Simulating Self-Organization during Strategic Change: Implications for Organizational Design
11 Simulating Self-Organization during Strategic Change:Implications for Organizational Design
ANANYA SHETH and JOSEPH V. SINFIELD, Purdue University
1. INTRODUCTION
Self-organization- a characteristic of complex adaptive systems (CAS) has been explored in organiza-tional research- in management theory [Mihm et al. 2003; von Foerster 1984], firm internationaliza-tion [Chandra and Wilkinson 2017], organizational design [Clement and Puranam 2017], and strategicchange [Foster 2015]. Newer organizational forms such as networks and zero-hierarchy companies thathold the promise of self-organization are gaining prominence [Puranam et al. 2014], and theoreticalorganizational modeling is a useful technique to study them proactively via simulation [Puranam et al.2015; Simon 1976]. In this paper, we introduce a nature-inspired model to understand self-organizationof collaborative groups in three archetypal organizational designs- i. fully-networked, ii. siloed, and iii.dynamic, where each design controls intra-managerial communication in specific ways, and each mem-ber has reactive or perceptive behavioral tendencies.
We are perennially in search of improved ways of organizing such that CAS components act coher-ently without explicit communication [Simon 1996]. Nature reveals numerous examples: ants- collec-tively establish ideal pathways to food, bees- work together to hive, fish- school to protect themselvesfrom predators, and human immune system cells- multiply at intrusion for immunity. These tasks re-quire complex coordination but are achieved with minimal communication. Furthermore, goal-orientedagents in these systems learn by cooperative mechanisms rather than competition. Factually, cooper-ative management has been a dominant feature of pre-industrialization human organization whichis a disproportionately large part of our existence [von Rueden and van Vugt 2015]. Even in today’slarge-scale enterprises, often termed ‘mechanistic’, there exist goal-oriented collaborative groups. Theessence of self-organization is response to stimulus with minimal inter-agent communication, and be-cause organizational design governs communication channels, in this paper, we hypothesize that com-paring cooperative learning patterns of individually biased actors operating in differently structuredorganizations [Granovetter 1985] via simulation can provide insight into organizational design.
Agent-based models (ABMs), which are useful in simulating complex social systems [Miller and Page2009; Schelling 1971], take a bottom-up approach i.e., rules governing agent behavior are defined (bot-tom) that in aggregation generate system behavior (top). Humans- the agents of our CAS are boundedly[Simon 1991] as well as ecologically rational [Todd and Gigerenzer 2012] i.e. they make decisions ac-cording to what they know and according to the environment in which they are acting. We simulateconditions of change in enterprise strategy such that a new strategic vision is set at the top that isthen to be explored and implemented by members of a cooperative group at the lower levels. Tverskyand Kahneman [1974] have highlighted individual level biases in humans that affect the action de-cisions to be made by group members. We focus on three such biases- i. Personal experience i.e., themembers’ tendency to lean on their own previous learning, ii. Prestige-bias i.e., the members’ tendency
Collective Intelligence 2019. a r X i v : . [ c s . M A ] J u l :2 • A. Sheth, J.V. Sinfield to follow other more successful members of the cooperative group, and iii. Inertia i.e., the members’tendency to remain at status-quo. Cooperative groups often have flat structures without formal hierar-chy between group members as would normally exist (formally represented in organizational charts)between members of a standard team. Rather, in such groups, individual action decisions are based onperformance outcomes of the individual and that of the group. The individual and group performanceis defined by how far the individual and the group is from the strategic vision. The system is temporaland as it progresses, members’ search for the mandated goal and adaptively learn paths to it. Hence,the model measure is the number of iterations it takes individuals as well as the groups to achieve thecommon goal. Note that this is guided search and not a random search i.e., at each recursion a crite-rion helps individuals evaluate their solutions. Hence, overtime, the groups’ outcome tends to move inthe direction of the goal due to accumulation of individual and group memory. Further, a member actson the received individual performance feedback in two hypothesized manners- 1). Reactively- i.e., byimmediately shifting behavioral tendencies with performance feedback, and 2). Perceptively- i.e., bygradually responding to accumulated feedback, yet moving to reactive behavior as (and if) performancepressure increases with time. In addition, the three organizational designs constrain members’ com-munication. For instance, all members of a fully-networked group such as builder’s of an open-sourcedigital library may communicate freely whereas those in a siloed structure such as a traditional in-dustrial conglomerate might not. Similarly dynamic organizations having regular reshuffling of teamsmight lead to novel scenarios. Therefore, self-organization in the three designs can be compared.
2. MODEL
We choose a model mechanism that is inspired from the kinematic motion of bird flocks. Hence, thereare spatial and motion elements as shown in the position and velocity equations 1 and 2 respectively.The model search space is a 25-dimension binary space and each strategy is a ’position’ in that spacerepresented by a string vector of binary bits. Each bit is abstracted as a strategic dimension such as‘target some market’, ‘produce economically’, ‘undercut market price’, ‘maintain wallet share’. The fit-ness of each member is how far s/he is from the ideal strategy. This is measured by the Hammingdistance [Warren 2013; Steane 1996] and stored in the individual’s memory. Memory affects the indi-vidual’s decision in the following three ways:
Organizationally, this factor captures the effect of the member’s own best achieved strategic positionsthus far on the next strategy selection decision. In the equation of motion, this term accelerates themember in a direction. Mathematically, it is achieved by multiplying a self-belief coefficient of the agentwith the difference of the agent’s current strategic position and its previously achieved best strategicposition. Members with reactive tendencies will readily gain/lose self-belief with positive/negative per-formance feedback.
Prestige bias is defined as the individual’s bias in identifying best performing members and followingthem over pursuing other paths such as the individual’s own. This principle was developed from theBoyd and Richerson [1988] theory of cultural evolution and has been known to exist in groups wherelearning occurs. Similar to self-belief, it is an acceleration term, and is mathematically achieved bymultiplying a prestige-bias coefficient of the member with the difference of the agent’s current strategicposition and the hitherto globally achieved best strategic position. A member having received negativeperformance feedback is likely to follow a more successful (prestigious) member.
Collective Intelligence 2019. imulating Self-Organization during Strategic Change: Implications for Organizational Design • This factor incorporates the members’ momentum accounting for the recent most decision. The mo-mentum can be unpacked as the member’s inertia multiplied by his/her immediate previous velocity.Hence, between two recursive steps, as the member changes his/her velocity, the new velocity is af-fected in part by its immediate previous velocity. In the equation of motion, this is the initial velocityterm. Organizationally, this is the difficulty in or obligation to not deviate radically from the ‘legacy’ ofpast decisions.
We use kinematic logic to govern the spatial drift of the agents’ strategic positions. At each iteration,equations 1 and 2 govern the model. Equations 3 and 4 are representations of equation 2 for explana-tory purpose. P i ( t + 1) = P i ( t ) + V i ( t + 1) (1)where P i is the strategic positioning of an agent in consecutive evolutionary states and V i is the agent’svelocity governed by: V i ( t + 1) = W ∗ V i ( t ) + C [( P b ( t ) − P i ( t ))] + C [( G b ( t ) − P i ( t ))] (2)where W is the agent’s inertia, C and C are acceleration constants representing self-belief and pres-tige bias respectively, P b is the agent’s personal previously learned best position, and G b is the currentposition of the most fit agent, P i is the agent’s position at iteration (t). The change in strategic positionsis by means of a spatial drift and therefore, is governed by the equation of motion: N ewV elocity = Initial V elocity + Acceleration ∗ T ime (3)The drift governing coefficients depend on individual member tendencies of inertia to change, self-belief, and prestige bias. Therefore, equation (3) is re-written as
N ewV elocity = Inertia ∗ a + Self belief ∗ b + P restige bias ∗ c (4)where a = previous velocity, b = current distance from previous learned personal best, andc = current distance from most fit agent .Hence, for our model, the individual’s velocity is dependent on his/her immediately previous velocityand the two accelerations based on the complementary biases held by the individual. Organization-ally, it signifies every member’s decision of balance between willingness to change, individuality, andfollowership. Members in each design are uniquely biased to simulate diverse organizations.
3. DISCUSSION
Simulation results reveal that rate of self-learning across organizational designs is different. A fully-networked organization with members having reactive tendencies is the quickest to self-organize fol-lowed by dynamic and siloed designs. Organizational structures with perceptive members have slowerlearning. However, among structures with such members, fully-networked structure take less thanhalf the time to self-organize as compared to siloed or dynamic structures. Dynamic reshuffling of re-active and perceptive members results in different outcomes- reshuffling perceptive members does notlead to improved self-organization hinting at potential inefficiencies such as over-networking percep-tive members. The work indicates that learning in goal-oriented groups varies by member tendencyand communication structure, and therefore organizational design. Hence, coupling member tenden-cies and organizational design is a significant research focus to understanding self-organizing groups.This modeling technique can be employed to test specific hypotheses in this regard.
Collective Intelligence 2019. :4 • A. Sheth, J.V. Sinfield
REFERENCES
Robert Boyd and Peter J Richerson. 1988.
Culture and the evolutionary process . University of Chicago press.Yanto Chandra and Ian F Wilkinson. 2017. Firm internationalization from a network-centric complex-systems perspective.
Journal of World Business (2017).Julien Clement and Phanish Puranam. 2017. Searching for Structure: Formal Organization Design as a Guide to NetworkEvolution.
Management Science (Accepted) (2017).Kenneth William Foster. 2015. A Network-Learning Model of Strategic Change. In
Academy of Management Proceedings , Vol.2015. Academy of Management, 17177.Mark Granovetter. 1985. Economic action and social structure: The problem of embeddedness.
American journal of sociology
91, 3 (1985), 481–510.J ¨urgen Mihm, Christoph Loch, and Arnd Huchzermeier. 2003. Problem–solving oscillations in complex engineering projects.
Management Science
49, 6 (2003), 733–750.John H Miller and Scott E Page. 2009.
Complex adaptive systems: An introduction to computational models of social life .Princeton university press.Phanish Puranam, Oliver Alexy, and Markus Reitzig. 2014. What’s new about new forms of organizing?
Academy of ManagementReview
39, 2 (2014), 162–180.Phanish Puranam, Nils Stieglitz, Magda Osman, and Madan M Pillutla. 2015. Modelling bounded rationality in organizations:Progress and prospects.
Academy of Management Annals
9, 1 (2015), 337–392.Thomas C Schelling. 1971. Dynamic models of segregation.
Journal of mathematical sociology
1, 2 (1971), 143–186.Herbert A Simon. 1976. From substantive to procedural rationality. In
25 years of economic theory . Springer, 65–86.Herbert A Simon. 1991. Bounded rationality and organizational learning.
Organization science
2, 1 (1991), 125–134.Herbert A Simon. 1996.
The sciences of the artificial . MIT press.Andrew M Steane. 1996. Error correcting codes in quantum theory.
Physical Review Letters
77, 5 (1996), 793.Peter M Todd and Gerd Gigerenzer. 2012.
Ecological rationality: Intelligence in the world . OUP USA.Amos Tversky and Daniel Kahneman. 1974. Judgment under uncertainty: Heuristics and biases. science
Self-Organization and Managementof Social Systems . Springer Berlin Heidelberg, 2–24.
DOI: http://dx.doi.org/10.1007/978-3-642-69762-3 1Christopher von Rueden and Mark van Vugt. 2015. Leadership in small-scale societies: Some implications for theory, research,and practice.
Leadership Quarterly
26, 6 (2015), 978–990.
DOI: http://dx.doi.org/10.1016/j.leaqua.2015.10.004Henry S Warren. 2013.
Hacker’s delight . Pearson Education.. Pearson Education.