Saturday, October 10, 2009

Paper Session One

8:30 am – 10:00 am

Alzheimer’s Disease Research: A COIN Study Using Co-authorship Network Analytics

Aaron Sorensen, Collexis  | Ken Riopelle, Wayne State University  |  Andrew Seary, Simon Frasier University

There is a mounting body of evidence in the literature suggesting that in the era of “big science,” high impact biomedical research is typically achieved through collaboration rather than through the lone efforts of a single investigator or often even a single research lab.  Although a recent study was published measuring the productivity and impact of the world’s top Alzheimer’s investigators, no one, to the best of our knowledge, has undertaken a study to establish which forms of collaboration tend to lead to the greatest increase in impact and productivity for a given set of Alzheimer’s investigators.  Put another way, there has never been an attempt to determine which collaborative patterns come closest to optimizing output, as measured by scientific impact and productivity, for researchers working in the area of Alzheimer’s Disease.  Using MultiNet and Negopy, the BiomedExperts coauthor networks of the most productive Alzheimer investigators (as determined by a count of all of a given author’s PubMed-indexed papers containing the word, “Alzheimer” in the abstract or title) and the highest-impact Alzheimer scientists (as measured by an Alzheimer-specific H-index as well as by the total number of citations generated in Thomson’s Web of Science by that author’s post-1984, Alzheimer-specific body of work) were analyzed. A coauthor collaboration network of this group of 269 scientists was created with each investigator being assigned a specific role within the broader network or within the sub-network to which each belonged.  Regression analyses were performed to determine any statistically significant correlations between the network role and author-level productivity and impact metrics. The findings of the correlation analysis are presented in this paper.

Complexity Leadership in Healthcare: Leader Network Awareness

William R. Hanson, College of Business, Anderson University  |  Randal Ford, Director of OD, Spartanburg Regional Healthcare System

Scholars claim a complexity leadership perspective is needed for organizational adaptation in knowledge-intense organizations.  Accordingly, the authors argue that to operationalize a complexity leadership perspective requires new leader competencies embedded within complexity concepts. The healthcare industry is a case in point. Processes in healthcare are highly complex networks of interactive and mutually supporting agents, which frequently overburden bureaucratic organizational structure and leadership conventions. Leaders thus are challenged to view both organizational structure and leading processes as a means to catalyze organizational learning, problem solving, creativity, knowledge and other outcomes to successfully adapt. We support this perspective by using qualitative methods and Dynamic Network Analysis to explore leader dynamics among 15 hospital laboratory subunits. Data were collected in two stages, through structured interviews based upon grounded theory theoretical sampling methods, and a subsequent online survey open to the population under study; the online survey response rate was 66%. Employing network analytical measures and values at the laboratory subunit level of analysis, the study’s findings support earlier leadership research by Schreiber and Carley (2008). These network measures reveal leader tendencies that underlie collaborative dynamics within the hospital laboratory. Our findings underscore an alternative approach to leadership thinking and analysis, and support the need for developing complexity leadership competencies.

Analyzing the Flow of Knowledge with Social Badges

Kai Fischbach¹ |  Peter A. Gloor² |  Casper Lassenius^{3  |} Daniel Olguin Olguin^{2  |} Alex (Sandy) Pentland ^{2  |} Johannes Putzke^{1  |} Detlef Schoder¹

¹University of Cologne, Germany, {fischbach|Putzke|schoder}@wim.uni-koeln.de

² Massachusetts Institute of Technology, USA, This e-mail address is being protected from spambots. You need JavaScript enabled to view it , This e-mail address is being protected from spambots. You need JavaScript enabled to view it , This e-mail address is being protected from spambots. You need JavaScript enabled to view it

³Helsinki University of Technology, Finland, This e-mail address is being protected from spambots. You need JavaScript enabled to view it

The author names are in alphabetical order.

This paper presents a collection of “best practices” for the use of “Social Badges" that support automatic collection of informal, personal interaction between knowledge workers within an organization. Our goal is to introduce a novel approach which improves data quality over legacy methods and that allows insights into the processes and structures of an enterprise’s informal knowledge networks. The approach uses dynamic Social Network Analysis (dSNA) to make it easier for executives to analyze and manage informal communications networks. Its practical applicability was evaluated by case studies conducted in three different organizations: (1) the marketing department of a medium sized bank in Germany, (2) the post-anaesthesia care unit at a large US hospital, (3) teams of software developers in a Nordic European country. For the analysis, we tracked, amongst others, all personal interactions between the knowledge workers in a department or team using social badges worn by each employee for the duration of four weeks. We analyzed this data as well as all emails exchanged between the employees and compared it with performance data of individuals and teams.

The paper highlights 19 key lessons learnt during these studies. Whereas the first 11 lessons focus on overcoming the employee's privacy concerns to set up the necessary technology infrastructure, the second 8 lessons provide some general findings for efficient management of knowledge workers.

Finding Trendsetters in Longitudinal Networks

László Gulyás¹,² |  George Kampis¹,² |  Zalán Szakolczi² |  Ferenc Jordán³,⁴

¹ Collegium Budapest, The Institute for Advanced Study

² Eötvös University Budapest

³ Microsoft Center for Systems Biology, University of Trento

⁴ Center for Network Science, Central European University, Budapest

We consider networks as parts of nature or society, to be discovered, analyzed and understood. Nature and society are inherently dynamic, thus we deal with genuinely dynamic networks. A particularly challenging problem is the identification of trendsetters and other causally relevant agents in longitudinal network data, such as in viral infections and other social contact networks, or in scientometric data sets. We look for agents responsible for the persistence and emergence of structures in the networks of future time slices (rounds of infections, or decades of publication, etc). We develop several measures to approach this problem, and realize them in an easy-to use toolkit that facilitates analysis by non-experts.

Using novel, dynamic extensions of classic static measures is our first approach. We apply node-level measures (local clustering, betweenness centrality, closeness centrality, eigenvector centrality, Kleinberg centrality, page rank scores) and analyze their dynamics. We identify nodes with high centrality scores in more than one time slice; this way we can identify nodes which are important in a more permanent fashion. We also add stability measures to express how stable the environment of these selected nodes is. To that end, we examine the dynamics of the neighborhood of a central node: permanent neighbors of the node, changes in the strength of its connectivity, average link strength, and monotonity trends in the link fluctuations. Results are shown in intuitive graphic plots.

Our second approach is based on counterfactuals as exemplified by the traditional kick-off experiments in ecology or classical genetics. Removing which nodes will change the topology and dynamics of the network most radically - this is a question that leads to new, indirect centrality measures that identify social analogs of keystone species. We show the usefulness of these concepts on application examples.

The research is partly funded by an EC FP7 FET Open grant (www.dynanets.org).

An Autopoietic System Theory for Creativity

Takashi Iba, MIT Center for Collective Intelligence, Faculty of Policy Management, Keio University

What is creativity? This kind of question has been made from time immemorial, however we never reached to satisfactory explanation. There are several approaches to creativity, while they can be summarized to two approaches: psychological and social ones. The psychological approach generally focuses on the individuals who create something, and the social approach emphasizes the contextual aspect behind the creative activities. Although understandings by these approaches are important, one will be confronted with a problem when understanding the creative collaboration done by two or more people. How can we understand creativity in collaboration? Is that just collection of individual creativities, or is there transcendental anything beyond the individuals? Furthermore, if the creativity in collaboration is transcendental, how can we understand it, and does it differ from the individual creativity?

In this paper, we propose an alternative approach to understanding creativity rather than psychological and social ones. Our approach is based on autopoietic system theory, where an autopoietic system is defined as a unity whose organization is defined by aparticular network of production processes of elements. While the theory was originally proposed in biology and then applied to sociology, we apply it to understand the nature of creation, which we call "creative system".

Creative system is an autopoietic system whose element is "discovery", which is emerged only when the following three-part selection is occurred: "idea", "association", and "applying". Creative system is historical system, because its element, that is "discovery", is produced based on the system itself in a circular fashion. Intrinsically, nexus of "discovery" is hardly to happen, hence any kinds of "media" are necessary to emerge discoveries over time. Thus, with using these concepts, we open the way to understand creation itself separated from psychic and social aspect of creativity, and also relationship between them.