Organization

"The ELICIT Experiment: Eliciting Organizational Effectiveness and Efficiency under Shared Belief" Joshua Lospinoso Advisor: Dr. Frederick I. Moxley ABSTRACT The United States Military has undertaken a five year experiment to better understand shared belief in hierarchical and self-organizing organizations. Through the ELICIT framework, data collected from a sample of thirty-four cadets operating in the one hour exercise indicates that hierarchical organizations more effectively promote ubiquity of correct shared beliefs. Because of omitted variable bias in the linear models available, organizational efficiency analysis was inconclusive. Self-organizing networks, however, are demonstrated as likely more efficient than hierarchical networks once experimental modifications are made. The intent of this paper is to provide preliminary analysis of the first iteration of ELICIT experiments, and to submit a methodology for analyzing organizational efficiency and effectiveness under a shared belief concept.

KEYWORDS: Shared belief, ELICIT, command and control, self organizing network, hierarchical network, linear regression. INTRODUCTION Much effort has been put forth to understand shared belief within an organization. Many problems manifest themselves in analysis of organizational shared belief. These obstacles stem from both the difficult nature of quantifying shared belief and measuring organizational performance. By utilizing an inter-departmental study conducted at the United States Military Academy, this article provides methodology to overcome both of these obstacles. The experiment, ELICIT (Experimental Laboratory for Investigating Information-sharing Collaboration and Trust), entails "a series of online exercises to compare the relative efficiency and effectiveness of traditional command and control (C2) vs. self-organizing, peer-based edge (E) organizational forms in performing tasks that require decision making and collaboration." 1 Through data manipulation and panel-data linear regression, it is possible to understand what characteristics on the individual- and organizational-level contribute or detract from efficiency and effectiveness. Regression analysis shows that hierarchical organizations are better suited to achieving correct self belief and are therefore more effective. Self-organizing networks show signs of high efficiency, but more thorough experimentation in larger numbers is needed to confirm this result. Not surprisingly, the most consistent factor in efficiency and effectiveness is simply the amount of time that an organization has to achieve shared belief. Further research is planned for the next four years. With the results from this first iteration, researchers will conduct analysis that promises to further isolate the contributory factors to organizational efficiency and effectiveness.

1

Reference ELICIT Report

BACKGROUND Installed on client computers, the ELICIT software package 2 serves as the platform application for studying organizational efficiency and effectiveness. The four phase experiment entails an introduction, practice round, a one hour exercise, and a wrap up. During both the practice round and the actual exercise, thirty four subjects are randomly assigned to one of two organizations: a typical hierarchically arrayed organization (C2) and a control-free, self-organizing organization (E). These two organizations operate independently for the duration of the exercises. The goal of the organization is to identify a terrorist attack based on bits of information distributed around the organization. After ten minutes of the one hour experiment, all of the correct information has been issued to the organization. Among the correct bits of information, or factoids, are also distributed false factoids. Each entity receives four factoids, and they must corroborate within the organization to come up with the correct arrangement of who, what, where, and when. The C2 group is comprised of a squad leader, four team leaders, and twelve team members. Communications among these entities are restricted to the following graph in Figure (1): Figure (1): C2 Communications Hierarchy

Squad Leader

Team Leader

Team Leader

Team Leader

Team Leader

Team Member (x4)

Team Member (x4)

Team Member (x4)

Team Member (x4)

Each team is dedicated to identifying one key element of the terrorist attack: who, what, where, and when. The E group is comprised of seventeen entities with full communication capability across the organization. There are no defined teams, but the goal remains the same: positively identify the terrorist attack. All entities have the ability to post their information on their organization's website. Within the E group, this website is global to the organization. The C2 group has separate websites for each echelon (four teams and one squad site). The hierarchy in Figure (1) describes where each entity can post information. Entities can also share information with other individual entities. Once an entity believes that it knows any number of correct factoids, it can report its belief through the "identify" function. Table (1): Excerpt from Parsed Dataset

Id Morgan1 Chris1 ... Morgan1 3592 3 14 10 0 12 0 44 85 0 0 0 1 time 245 374 correct 0 2 who_n 0 0 what_n 0 0 where_n 0 0 when_n 0 0 complete_n 0 0 share 0 0 receive 0 0 sl 0 0 tl 0 0 tm 0 0 cf 1 1

Parity Communications in collaboration with the Higgins Trust Framework and the SocialPhysics project constructed the ELICIT software package.

2

DATA The experiment was run by cadets at the United States Military Academy 3 . Likewise, all participants were cadets. Subsequently, the data from the experiments is compiled into a multi-sheet Excel workbook. Each entity has a log organized by time with the annotated time elapsed, what kind of action the entity took, and information about that action. Identify actions include text-box style input for who, what, where, when, and why. Shared and received factoids are logged according to time, receiver, and sender. Role assignment into E group and C2 group (including position) are logged at time zero. Manipulating this rich databank provides a challenge. Through scripting techniques, the data was parsed into a .csv. Table (1) is an excerpt from the ported data set. Each observation