Companies are increasingly turning to Enterprise 2.0 technologies to harness collective intelligence for information management and analysis. In this context, social network analysis (SNA) provides insights into which knowledge workers interact with which information assets and thus improves cooperation among them. However, current approaches to SNA disregard semantic relationships between the information assets or the relevance of the applicability of SNA by end-users. In the paper at hand we address this issue by proposing the social information ow graph (SIFG) as a tool for end-user-oriented SNA. In this sense, the SIFG provides a holistic and social perspective on relationships between individuals and information assets within an Enterprise 2.0 environment. We showcase its technical applicability by a prototypical implementation. By conducting case studies in three di erent application domains, we show that the SIFG and its explorability enable novel opportunities and use cases for end-user-oriented SNA.
Today's enterprises are facing not only the challenge of managing an
increasing amount of digital information, but also to do this in an e cient way [
On another note, data has to be processed and presented according to the speci c needs of knowledge workers in order to support decision making based on it. Knowledge workers often have di erent requirements regarding the representation of the same information, e.g., they might be interested in di erent levels of detail. End-user-oriented business intelligence (BI) tools enable them to process and represent a given information according to their needs without having to rely on IT specialists. Thereby, knowledge workers are able to de ne custom data transformations and tailored data visualizations. However, if multiple knowledge workers de ne data transformations and visualizations based on the same information, it is challenging to keep track of which information is transformed and visualized in which way.
To address this issue, information ow diagrams make the information ow from its source through its transformations to visualizations transparent. While in a non-collaborative setting, individual knowledge workers de ne how information is processed and represented based on their individual needs and independently from co-workers, in a collaborative environment users not only cooperate in managing of information, but also in its processing and presentation, e.g., by sharing transformations or visualizations, or by de ning them jointly. This inevitably leads to a more complex information ow structure, which makes it more di cult for users to understand connections between information assets.
In the present paper, we propose a concept for the utilization of information ow graphs for end-user-oriented SNA. In this sense, a social information ow graph (SIFG) not only visualizes how the information ows through a software system, but also which users are interacting with it. Therefore, it provides a social and holistic perspective on information assets and their relations. We also address the challenge of making the SIFG explorable by end-users to ensure its practical applicability. Finally, we identify concrete concerns in di erent application domains which can be addressed by a SIFG. 2
As described by Chi and Riedl's [
The concepts Data, View, and Transformation are considered as
InformationAsset s as de ned by Khatri and Brown [
* DataTransformation
InformationAsset
Data
* transforms to ► DataBinding
has ► View
* visualized in ► *
MetaData Ownership
* ◄ has
1 User *
Usage *
…
◄ captures
who uses it [
The SIFG enables the derivation of new social relationships by considering
interactions with di erent information assets which are connected to each other
through a chain of transformations. For example, the owner of a particular data
object can be associated with a user consuming the same data object through a
corresponding visualization. As a consequence, the proposed model opens new
opportunities for improving organizational social network analysis [
The prototype is built upon a model-based Enterprise 2.0 platform called
Hybrid Wikis [
Regarding the conceptual model as described in the previous section, the Hybrid Wiki entities as Data, dashboards and visualizations as View s, and MxL expressions as both DataTransformations and DataBinding s form the system's InformationAsset s. Each of those information assets is de nable and editable by users, and possesses a speci c set of meta data attributes. For example, each information asset has a version history capturing its temporal evolution, or authorization rules controlling the access to it in a collaborative environment. These meta data attributes|and particularly the user-related meta information|can be used by the prototype to enrich the aforementioned information ow graph and to generate the SIFG. Therefore, the prototype represents a full implementation of the conceptual model as presented in Section 2.
In the following, we illustrate the prototype by implementing a simple sales scenario which involves customers, orders, products, and categories. A customer's turnover is de ned as derived attribute, i.e., the value is calculated automatically according to a given computation prescription. Derived attributes represent data transformations as de ned in Section 2. In the prototype, a customer's turnover can be de ned by the MxL expression Orders.sum(Product.Price). Based on this, a query computing the total turnover of all customers can be de ned as nd Customer.sum(Turnover). The nd construct retrieves all customers. Based on this collection as well as a lambda expression referring to the turnover of each customer, the sum function computes the total turnover. In this example, there is a semantic dependency from the query to the derived attribute turnover, which in turn depends on the order's product and the respective product's price. Therefore, there is an information ow from the product's price through the derived attribute turnover to the total turnover query.
Figure 2 shows a dashboard illustrating, e.g., the temporal evolution of the daily turnover within a line chart, or a cluster map organizing the products within their categories. Each visualization is con gured separately, i.e., for each visualization users have to de ne a corresponding data binding by using MxL expressions. Based on the data model, data transformations (e.g., derived attributes), as well as visualizations and their data bindings, the prototype automatically generates the SIFG in real-time. As shown in Figure 3, the information ow's starting points are the Hybrid Wiki system's entity types (e.g., Order and Customer ) which are organized in so-called workspaces (the dark-blue boxes on the left, e.g., Northwind ). The internal structure of the types (e.g., attributes and derived attribute) is not shown, since this would add to much complexity to the view. Visualizations are illustrated by respective icons and composed to dashboards (the red boxes on the right, e.g., Northwind ). In between, there are functions representing generic data transformations, e.g., turnoverPerDay.
The prototype provides UI features inspired by Heer and Boyd [
To abstract and hide certain parts of the graph view, users can collapse workspaces or dashboards. This can reduce the complexity of the graph view and increase its understandability. As shown in Figure 5, collapsing a dashboard hides all its visualizations. All information ows which initially targeted those visualizations are now heading to the dashboard instead. 4
We conducted a case study including three cases in di erent application domains and di erent German companies. In each of those cases, knowledge workers already use a BI tool for analyzing and visualizing domain-speci c data in an end-user-driven way. The study shows concrete concerns those knowledge workers can address when using the SIFG prototype. To this end, we asked them for their opinion on the SIFG and for related concerns which they face in their daily work, and which could be resolved by our approach. The rst interviewee is an enterprise architect who currently manages the internal architecture of a German IT services provider (5,000 - 10,000 employees), while the second one is an IT infrastructure manager of a subsidiary of a German investment company (10,001+ employees). The third interviewee is a data quality manager of a German company (10,001+ employees). For the sake of discussion, we exemplarily implemented a representative dashboard for each of the observed cases.
The interviews revealed the following concerns which are addressable by the demonstrated prototype in particular, and the SIFG in general: Usage Analysis: As stated by the interviewees, "the collection of data induces the highest cost in our daily business". By using the SIFG, data model maintenance process can be optimized by stopping to maintain unused data. Stakeholder Identi cation: The interviewees stated that the SIFG could be used to identify stakeholders which are related to certain information assets, enabling them to proactively contact the users which express their interest in a given dashboard.
Impact Analysis: According to the interviewees, users tend to avoid changes of certain information assets, since they are afraid to "break something". Faced with the SIFG, the interviewees expressed that "this would certainly help to make the models and relations more transparent".
Support for Data Provenance: As stated by the interviewee, sharing data among colleagues is "part of the daily business". Thus it is "interesting to know what the story behind a given information asset is", raising questions such as "Where does the data come from? Who worked on it? When was the last change and by whom?". Providing an environment to visually explore the SIFG may reduce the "signi cant operative cost" induced by struggles when identifying co-workers that faced similar problems in the past. Addressing Compliance Demands: Driven by compliance demands, certain organizations (particularly in the nancial services domain) have to be able to reveal the full calculation process behind a given measure or visualization.
The SIFG may help to meet these demands.
Support for Data Consolidation: The interviewees stated that the SIFG is helpful to initiate consolidations of dashboards and views based on overlapping information, i.e., information which is visualized in di erent dashboards.
Huner et al. [
On another note, Dinter et al. [
Hermans et al. [
In addition to researchers, software vendors also try to utilize the
dependencies and relationships between both information assets and users in order to
expose new knowledge [
This work presents a concept and a prototypical implementation for a SIFG, which not only makes semantic dependencies between information assets visible and explorable for end-users, but also integrates relations to users of the system. In this way, the SIFG supports the identi cation of new social relationships between users, and thus fosters collaboration among knowledge workers. Its conceptual model as described in Section 2, the UI features described in Section 3, and the interviews as described in Section 4 represent the answers to the research questions raised in the present paper's introduction.
With respect to the validity of the exploratory case study, three cases form
a small foundation for both an evaluation of the SIFG as well as the identi
cation of concrete concerns addressable by it. However, the interviews represent
a preliminary study for a more extensive study on concerns in one particular
domain. While the interviewees agreed on the high potential of the SIFG, they
also outlined ethical and privacy issues [
Based on the results of the paper in hand, our future research activities will focus on the improvement of the SIFG with respect to its application in a speci c domain, namely EAM. Therefore, we aim for an extension of this concept and the prototype in order to address the speci c concerns as identi ed in the interview with an enterprise architect. Furthermore, we plan to do an extensive evaluation within a research community including more than 20 enterprises.