution completely and use the traditional node-link diagram. An example is the website LittleSis [15] in the U.S., “a free database of who-knows-who at the heights of business and government”. The database combines data about members of Congress, political contributions, corporate boards, lobbying data, and government contract data. The published visualizations on the platform are mostly static node-link diagrams and only used for illustrative purposes and storytelling, instead of allowing data exploring and analysis, as Clique intends to support. A more interactive example is the tool Lynksoft [8], an online network visualization tool. Lynksoft is currently used to visualize the same political influence network data source as we use for this thesis. It is a powerful tool, since users are given the possibility to filter and remove nodes, though it still does not address the problems of the node-link diagram: complexity and temporal context.

2.1 Visualizing egocentric networks

Shi et al. [21] developed the 1.5D Egocentric Dynamic Network Visualization, which focuses on the egocentric dynamic network, and encodes time vertically in one dimension. This approach smoothly integrates multivariate data next to the dynamic network structure by including trend glyphs to reveal interesting patterns. Nodes are connected to moments on the timeline and to each other. The colors and thickness of the edges reveal information on the node attributes. Although the timeline gives the user a clear overview of the network evolution, the display of complex networks makes the visualization hard to interpret due to clutter of network elements.

EgoNetCloud adds the event-based scenario to the 1.5D technique [21]. This event-based egocentric dynamic network visualization [16] displays events, i.e. temporal network dynamics as well as the relationship between the ego node and the first degree nodes. Events are time points when an edge appears, i.e. the publication of a paper, a call, or befriending someone on social media. Although it is nice to have all aspects in one visualization, the addition of these events to the displayed timeline makes the visualization even harder to interpret than the 1.5D method, especially when the number of nodes and edges increases.

2.2 Visualizing changes in dynamic network evolution

Graphdiaries as introduced by Bach et al. [6] is an

interesting visualization example that focuses on animated transitions, and interactively highlights changes in between time steps. The use of interaction importance. To diminish the complexity of the network

we need to search in another direction for a solution. Instead, this project explores a single politician’s history and associated circle of influence. The complexity is reduced without obscuring information, by generating a subset of the network that only targets an individual node, and its egocentric network: its neighbors and associated interconnections.

The second challenge is the visualization of temporal political influence network dynamics. The two major current visualization methods of network dynamics are time-to-time mappings (animation) and time-to-space mappings (small multiples1_{) [7, 24]. Both have their own} (dis)advantages. The animation approach [17] has been proven ineffective for network analysis compared to the small multiples technique as users need more time to understand the network dynamics, plus it is hard to track changes over longer periods of time [5, 9, 25]. Time-to-space mappings seem more promising for analytical purposes, although it is challenging to find a balance between a few visualizations that lack detail, and many visualizations that need large screens [21] and are hard to monitor simultaneously [24]. It appears that other visualization methods are appropriate when we combine the temporal aspect of dynamic networks with the analytic journalistic approach, and the extensive amount of links an egocentric network still contains. To track changes in the network over time, we propose to focus on the essential part, i.e. the change. Using only the changes further reduces the amount of links, which makes it easier to research extensive egocentric dynamic networks.

Summarized concisely: this project develops an innovative, interactive visual interface that makes it possible for journalists to research the extensive amount of open data on politicians and their circle of influence. Instead of visualizing the entire network, the focus is on the visualization of change in the egocentric network. This will emphasize the historical nature of the data and declutter large network visualizations. To put it in a research question: How to enable journalists to research political influence networks by visualizing change?

The name that is chosen for the interactive tool is ’Clique’, which means interconnected network in network theory as well as a secluded group of people who do not readily allow others to join them; a description often used for the group of governing politicians2_.

1. Small multiples uses a series of several graphs or charts with equal scale and axes to visualize different parts of the same dataset.

2. In Dutch the same word is used describing the same thing with equal pronunciation, though written differently: ’kliek’.

(4)

of links, and clear communication on the computational process. As for the functionality of Clique, the interviewees consider the exploration of an individual timeline an interesting approach because it allows them to research the career and relations of a person of interest.

3.2 Dataset

To develop our visualization approach we use a real-world dataset: Transparant Nederland (TNL) [26]. TNL combines diverse open data sources, which hold information on politicians, their resumes, and circle of influence.

The assembled data is published as an Application Program Interface (API). A detailed description of the data can be found in appendix A.

The TNL dataset is fulfilling almost all of the requirements of a good dataset as formulated by Few [10]. However, some remarks must be made. The data is clean to a certain extent, but is often incomplete. For example, a list of positions that lacks dates, or the use of various labels describing the same entity. Besides the incompleteness of the data, the sincerity of politicians is important here as well. A secondment is only visible when explicitly stated, otherwise the placement organization is displayed instead. In addition, the TNL data is regularly updated though not realtime, which can lead to outdated information. To understand the behaviour of the nodes in the visualization, we should first define the extent of an egocentric subset of the dataset. Egocentric networks always begin with the ego-node. Organizations constitute the nodes in the first degree. Their other employees during the same period, form the second degree. The third degree is constituted by organizations connected to the ego-nodes in the second degree. Since our visualization focuses on the egocentric network, it focuses only on the first and second degree of the subset. Of course, the third degree is still accessible by selecting another ego node, which transforms it to the first degree. A connection between two actors in the network is brought to existence when they share the same organization during the same period. This assumption should be carefully considered, as there are many reasons two actors may be completely unaware of their connection in reality, e.g. working in different geographical locations for the same organization. All nodes in the same degree are only connected through another degree, never directly to each other. The egocentric network structure therefore resembles a tree, starting from the ego and dilating to the egocentric social network.

3.3 User goals and tasks

Change is the transformation of one state to another, and only exists by virtue of time. Basic questions a journalist tries to answer regarding change are: who, what, where, when, why, and how. To answer all these questions with and animation to research changes in the network instead

of creating multiple static visualizations is an interesting approach, although not reducing the network complexity. An inspiring research concerning the visualization of dynamic network evolution is the ’reduction of snapshots to points’ method developed by van den Elzen et al. [24]. This technique involves reducing small multiples to a single point in high-dimensional space, mapping it to 2D, and showing the evolution between two points with the help of color. This technique easily identifies outliers and reoccurring states. A drawback of this visualization technique for Clique is its high abstraction level. However, this technique could potentially bring unexpected outliers to the attention quickly.

Archambault [3] presents the technique of a coarsened difference map3_{to visualize the differences between} two graphs (or in this case, two time steps in a dynamic network). The visualization combines the two graphs in a colour-coded difference map in which first degree nodes with equal difference values are coarsened to one big node. This technique takes into account node as well as edge changes, preserves the structure of the network, and reduces visual complexity simultaneously. This method has two drawbacks when deployed for Clique. Firstly, the loss of information on the exact timing of the change in the nodes. Secondly, although the structure of the network is preserved, this technique does not include spatial cues to map similar elements to similar parts of the graph, which could be confusing for the user.

3. CLIQUE: USER AND TASK DESCRIPTION

This section describes the approach of Clique in more detail. First the target group (paragraph 3.1) and the dataset (paragraph 3.2) will be introduced, followed by a discussion of the user goals and tasks (paragraph 3.3).

3.1 Target group: journalists

An emerging practice in the field of journalism is data-journalism. The new possibilities of mobile and digital media are the main motor force of this development [12]. However, there is no consensus on the definition of ’the data-journalist’, although all data-journalists believe that their colleagues should have some affinity with data, their data-related skills and educational background vary widely [11].

To discover the needs of the target group, we performed some exploratory interviews with several journalists. One of the most frequently mentioned issues is that Clique should address the clarity of the information: constant available clear referencing, information on the nature 3. A difference map is the combination of two input graphs in a single graph.

(5)

filter the information. Sorting (e.g. by time) and clustering is necessary to assist the user making sense of the data. Of course, details-on-demand, as described by

Shneiderman [22] in his Visual Information Seeking Mantra “Overview first, zoom and filter, then details-on-demand”, and clearly stated as a requirement by our target group, should be added to the list of user tasks as well.

3.4 Summary

Clique intends to facilitate the research of data-journalists on the evolution of political influence networks. In order to develop a visualization approach, a real-world dataset is used, TNL, which holds information on politicians, their resumes, and circle of influence. Journalistic queries on the evolution of change can be described in three dimensions: what (behaviour of the nodes), which (location in the network: which nodes), and when (timing). Each user task is a compound tasks combining two dimensions to derive the answer on the remaining one.

4. VISUALIZATION APPROACH

The three compound low-level tasks are difficult to capture in one comprehensible visualization, as we saw in the chapter on related work. The amount of data is extensive, and the temporal nature of the data makes it even harder to make a visualization, especially given the varying data-related skill-level of the target group. This project proposes an application that uses three different, successive, interactive visualizations, each focused on a different compound task (which, what, and when). These three

visualizations enable the user to explore the changes in the network by selecting parts of the data in one visualization and reviewing them in the next.

The change in itself occurs on a single moment in time, which implies the existence of ’static’ parts over longer periods of time. The change in the egocentric networks of politicians is associated with the positions they hold. Every position is connected to an organization, which glues together other egocentric networks. When positions change, the egocentric network changes: the associated organizations in the first degree, and the other linked egocentric networks in the second degree. To identify these changes, we need to relate the egocentric network to similarities in other egocentric networks.

The overall approach of Clique is visualized in figure 1. A user starts by selecting a person of interest. The selected egocentric subset is displayed in the egocentric timeline visualization (which). This visualization depicts change for the first degree nodes over time, by identifying static periods in which politicians are associated to organizations. The timeline allows the user to define a period, and to research the changes during that timeframe the TNL dataset would be quite ambitious since the data

only possesses factual information: the actor (who?), the event, in this case the position at a certain organization (what?), and the timing (when?). Peuquet [18] introduces a framework to describe the dimensions necessary to find the answer to spatio-temporal queries: space (where?), time (when?), and object (what?). She describes the answer to the question in one dimension as a derivative of values from the remaining two. In order to apply these dimensions to a dynamic network context, we use the redefined version of the framework as proposed by Bach et al. [6]. Temporal tasks are still described by time (when?); object (what?) captures the event, describing what happened: in this case the appearing and disappearing of nodes; space (where?) clarifies the location within the graph structure: which nodes? Which paths? Which attributes? As the location question (where?) transforms (which?), when we ask for the location of network elements, the latter is used to avoid confusion. The visualization of each user task as a derivative of the values from the two remaining dimensions allows us to characterize the user goals as compound low-level tasks. Along with their description we provide the high level implications for interactive visualizations as formulated by Yi et al. [27]:

• what & when -> which: To determine exactly which node(s) to show can be derived from the combination of the knowledge on the timing and the event.

In order to visualize the combination of timing and the changing of events, it is important that the visual element(s) concerning the events should be easily identified, filtered along search queries, and tracked over periods of time.

• when & which -> what: What happens to the nodes can be derived from the combination of knowledge on the nodes and the timing.

To comprehend the nature of the changing events in the visualization -- the appearing and disappearing of nodes -- it should be possible to explore the events, and easily locate the involved actors in the network.

• which & what -> when: Knowledge of the timing can be derived from combining knowledge on the nodes and the event.

To understand the timing of the changes, filtering, connecting and comparing various egocentric networks are needed, when multiple ego nodes are involved in events.

These low-level compound tasks have some generic low-level implications on the visualization level as well. We describe them along the taxonomy as described by Amar et al. [2]. For all queries a user needs to be able to retrieve the value of specific attributes in order to determine how to find ranges within the dataset, and

(6)

space. Aggregation is not an option in this case, as all intervals are equally significant. Instead, the Triangular Model (TM) [19] is applied, which is introduced by Kulpa [14] and presents time segments in a two dimensional space, as illustrated in figure 2. By transforming the time segment to the base of an isosceles triangle, the height (h) of the triangle represents the duration of the interval. Each interval is mapped to one point with an x and y coordinate: the top of the triangle (I). The use of TM gives us the possibility to display a politician’s career in one single scatter plot-like overview with a controllable height. The addition of interactivity guides the user through the visualization (as shown in appendices B.2, B.3 en B.4). Hovering over a coordinate highlights the corresponding period on the timeline and details of the data. Interactivity solves the problem of overlapping points, which obscures data. The points are loaded with a quick animation, combined with a number identifying the number of overlapping points. Clicking on a point enables the user to browse through the underlying points. Another feature in the second degree network (paragraph 4.1). The change

visualization (what?) identifies the changes in the first and second degree nodes during the period defined in the egocentric timeline visualization (paragraph 4.2). The change visualization can be dynamically adjusted by selecting other periods in the timeline visualization. By selecting another actor in the network, the relational visualization (when?) is triggered, which depicts the changes and similarities in the shared nodes of the two egocentric networks for the entire period (paragraph 4.3). Details-on-demand on positions/organizations and involved actors can be requested from all levels of the tool. By clicking a button in the relational visualization, the selected actor becomes the person of interest: the egocentric network is compiled and the process starts from the beginning.

4.1 What & when -> which: Egocentric timeline

approach

The egocentric timeline displays the changes in the first degree nodes (which), by combining knowledge on the time intervals (when) of the event, i.e. the held positions (what). This visualization gives the user the possibility to explore historical network data on a person of interest, and to filter the data to user-defined periods. To enable a quick review of the sectors involved, and intuitively highlight outliers, attribute information is added to the first degree nodes: a color indicates the sector categorization of the involved organizations.

Time intervals are usually displayed as linear segments with one dimension. Some models use an illustrative extra dimension without temporal meaning, e.g. time tables and Gantt charts. Visualizing a politician’s career by these methods generates an uncontrollable amount of lines, and could potentially take up excessive screen

𝛪𝛪⁻ 𝛪𝛪 𝑑𝑑𝑢𝑢𝑟𝑟�𝛪𝛪� 𝑚𝑚𝑖𝑖𝑑𝑑�𝛪𝛪� dur ation time (a) (b) time 0 𝛪𝛪⁺ 𝛪𝛪⁻ 𝛪𝛪⁺ 𝛼𝛼1 𝛼𝛼2 ℎ 𝑙𝑙

Fig. 2 (a) An interval point (I) using two dimensions, constructed with the Triangular Model technique. (b) The same interval as a Gantt chart in one dimension.

select ego select period T

abstract abstract select person adjust ego to T D A TA PR OCESSIN G USER INTERA CTION searchfield for egonode visualization: egocentric timeline visualization: network changes visualization: shared nodes

full network egocentric network

which? what? when?

egocentric network shares egocentric network changes time restricted egocentric network selected ego node two separate selected nodes 𝛪𝛪⁻ 𝛪𝛪⁺ _P1 _P2 + - =

i

Fig. 1 The approach of Clique, which shows the user interaction at the bottom and the corresponding data processing on top. A user selects a focus person to research the egocentric network. This is visualized as a career overview in a timeline visualization (which?). The user can select part of this timeline to review the egocentric network changes during the selected period in the change visualization (what?). By selecting another actor, the changes in the relationship between the two actors are displayed in the relational visualization (when?). Details on positions/organizations and actors in the network (i) can be requested from all visualizations.

(7)

screen space versus detail. The focus on change gives a second problem, it means we completely discard existing, activated and deactivated contacts: contacts that remain static in between the two time points. From a journalistic point of view all contacts could be significant, not just the changed ones. However, including this contacts in the visualization means extra clutter. To solve these two problems we use an adjusted coarsened difference map as introduced by Archambault [3]. A difference map merges two input graphs in a single graph while preserving the network structure. By coarsening the branches to their difference value, each set of connected nodes and edges is clustered as illustrated in figure 5. The coarsened difference map technique makes it possible to show all changes in a single visualization, and include existing contacts without adding a lot of clutter.

Due to the tree-like structure of the egocentric network, we need to make some adjustments to the coarsened difference map technique. Archambault et al. [4] defined the path-preserving hierarchy definition, which merges degree one nodes (in this case begin- or end nodes) into metanodes based on their difference value. Although this technique relates to complicated networks, it is easy to apply to the straightforward hierarchy of the egocentric network. Since change in first degree nodes often influences change in second degree nodes, it makes sense to use this technique to coarsen both of the hierarchic layers in the graph. The addition of interactivity makes it possible to ’unfold’ coarsened branches to learn which nodes it contains (an example is shown in appendix B.5 and B.6). To make a visual connection with the egocentric timeline visualization, this project matches branches not only on their difference values, but on their sector categorization as well. The sector categorization is represented by the same color as used in the timeline visualization. The difference value is represented by a symbol. The structure of the egocentric network has that allows the exploration of the timeline in more detail

is the ability to hide sectors by clicking on the legenda. Clicking and dragging in the timeline enables the user to select a desired timeframe, which triggers the change visualization below the timeline.

4.2 When & which -> what: Change visualization

approach

The egocentric timeline visualization allows the user to define a period and pass the data selection to the second visualization, the change visualization. In this visualization the behaviour of the nodes (what) is displayed: the changes in the network elements (which) between start- and end time of the selected period (when), as illustrated in figure 3.

Apart from a clear distinction in the degree of the nodes in the visualization, we should carefully consider the various changes that can occur during a period. A political influence network can be considered a dynamic social network, which implies that certain network changes should be carefully interpreted. A lost connection is often not (immediately) lost in the real world. Therefore we choose to use the term ’deactivated’. Figure 4 identifies five types of change during a period compared to a selected period, based on the relations between two time intervals as described by Allen [1]. When first degree nodes are categorized to one of these types, the second degree nodes sometimes automatically share the same type of change: • Deactivated organizations always contain deactivated

members.

• Activated organizations contain members that can have any state (except from members in the future, which are irrelevant).

Visualizing change in a comprehendible manner has never been an easy task. Focusing on temporal change is a relatively new approach, though still a time-to-space mapping (as are, for example small multiples). This leaves us with the usual problem for these kind of visualizations:

time Activated Deactivated _𝛪𝛪⁻ 𝛪𝛪⁺ Allen-relations X contains Y X overlaps Y X before Y X overlapped by Y X during Y X after Y Changed x x x New x x X versus Y Symbol = + / _ ± a b c d e f Y

Fig. 4 Five kinds of node changing during a period, modified from Allen [1], categorized by a combination of two factors: changed/unchanged and activated/deactivated. This creates four types of nodes: (a) activated and unchanged, (b) activated and changed, (c) deactivated and unchanged. The last one, deactivated and changed, is divided in two parts to indicate whether the deactivated actors were already part of the network on the start time or newly added: (d) existing deactivated and changed, (e) new deactivated and changed. The last type (f) is included for completeness, but is irrelevant because it lies in the future.

Timeline

Snapshots (time restricted egocentric network )

In-between change (egocentric network changes)

𝛪𝛪⁻

𝛪𝛪⁺

Fig. 3 The visualization of the change in a network instead of network snapshots or animation.

(8)

shared first degree nodes. Other actors can be present in both egocentric networks through various organizations on different moments in time. These are indirect links between two actors, and indicate the closeness

of the two networks. The shared second degree nodes are displayed in a table, which shows the name of the shared contact, and for both actors the name of the organization through which they are related.

• Shared attributes: due to time restrictions this project focuses on the shared nodes. However, an interesting shared attribute in this case would be the sector, which could offer a high information gain.

The relational visualization is split in two parts: on top the shared first degree nodes, below the shared second degree nodes. Hovering shows details on demand. In the bottom of the screen is a button, which allows the user to generate the egocentric network of the selected actor (illustrated in appendix B.7).

4.4 Summary

Clique uses three successive visualizations, which enables the user to select parts of the data in one visualization and review them in the next. The egocentric timeline visualization shows the time intervals of the first nodes, making use of the Triangular Model. The data can be filtered by selecting parts of the timeline. The changes during the selected period are shown in a change visualization, which consists of an adjusted interactive coarsened difference map. By selecting a second degree node, a pop-up shows the relational visualization, exploring the relationship between the two actors.

5. EVALUATION

To answer the research question, this project tackles two research objectives:

• Effectiveness: Can the application provide insights to the journalists on the data they were not able to obtain before? How? Why?

• User experience & intelligibility: Does the user understand the results? What are suggestions of improvement according to the target group? consequences for the spatial cues as well. The ego node

itself is a constant, can never be coarsened, and is always present, preferably on the same spatial position, i.e. in the approximate center of the visualization. Additionally, spacial cues are added for the approximate location of the nodes in the first degree, which facilitates the location of organizations by the user. This is accomplished by building the graph chronologically.

4.3 Which & what -> when: Relational visualization

approach

Although the timing of first degree nodes changes can be derived from the timeline visualization, the exact timing of second degree nodes changes gets lost in translation. Even by selecting a period, one can only determine the behaviour of that node in between the begin and end time of that period using the change visualization. The exact moment of change (when) in the relationship (what) between the egonode and the selected second degree node (which) is therefore visualized in the third visualization, the relational visualization.

Since the relational visualization is examining the evolving relationship between two actors, the focus is on the entire network and the entire timeframe. We should be careful not to confuse the user, who already defined a period on the timeline, and link a clear timeframe to the information. This visualization is therefore placed in a pop-up window, accessed by right clicking a second degree node in the change visualization. By visualizing the entire period the user can not only derive the exact moment of change in the relationship, but also identify other shared connections:

• Shared first degree nodes: the positions held at the same organizations during the same period. Since the amount of shared intervals is expected to be negligible, the intervals are displayed in a Gantt chart. The intervals of one actor are placed above the x-axis and those of the other actor below, as shown in figure 6. This makes it easy to compare time intervals, and to find the corresponding period.

• Shared second degree nodes: these shared nodes are not equal to the second degree nodes directly attached to the

time

Egonode Selected person

Fig. 6 The mirrored Gantt chart explores the shared first degree nodes. Only time intervals at the same organization on overlapping periods in time are displayed.

(a) input graph (b) input graph (c) difference map (d) coarsened difference map

Fig. 5 The construction of a coarsened difference map modified from Archambault [3] (d) is based on two input graphs (a) and (b) that are merged into one single difference map (c). The grey nodes and edges in graph (c) correspond with graph (a), the black nodes and edges with graph (b) and the blue nodes and edges in both (a) and (b). Graph (d) displays the clustered nodes and edges into metanodes based on their connection and difference value

(9)

network. Another point with the same level of agreement was, that the use of Clique speeds up the process of collecting data. Three out of five felt that the tool helps them find unexpected connections between involved actors, which can lead to new research impulses. Information need - Four out of five respondents had bigger information needs than Clique could provide. This can partially be solved by including more data from the database, e.g. the use of an extended job description instead of the current abbreviated version. Another quick fix is the addition of more details on demand. All respondents felt the need to answer more profound questions that arose from the questions answered: who is that contact? To which political party does he belong? What does that organization do, and what sector is it in? Background information gives meaning to the presented content and is indispensable while working on a case.

Usefulness - Although all respondents had the option to search the answers to their questions in the browse-environment, they all spend most of their time in Clique. The shared opinion was that Clique contributes to establishing connections and drawing conclusions. Besides that, four out of five felt that Clique supports free exploration of the content combined with a certain serendipity; by clicking around the user discovers new connections.

Steep learning curve - It took all users some time to get used to the visual idiom of Clique. Respondents who are used to work with data mastered the tool more quickly than those with less experience. However, at the end of the interview four out of five felt confident that the tool would be useful, once familiar with it.

Misunderstandings - Getting accustomed to the egocentric timeline visualization took some time, but caused few difficulties. The change visualization on the other hand, produced quite some misunderstandings. The used icons are hard to interpret, as is the theory of Allen relations behind it. A lot of wrong conclusions originated from the simple assumption that the coarsened difference map points were endpoints and stopped there. Another misconception was caused by the perceived meaning of the values behind the edge labels. Additionally, including of the inactivated/unchanged nodes (the former positions) led to quite some confusion.

Improvement suggestions - The respondents frequently mentioned the issue that the search field requires a full name -- an unfinished part in the prototype. As the search field should be guiding the user, and make it easier to find people by giving suggestions while you type. In the change visualization improvements can be made by adding a sector legenda and rethinking the way large numbers of nodes are displayed.

5.1 Evaluation Method

Five journalists, experts on the field of politics, were asked to individually interact with the prototype (screen prints are provided in appendix F). This individual approach is a deliberate choice, since the target group consists of expert users with completely different backgrounds, who can be bound to professional secrecy. By combining the results, it is possible to study the prototype in the context of differences among users, which will reveal the points with and without user consensus.

The prototype is tested by interviewing the target group on the use of Clique compared to the raw database data, accessed through a ’browse’-environment where the information is formatted in lists. The insights, needs, misunderstandings and improvement suggestions of the target group are necessary for further development. The prototype is not yet in the phase of usability testing; therefore a walkthrough of the system is provided. Then, the journalist will work on either a given case (presented in appendix C) or a case of their own, and answer a set of questions, while leaving room for personal input. They can freely use the database and Clique, and are asked to elaborate on their decisions out loud. The results are analyzed by coding the transcripts on the main research objectives.

5.2 Discussion of Evaluation Results

The interviews were conducted with five journalists, four male, one female, in varying ages, fields of expertise and data-skills. Except for one, all respondents worked on the predesigned case. The first three questions, which focused on a single compound task, were quite easy to answer for all the respondents. The subsequent questions provoked more discussion and revealed a glimpse of the respondents’ individual research approach.

Limitation - The biggest limitation in this research is the incomplete data of TNL, which becomes even more evident as it is visualized in Clique. The judgement of the effectiveness of Clique becomes biased when Clique is designed to answer questions, without containing the answer. Another limitation is the fact that Clique intends to support journalists in their research, which is a situation that is difficult to imitate during the tests. Respondents that chose their own case did not know exactly what to expect, thus came unprepared. Respondents that chose the predesigned case, lack

background knowledge on this case, which makes it harder to interpret results, and find more insights than

the answers to the questions.

Effectivity - Overall, the journalists were quite positive on the effectiveness of Clique. Four out of five agreed that Clique is a good starting point for an investigation, because it provides them an impression of a politician and his

(10)

7. ACKNOWLEDGEMENTS

I thank Marcel Worring, my supervisor at the University of Amsterdam for his valuable feedback in the process of this thesis. Thanks to Maarten Marx (UvA) for admitting to the role of co-assesor. Special thanks to the people of Waag Society, especially to Ivonne Jansen-Dings for her enthusiastic coaching, and Stefano Bocconi for his cooperation and time. My gratitude goes out to all the journalists that put their time and effort in the exploratory interviews and prototype tests. Last but not least I would like to thank my husband Mark Pot for his patience and support, and Hanneke Derksen for her valuable improvement suggestions.

Nice-to-haves - All respondents indicated that they would like to adjust the data to fit their own research: add nodes, remove nodes, or adjust the categorization. This calls for an import and export function. The comparison of two random egocentric networks instead of following the temporal research path is another appealing idea to the respondents. Also they would like to have possibilities to use the data in an article, for example the option to embed Clique in their article or generate statistics.

6. CONCLUSION AND FUTURE WORK

Clique enables journalists to research political influence networks to a certain extent. Clique supports exploration of the database content by visualizing change, allowing the respondents of the tests to establish connections they were not able to obtain before. Clique is considered useful, particularly during the initial phase of research while collecting data. We think this process could be facilitated even more by adding an export function to Clique. The extent in which Clique can support journalists in their research depends heavily on the quality of the data displayed. When the data becomes more complete, Clique becomes more effective and will be able to provide more reliable insights to journalists. Therefore we recommend extending the TNL database, especially the information related to organizations.

Although Clique is accessible for all journalists, the interviews indicated that the lower the data-related skills of the respondents are, the steeper the learning curve becomes. The usability should therefore be given a lot more attention, for Clique to be useful without an expert guide. Especially the change visualization needs extra attention, because it caused most faulty conclusions, but is also considered the most interesting visualization and examined intensely during the tests.

A very interesting direction for further application development is the comparison of two random egocentric networks. This would provide journalists the opportunity to gain more insight in the relationship between two actors, and calculate the proximity of their careers and networks. In the current version of Clique, it is possible to gain basic insight in the similarities in the networks of two actors, that are linked through their egocentric networks. The utility of Clique would be greatly enhanced if an extended version of this research option would be added.

(11)

[16.] Liu, Q., Hu, Y., Shi, L., Mu, X., Zhang, Y., and Tang, J. (2015). Egonetcloud: Event-based egocentric dynamic network visualization. In Visual Analytics Science and Technology (VAST), 2015 IEEE Conference on, pages 65–72. IEEE.

[17.] Moody, J., McFarland, D., and Bender-deMoll, S. (2005). Dynamic network visualization1. American Journal of Sociology, 110(4):1206–1241.

[18.] Peuquet, D. J. (1994). It’s about time: A conceptual framework for the representation of temporal

dynamics in geographic information systems. Annals of the Association of american Geographers, 84(3):441–461. [19.] Qiang, Y., Delafontaine, M., Versichele, M., De Maeyer,

P., and Van de Weghe, N. (2012). Interactive analysis of time intervals in a two-dimensional space. Information Visualization, 11(4):255–272.

[20.] Shamir, A. and Stolpnik, A. (2012). Interactive visual queries for multivariate graphs exploration. Computers & Graphics, 36(4):257 – 264. Applications of Geometry Processing.

[21.] Shi, L., Wang, C., Wen, Z., Qu, H., Lin, C., and Liao, Q. (2015). 1.5 d egocentric dynamic network visualization. Visualization and Computer Graphics, IEEE Transactions on, 21(5):624–637.

[22.] Shneiderman, B. (1996). The eyes have it: A task by data type taxonomy for information visualizations. In Visual Languages, 1996. Proceedings., IEEE Symposium on, pages 336–343. IEEE.

[23.] Thomas, J. J. (2005). Illuminating the path:[the research and development agenda for visual analytics]. IEEE Computer Society.

[24.] van den Elzen, S., Holten, D., Blaas, J., and van Wijk, J. J. (2016). Reducing snapshots to points: A visual analytics approach to dynamic network exploration. Visualization and Computer Graphics, IEEE Transactions on, 22(1):1–10.

[25.] von Landesberger, T., Kuijper, A., Schreck, T.,

Kohlhammer, J., van Wijk, J., Fekete, J.-D., and Fellner, D. (2011). Visual analysis of large graphs: State-of-the-art and future research challenges.

[26.] Waag Society, University of Amsterdam, and University of Leiden (Centre for Innovation) (2016). Transparant Nederland; bestuurlijk nederland in beeld.

https://transparantnederland.nl/.

[27.] Yi, J. S., ah Kang, Y., Stasko, J. T., and Jacko, J. A. (2007). Toward a deeper understanding of the role of interaction in information visualization. Visualization and Computer Graphics, IEEE Transactions on, 13(6):1224–1231.

REFERENCES

[1.] Allen, J. F. (1984). Towards a general theory of action and time. Artificial intelligence, 23(2):123–154.

[2.] Amar, R., Eagan, J., and Stasko, J. (2005). Low-level components of analytic activity in information visualization. In Information Visualization, 2005. INFOVIS 2005. IEEE Symposium on, pages 111–117. IEEE.

[3.] Archambault, D. (2009). Structural differences

between two graphs through hierarchies. In Proceedings of Graphics Interface 2009, pages 87–94. Canadian Information Processing Society.

[4.] Archambault, D., Munzner, T., and Auber, D. (2008). Grouseflocks: Steerable exploration of graph hierarchy space. IEEE transactions on visualization and computer graphics, 14(4):900–913.

[5.] Archambault, D., Purchase, H. C., and Pinaud, B. (2011). Animation, small multiples, and the effect of mental map preservation in dynamic graphs. Visualization and Computer Graphics, IEEE Transactions on, 17(4):539–552. [6.] Bach, B., Pietriga, E., and Fekete, J.-D. (2014).

Graphdiaries: animated transitions andtemporal navigation for dynamic networks. Visualization and Computer Graphics, IEEE Transactions on, 20(5):740–754. [7.] Beck, F., Burch, M., Diehl, S., and Weiskopf, D. (2016). A

taxonomy and survey of dynamic graph visualization. In Computer Graphics Forum.

[8.] Centre for Innovation, Leiden University, in

collaboration with the University of Amsterdam (2016). Lynksoft; leverage your network. https://lynksoft.com/. [9.] Farrugia, M. and Quigley, A. (2011). Effective temporal

graph layout: A comparative study of animation versus static display methods. Information Visualization, 10(1):47–64.

[10.] Few, S. (2009). Now you see it: simple visualization techniques for quantitative analysis. Analytics Press. [11.] Fink, K. and Anderson, C. W. (2015). Data journalism

in the United States: Beyond the “usual suspects”. Journalism Studies, 16(4):467–481.

[12.] Franklin, B. (2014). The future of journalism. Journalism Practice, 8(5):469–487.

[13.] Janssen, M., Charalabidis, Y., and Zuiderwijk, A. (2012). Benefits, adoption barriers and myths of open data and open government. Information Systems Management, 29(4):258–268.

[14.] Kulpa, Z. (1997). Diagrammatic representation for a space of intervals. In Machine Graphics & Vision, pages 5–24.

(12)

LIST OF APPENDICES

A Detailed Description Transparant Nederland Dataset

B Initial Design Clique

B.1 Home

B.2 Egocentric timeline visualization

B.3 Egocentric timeline visualization with node-selection B.4 Selection period in the egocentric timeline visualization B.5 Change visualization

B.6 Unfolded change visualization B.7 Relational visualization

C Predesigned case for test

D Transcripts

D.1 Coding scheme D.2 Respondents D.3 Interview Remy Koens D.4 Interview Teun Gautier D.5 Interview Dimitri Tokmetzis D.6 Interview Gemma van der Kamp D.7 Interview Roy Mevissen

E Analysis

E.1 Perceived effectivity E.2 Insightfulness E.3 Data reliability E.4 Extra information need E.5 Perceived Usefulness E.6 Nice-to-haves

E.7 Perceived understandability E.8 Misunderstandings E.9 Improvement suggestions

E.10 Background jounalistic research approach

F Prototype Clique

F.1 Home

F.2 Egocentric timeline visualization F.3 Change visualization F.4 Relational visualization

12

13

13 14 15 16 17 18 19

20

21

21 21 22 28 36 40 46

50

50 52 54 55 56 57 59 60 61 62

63

63 64 65 66

(13)

APPENDIX A: DETAILED DESCRIPTION TRANSPARANT NEDERLAND DATASET

Included sources

• Secondary positions list Members of House of Representatives.

• Governmental organizations; information on all the governmental organizations in the Netherlands, publication by the Ministry of the Interior and Kingdom Relations.

• PDC Staten Generaal; information on the members of the parliament since 1996. • Statistical Classification of Economic Activities in the European Community, Rev. 2.

• Netherlands Authority for the Financial Markets (AFM); overview of all substantial holdings and gross short positions in issuing institutions and shares with special controlling rights.

• Members of the BVPA Dutch Association for Public Affairs (BVPA: Beroeps Vereniging Public Affairs). • Chamber of Commerce (Dutch: KvK) numbers from the Company Register.

The different variables the data holds can be described by

• Actors; the politicians (name).

- Gender

• Positions; the positions the politicians hold, now and in the past.

• Organizations; the company or organization to which somebody is associated, can be a non profit organization or a political party as well.

• Sector; the sector an organization belongs to. • Timepath;

- Begin date - End date

(14)

APPENDIX B: INITIAL DESIGN CLIQUE

B.1 Home

CLIQUE by Marieke van Kouwen in collaboration with Waag Society Master Thesis Information Studies Human Centered Multimedia University of Amsterdam

User guide

CLIQUE

Explore the political influence network in the Netherlands. Dig in the Transparant Nederland data set. Start your exploration here by typing the name of a politician.

(15)

B.2 Egocentric timeline visualization

CLIQUE by Marieke van Kouwen in collaboration with Waag Society Master Thesis Information Studies Human Centered Multimedia University of Amsterdam User guide CLIQUE 15 10 5 0 D ur ation in y ears time in years Political Party 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 2

Other services & activities 9 / 12 Education 5 / 15 Electricity, gas and water supply 1 / 1 Healthcare 4 / 5

Sectors Click to show/hide

ANNE FLIERMAN / timeline

Click and drag to select a period Scroll down for the result

(16)

B.3 Egocentric timeline visualization with node-selection

(17)

B.4 Selection period in the egocentric timeline visualization

(18)

B.5 Generated change visualization

ANNE FLIERMAN / social network changes 03/89 - 01/04

Symbols + added active ± added deactive - changed to deactive = unchanged active / unchanged deactive active connection deactive connection same person -+ -± ± + / = = + =

(19)

B.6 Unfolded change visualization

Ella Vogelaar

Jo Ritzen André Postema

(20)

B.7 Relational visualization

CLIQUE by Marieke van Kouwen in collaboration with Waag Society Master Thesis Information Studies Human Centered Multimedia University of Amsterdam

User guide

CLIQUE

Explore the political influence network in the Netherlands. Dig in the Transparant Nederland data set. Start your exploration here by typing the name of a politician. 15 10 5 0 D ur ation in y ears time in years Political Party 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 2

Ella Vogelaar

Jo Ritzen André Postema

ANNE FLIERMAN / shared connections ELLA VOGELAAR

X

INDIVIDUAL TIMELINE

Shared positions

Shared sectors

Shared (assumed) contacts Ella Vogelaar Contact André Postema Jo Ritzen Pauline Meurs Ivo Opstelten Margot Vliegenthart NVAO NVAO NVAO

Ministerie van Binnenlandse Zaken Adviescommissie Normering Inburgeringseisen

NVAO NVAO NVAO RVZ

Universitair Centrum Sportgeneeskunde Ella Vogelaar’s link Anne Flierman’s link Anne Flierman

80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16

(21)

APPENDIX C: PREDESIGNED CASE FOR TEST

Wet Stroom

korte beschrijving (uit: “Agressief lobbyen, een reces-rel en een bezoedelde reputatie” - NRC Handelsblad 9/1/16)

“De Wet Stroom regelt uitvoering van het energieakkoord – zoals de aanleg van windmolens op zee. Ook preciseert hij eerdere wetgeving, zoals de zogenoemde splitsing: energiebedrijven mogen niet langer beheerder van stroomnetten en tegelijk stroomproducent zijn.

Den Haag voerde deze ontvlechting eerder op last van de EU in, waarna bedrijven als Nuon en Essent voor miljarden werden verkocht. Vervolgens bleek dat geen andere EU-lidstaat dit deed. Eneco en het Zeeuwse Delta bleven de splitsing bestrijden, en procedeerden er – vergeefs – tot bij de hoogste rechter tegen.

De bedrijven hadden pech: de coalitie en minister Henk Kamp (Economische Zaken, VVD) waren compromisloos. Zij redeneerden dat splitsing tot lagere consumentenprijzen en betere marktwerking leidt. De aanleg van windmolens op zee maakte het voor de coalitie ondenkbaar de wet te herzien. En Kamp wees erop dat splitsing al door vier ministers was verdedigd, inclusief twee CDA’ers.”

(…) Senatoren van verschillende partijen vertelden me ook dat ze merkten dat de CDA-woordvoerder, Anne Flierman, man met veel ervaring in lokaal bestuur en onderwijs, in informele contacten uitgesproken tegen de splitsing was. “Bijna verbeten”, zei Verheijen.

En alles escaleerde toen een meerderheid van de senaat, onder wie Flierman namens het CDA, op 22 december al in de eerste termijn een motie tegen de splitsing indiende.

Hoogst ongebruikelijke powerplay: normaal is dat je eerst het antwoord van de minister aanhoort. VVD-woordvoerder Helmi Huijbregts fluisterde daarop diverse senatoren in dat Flierman nota bene commissaris bij Cogas is, een

gasnetbeheerder die soms ook stroomproductie doet. Belangenvermenging. Zowel Verheijen (PvdA) als Pijlman (D66) keken hiervan op, vertelden ze me – hoewel Fierman dit commissariaat gewoon had opgegeven bij de Eerste Kamer.

Huijbregts vermeldde in haar bijdrage aan het debat haar eigen toezichthoudende rol bij Intergas, een gasbedrijf in liquidatie. „Ik had verwacht”, zei ze me deze week, „dat Flierman hetzelfde had gedaan met Cogas.”

Flierman vertelde me dat zijn fractie wist van zijn commissariaat bij Cogas, en niettemin instemde met zijn woordvoerderschap. „Er speelde geen persoonlijk gewin, en Cogas had geen belang.”

Dit laatste bestrijdt Economische Zaken. „Als de minister de motie van een meerderheid van de senaat tegen de splitsing had uitgevoerd, was het voor Cogas mogelijk geworden weer productiewerk te doen’’, zei een woordvoerder van Kamp me.”

Vragen

Hoeveel banen houdt Flierman er op na naast zijn werk in de kamer van 16/06/2009 tot 01/06/2016?

10

Wie is/zijn er toegevoegd aan het netwerk van Flierman in 2003

Jo Ritzen

Wat is de periode dat deze persoon en Flierman voor dezelfde werkgever hebben gewerkt?

College van bestuur, universiteit Maastricht: f: 01-04-2001/01-08-2005 r:01-02-2003/01-02-2011 —> 01-02-2003 - 01-08-2005

Breng het netwerk rondom Flierman rond 2015 in kaart.

Binnen welke sectoren beweegt Flierman zich tot op dat moment?

Welke mensen van binnen het CDA heeft Flierman ook via andere bedrijven in zijn netwerk? Welke personen heeft Flierman al vanaf zijn eerste baan in zijn netwerk (1983)?

(22)

APPENDIX D: TRANSCRIPTS

D.1 Coding scheme

Effectivity (Can the application provide insights to the journalists on the data, they were not able to obtain before? How? Why? ) Perceived effectivity

Insightfulness

Data reliability

Extra information need

User experience (What is perceived as useful?) Perceived Usefulness

Nice-to-haves

Intelligibility: (Does the user understand the results?)

Perceived understandability

Misunderstandings

Improvement suggestions

Background information target group

Background information journalistic research approach

D.2 Respondents

Teun Gautier

For a long time Teun Gautier (1968) worked in the media sector, inter alia at Elsevier. He worked as a publisher of De Groene Amsterdammer and as financial director of the news website De Correspondent. He also formed the basis of Publeaks, a secure website for whistleblowers. He is currently working on De Coöperatie, a platform for independent journalists, copywriters, photographers, audiovisual journalists, designers, data journalists, web journalists and other freelancers in the media.

Remy Koens

Remy Koens is freelance datajournalist at one of the leading newspapers of the Netherlands, de Volkskrant. His professional interests concern politics, education, environment and social services.

Dimitri Tokmetzis

Dimitri Tokmetzis (1975) is datajournalist at the news platform De Correspondent. His work shows the hidden world behind computers, smartphones and other digital equipment. How these machines affect our daily lives, in different areas e.g. the security industry, technology, ethics and open government.

Gemma van der Kamp

Gemma van der Kamp is a trained anthropologist with a special interest in international development, journalist and teacher innovation & storytelling tools at Fontys University of Applied Sciences, Bachelor Journalism.

Roy Mevissen

Roy Mevissen (1973) is a experienced freelance journalist and teacher journalism at Fontys University of Applied Sciences, Bachelor Journalism. In the past, he worked for several (regional) newspapers and magazines. His field of expertise: innovation journalism.

(23)

D.3 Interview Remy Koens

Date: 23 June 2016

Duration: 72 minutes

Interview and transcript language: Dutch Transcript starts after explanation at 00:20 Key

RK = Remy Koens (interviewee) SB = Stefano Bocconi (technical support) MT = Marieke van Kouwen (author)

RK: Er zijn wat bronnen waarbij ik wat sneller mijn bedenkingen heb, dus ik wil altijd wel graag zien waar informatie vandaan komt als ik het dan via een derde partij opvraag.

MK: En hier gaat volgens mij nog iets mis, want geheid dat Ivo Opstelten en Mark Rutte gedeelde contacten hebben voor een groot deel, maar er komt niks naar boven.

RK: Nou op zich, het is het gat waar Rutte kamerlid was volgens mij, en Opstelten niet en toen werd hij premier, of hier pas? Nou ik weet het ook niet helemaal goed. Dat staat er nu ook niet duidelijk in toch? Dat ze samen in de regering hebben gezeten.

MK: Nee, niet zo, in principe staat dat er wel in volgens mij. RK: Ja in parlement en politiek staat dat geheid.

MK: Ja precies, maar af en toe zijn er nog wat hiccupjes, zeg maar. RK: Ik snap dat, hoeveel moeite erachter zit om de data goed te krijgen.

MK: Ja. En we wilden je vragen om er doorheen te klikken als je dat leuk vindt. Ik zal er een muis aanhangen als je dat prettig vindt?

RK: Ja hoor.

MK: Alsjeblieft. En ik heb een case als je dat leuk vind of misschien heb je er zelf een die je zou willen uitzoeken? RK: Ik ga even klikken en daarna wil ik die.

MK: Ja, dat is goed. Je moet de hele naam intikken, dat is ook nog een beetje.. er zit nog geen zoekhulp in. RK: Hele naam, o sorry. Dat is toch wel iemand die..

MK: Het klinkt als een voornaam.

RK: Het is toch Rintjes. Eens kijken of hij erin staat.

MK: Hij staat er wel in anders krijg je meteen een nee. En je zou anders even naar die andere kunnen schakelen, dat je af en toe even kijkt wat staat daar weer.

RK: Ja hoor. Ok.

MK: Ales wat trouwens niet geïndexeerd is qua sectoren, en dat is echt heel veel nog, dat staat onder Other. Dus kan nog veel meer informatie op die manier in.

RK: Hmhm, Shell… ja. Ok. Even kijken.. MK: Wat zou je van hem willen weten?

RK: Ik ben benieuwd of de vorige machtigste man ook in zijn netwerk zit. Of dat een beetje een ons-kent-onsje is. MK: Maar is hij ook politicus?

RK: Nee, hij is vooral...

MK: Dan staat het er niet in. Er staan echt alleen politici in want je zit natuurlijk met.. met allerlei dingen over persoonlijke gegevens publiceren en ze hebben ervoor moeten kiezen om dat..

SB: Ja dat ook en de bronnen zijn niet beschikbaar. MK: Het zou wel leuk zijn om te zien.

RK: Ok, even kijken dan moet ik daarop rechtklikken. Hans Weijers, Gerrit Zalm.. Ik wist niet dat hij dit heeft gedaan. Ja, bij Shell. Dat vind ik ook wel heel leuk om te zien.. O, ik zit er een beetje verkeerd naar te kijken. Ik zou het wel prettig vinden ik hier bij deze personen ook nog een korte beschrijving, een titel of iets dergelijks.. Nou ja, hier zitten heel veel

namen tussen die ik wel ken, maar degene die ik dan niet ken, dat is dan weer.. Nou ik kan wel inschatten; nou dat is

iemand ongeveer daarvan, maar… eens kijken. Naar de case. Die kent ie, o ja, hier kent hij er al een stuk meer.

MK: Het kan zo zijn dat die dus daar weer op een ander moment heeft gewerkt, toen hij er al niet meer zat. Daarom zijn die niet opgenomen in die andere.

RK: Ja precies.

MK: Omdat het nu suggereert dat hij die allemaal kent, maar..

(24)

Deze is er nu bij, Shell.

MK: Ja dat zag je net ook, he; want hij had eerder niet een link met hem.

RK: Ja, precies. Dat is ook een an de redenen waarom hij een van de machtigste personen van Nederland is genoemd door ons, de Volkskrant. Wij maken elk jaar een soort ranglijst waarin we proberen de top 200 machtigste mensen achter de schermen en nou ja, er een ranglijstje van te maken. Laatste keer was dat dus Hans Weijers en dat jaar ervoor was hij het voor het eerst. Dat ging dus over 2014. In 2014 werd die lijst dus gemaakt. Dus dat was net toen hij voor het eerst de heer Gerrit Zalm in zijn netwerk had en dat is een zware jongen om erin te hebben. Ja en dat komt wel overeen met dit. Oké en heb je een case voor mij?

MK: Ja, ik had wat dingetjes opgeschreven, het gaat over de wet Stroom. Ben je daar een beetje bekend mee? Ik heb hier een korte samenvatting, de wet Stroom, een energieakkoord. Je kunt het lezen; een van de dingen waar het op neerkwam was dat Anne Flierman ineens een motie tegen die wet indiende, enorm actie ging voeren tegen die wet en dat hij dus achteraf bleek commissaris van de Cogas te zijn. En dat was wel bekend maar dat wist eigenlijk niemand. Uiteindelijk is die motie aangenomen maar heeft de minister gezegd: Ik ga het niet doen. Dus uiteindelijk heeft iedereen niks gekregen. Daar kwam het op neer. Ik heb wat vragen geformuleerd met als laatste vraag; wat kun je zelf nog formuleren?

RK: Ok en ik mag gewoon ze alle twee gebruiken?

MK: Ja, of dat je het om en om probeert dus dat je bij allebei probeert of je het via die tool kan beantwoorden. Ik ben daar wel benieuwd naar.

RK: Ja dat is goed. Oké, laten we op de, in mijn ogen, minst praktische manier beginnen. Volgens mij is Clique hier veel beter voor. 16 juni 2009 tot … Nou die doet ontzettend veel dingen ernaast dus dan, ik weet het niet...moet ik ze met de hand gaan tellen… nou dat ga ik niet doen..

MK: Nou dan zou ik die andere doen, daar moet je enigszins met de hand tellen, maar.. RK: Ja precies… het is wel makkelijk dat er mensen voor me zijn geweest.

MK: Nog enigszins zoekhulp. Gewoon zelf creëren.

RK: Google autofill. Ok, 16-6-2009. Kun je hier eigenlijk nog op inzoomen of niet? O nee. MK: Hij schaalt naar het hele jaar, misschien moet je hem even halverwege trekken.

RK: Ja ok, kijken wat hier staat. Hier staat negen negen. Dus logischerwijs.. ik zou heel snel zeggen dus dat is een keer de Eerste Kamer.

MK: Ja dus die had hij al dan.

RK: Ja precies dus een, twee, drie, vier, vijf, zes, zeven, acht, negen, tien, elf, twaalf.. MK: Ja, dat is een beetje...

RK: Dat kun je beter weglaten. Maar even kijken,

MK: Soms staan ze er hier soort van dubbel in.

RK: O ja, dan is dit dezelfde maar ok. Ja, dan weten we dat. Dan zal deze ook hetzelfde zijn..

MK: Ze staan altijd wel onder een samenvatting. Dus een bedrijf is er altijd maar of afgegaan, of bijgekomen of er

bijgekomen en er weer afgegaan. Je kan niet… dit bedrijf kan niet ook daar nog staan. In dezelfde blauwe, dat gaat gewoon niet, dat zou heel raar zijn.

RK: Dat snap ik, maar om het antwoord te geven op de vraag: een, twee, drie, vier, vijf, zes, zeven, acht, negen, tien, elf, twaalf, dertien, veertien, vijftien, zestien, zeventien, achttien, negentien, twintig.

MK: Je vergeet een dingetje.. RK: O jee...

MK: Alle met die schuine strepen, dat zijn alle dingen die al in zijn netwerk waren, die dus inactief zijn, waar niks veranderd is. Dat geen actieve banen zijn dus die kun je in principe dicht laten. En deze zijn… die zijn in de periode changed dus die wel..

RK: Active, ja, ik snap het. MK: Dus die vallen eraf.

RK: Ja oke, dat is fijn dat Rotterdam nu weg is,

MK: Dat scheelt weer. Hij heeft er geloof ik drie keer gewerkt, ik denk dat het daardoor komt. Dus volgens mij zitten er drie punten in, omdat in Rotterdam in de gemeenteraad heeft gezeten ofzo.

RK: Aah, op die manier.

MK: En dit zullen waarschijnlijk ook twee, als je naar boven scrollt twee selecties zien, dus dan laat hij hem twee keer zien. RK: Aah dat zijn gewoon al de hele tijd..

MK: Ja, dat denk ik ook.

RK: Maar ok, een, twee, drie, vier, vijf, zes, zeven, acht, negen. MK: Ja precies.