Revealing Sub-Optimality Conditions of Strategic Decisions

1 REVEALI NG SUB-OPTIMALI TY CONDITIONS O F STRATEGI C DECISI ONS H. K EMAL İ LTER Department of Management, Baskent University, Eskisehir Yolu 20.km, 06530, Ankara, Turkey Draft version June 30, 2011 ABSTRACT Conceptual view of fitness and fitness measurement of strategic decisions on information systems, technological system s and innovation are becoming more important in recent years. This paper determines some dynamics of fitness lan dscape which are lead to t erminatio n of decision makers’ r esearch before reaching the global maximum in strategic decisions. These dynam ics are specified ac cording t o manag ement decision making models and supported with simulation results. This article det ermines simulation results by means of “ Fitness Va lue” and “Prob ability of O ptimality ”. Correlation between these two concepts may be remarkable according to revealing optimal values in innovative and research-based decision making approaches beside sub- optimal results of trad itional decision m aking approaches. Keywords: Strateg ic decision making; Fitness landscape theory; Sub-optima lity;Optimality; NK Landscape; Simulation. 1. INTRODUCT ION Fitness landscape t heory is becoming to use for answering t he search of developing species which desire to r each highest peak on the potential gene space in the field of evolutionary biolo gy (Wright, 1932; Gillespie, 1984). Development of the cost landscapes in related solution spa ce is used as an approach t o com binatorial optimization problems’ solutions (Holland, 1975; Kirkpatrick et al., 1983; Palmer, 1988) in computer engineer ing and operations resear ch fields. In recent years, different approaches are used in various fields of social sciences like organizationa l change ( Beinhocker, 1999; McKelvey, 1999; Reuf, 1997), evolution of social structures (Levinthal, 1996), innovation network s (Frenken, 2000 and 2006), selection of appropriate technology (McCarthy and Tan, 2000; McCarthy , 2003), economic structures (Kauffm an, 1993) and political systems (Kollman et a l., 1992). Introducing perspective with NK model which is simplify using fitnes s landscape theory in various fields is devoted to be possible in global optimum searching on a stochastic but easily cont rollable fitness landscape that composed of pos sible fitness v alues. Local optimum points besides global optimum s are also important in fitne ss landscapes (see for detailed information: Ilter, 2007; Ilter, 2008). Local optimum point s can be se en peak point s which isn’t allow for changing possible fitness values even if alternatives of selection are changed. After all, firms can terminate t heir search for global or local optimum value(s) because of sub -optimal value(s) are accepted as best value(s) for them. “W hy does the firm generally terminate their search on the fitness landscape before finding the local optimum value yet?” is still a n unanswered question in social sciences. 2 Decisions on informati on system s, technology and innovation have a different place in firm s’ decision mechanism in terms of some aspects. There are various factors seems to be i mpor tant which affect the selection of technological structures (production technology , information technology, etc.) in firms and the decision behaviors of decision makers in making decision related to these technologies. These behaviors can be revealed by affecting the optim ality of firm’s f inal decision and the inner-fitness of decision maker in a large ext ent. It is possible to say t hat organization pr opert ies and f actors in organization hierarchy effect decisions about inf orm ation systems, technology and innovation manag ement which can b e concluded as a part of complex systems. Decisions except optimal ones (sub- optimal decisions) couldn’t r ecognized while d esign of decision mechani sm in org anization targets the optim al decision in som e conditions. 2. METHOD In thi s article, we try to dete r mine some important dynam ics that may cause of termination of firms’ searching action of optimal ity even if they couldn't reach the l ocal optimum value. These dynamics are considered toget her with some factors of organizationa l decision-making models and then supported with sim ulation results to reveal sub-optim ality conditions. Various scenarios hav e reviewed by using N K fitness landscape theory fo r determining sub-optimality conditions that cou ld be in dec isions which are re lated to inf ormation syste ms, technology and innovation. Scenar ios are developed as som e conditions wh ich are include one dec ision mak er, various numbers of subordinates and various numbers of decisions (T able 1). The state of optimality divergence and the state of sub -optimality tol eration of decision mak er emphasized in this article are supported by simulations’ results after statis tically acceptable num ber of run s. As an example, scenario L07 is reflecting a case which includes two subordinates (subordinate A has to make a decision and subordinate B has to make thr ee decisions) and a passive deci sion mak er (has to make a decision of subo rdinates’ d ecision combinat ion). Several runs of various scenarios are inspected with using NK fitness theory for determining sub - optimality state s of decisio ns which are related to inform ation system s, technology and innov ation. 3 3. RESULTS In simulation, global optim um poi nts that on t he fitness landscape (space of decision alternatives) ar e determined by generation of a landscape wh ich includes fitness values of each scenario. Probab ility of achieving global opt imum in final decision of the decision maker after evaluation of deci sion a lternatives can be defined as the t erm of “Probability of Optimality”. Sub -optim ality is appeared if decision’s probability of optimality descents below 100%. Relationship between fitness value of the final decision of the decision mak er and global optimum value on the fitness landscape can be define d as “Fitness Rate” and it can be recognized as of success f actor of decision maker’s final decision. Successes of decisions can be determined by inspection on global optimum- decision’s fitness value bias for th e decis ion that can’t ach ieve optimality bu t sub -optimality. (Figure 1 ) 4 Figure 1: Relat ionship Between Prob ability of Op timality and Fi tness Rate in Var ious Scenarios 4. CONCLUSION Probabilities of optimality and fitness values of each scenario are correlated each other during the analysis of sim ulation res ults. These correl ations show that decisions of active decision m akers (scenario code in light color) are more efficient than decisions of passive dec ision mak ers (sc enario code in dark color) in terms of the fitness value. On the other hand, there are no optimal values for each of all decisions and noticed that these decisions are not optim al but sub -optimal. Addition to this determination, decisions of active decision m akers have higher v alues th an decisions o f passiv e decision makers in terms of the probability of optim ality. Success of decision maker i s limited from the point of view of the probability of optimality in despite of these two setting s ar e stated t hat the active decision m akers are m ore successful than passiv e decision m akers. REFERENCES Beinhocker, E. D., 19 99. Robust adaptive strategies, Sloan Management Revie w, 40/3, 95 - 106. Frenken, K., 2000 . A complexity approach to inno vation networks, Researc h Policy, 29, 257 - 272. Frenken, K., 2006. A fitness landscape app roach to technological complexity, modularit y, and v ertical disintegration, Structural C hange and Econo mic Dynamics, 1 7, 288 - 305. Gillespie, J.H., 1 984. Molecular evolution o ver the mutational landscape, Evolution, 38, 1116 -1129. Holland, J., 1975. Adaptation in Natural and Artificial Systems: An I ntroductory Analysis with Applications to Biology, Control, and Artificial Intelligence. Universit y of Michigan Press: An n Arbor. Ilter, H.K. , 2007. Optimality in the Technological Fitness Landscapes: Conceptual Vie wpoint o f Searching for Summit, Pr oceedings o f 27. National Oper ations Re search and Industrial E ngineering Congress, 2 -4 July, Iz mir, Turkey. Ilter, H.K. 2008. Research on Information Technology Strategies in ter ms of Fitness Landscap e Theory, Journal of Economics and Administrative Sciences (Hacettep e University), 26/2, 163 - 182. 5 Kauffman, S. A., 199 3. T he Origins o f Order: Self Organization an d Selectio n in Evolutio n, Oxford Universit y Press, New York, NY. Kirkpatrick, S.; Gelatt, J r., C.D. ve Vecchi, M.P., 1983. Optimization b y si mulated annealing. Scie nce, 220, 671 – 680. Levinthal, D., 199 6. Learning and Schumpeterian dynamics, Malerb a, G.D. ( Ed.), Organiza tion and Strategy in The Evolution of The Enterpr ise, Macmillan Press Ltd, Basingstoke. McCarthy, I . P ., 2003. Technology management – A co mplex adap tive systems approach, International Journal of Technology Manage ment, 25/8, 728 -745. McC arthy, I. P . ve Tan, Y. K. , 2000 . Manufacturing co mpetitiveness and fitness la ndscape theory, Journal of Materials Proce ssing Technology, 107/1 -3, 34 7-352. McKelvey, B ., 199 9. Self -Organization, complexit y, catastro phe, and microstate models at the ed ge o f chao s, Baum, J .A.C. ve Mc Kelvey, B. (Ed), Variations in Organization Science, Sage Publicat ions, T housand Oaks, CA, 279- 307. Palmer, R., 198 8. Statistical mechanics app roaches to co mplex opti mization problems, Anderson, P.W .; Arrow, K.J.; Pines D., ( Ed.), The Economy as a n Evolving Complex S ystem, Addison -Wesle y: Reading. Reuf, M., 199 7. Assessing organizational fitness o n a dynamic landscape: An e mpirical test of the relative inertia thesis, Strategic Ma nagement Jo urnal, 18/11, 837 -853. Wright, S., 193 2. The roles of mutation, i nbreeding, crossbreed ing and selection in evol ution, Proce edings of the Sixth International Congres s of Genetics, 356 - 366.