Yeni bir yöneylem araştırması yaklaşımı, bilişsel haritalar

Abstract

Bu çalışmada; üç farklı politikacının özelleştirme ile ilgili görüşlerinin ve düşünce yapılarının bilişsel haritalarının oluşturulması yoluyla, gerektiğinde karar verme sürecine yardımcı olması amacıyla, incelenmesi hedeflenmektedir. Bilişsel Haritaları oluşturulacak politikacıların seçiminde, karşılaştırmaya olanak tanıması amacıyla farklı politik sorumlulukları olan kişiler belirlenmiştir. Bu doğrultudan hare- * Başbakan ve DYP Genel Başkanı Tansu Çiller, * SNAP Genel Başkanı Mesut Yılmaz ve * CHP Milletvekili Mümtaz Soysal' m özelleştirme ile ilgili bilişim süreçleri incelenmiştir. Oluşturulan bilişsel haritalar, fayda/hedef/politik/dışsal değişken sıklığı; yol dengesi; politik tercih tutarlılığı; döngü sıklığı; harita yoğunluğu ve değişken sıklığı derecelerinin belirlenmesi sonucu yapısal olarak karşılaştırılmalardır. Ayrıca, tanım-spekülasyon puanı; matris uzaklığı ve uzaklık oranı ölçütlerine göre de içerik olarak karşılaştırılmalardır. Politikacıların; Türkiye'nin faydasını en büyüklemek için kendi görüş açılarına göre izlemeleri gereken politikalar saptanmıştır. Bu aşamada, matris uzaklığı ve uzaklık oranı ölçütlerine göre BH'leri birbirine en çok benzeyen Yılmaz ve Çiller' in "rekabetin hakim olduğu piyasa koşullarının oluşturulması" ve "KİT'lerin elden çıkarılması" politik tercihlerinin aynı olduğu; BH'si, çiller' in BH'sine oranla Yılmaz' ınkine daha yakın olan Soysal' m, ilk tercihte onlarla hemfikir olduğu ortaya çıkmıştır.

Although artificial intelligence, is a branch of computer science, it also deals with decision making problems such as decision support system & heuristic programming. If it's wanted to be stated as an input-output model; psychology, mathematical logic, computer science, linguistic and philosophy can be listed as the inputs of the model, where as the following applications can be concerned as the outputs of the model: * Game playing, * Theorem proofs, * Expert systems, * Natural language processing, * Planning and robotics, * Neural networks. People working with artificial intelligence, want not only to support decision making process, but also to create an intelligent machine which gives decisions itself. Activities which require intelligence, can only be achieved by using the following elements; * The symbol models that can put forward the important aspects of the problem, * The operations, that are done on these models in order to create potantiol solutions, * The search, that are used in order to choose a solution between these posibllities. The three elements listed below, are important since they form the base of the Newell and Simon's hypothesis named as "Physical Symbol System Hypothesis" which put forward the two main aspects of artificial intelligence, namely; "knowledge representation" and "search". People think about many alternative strategies when dealing with a problem, and so, problems are solved by choosing one of these potantial solution alternatives which is known as "state space search". Search systems have three main elements; * A data set, which concerns knowledge about the present state and the rules of the situation. * A goal definition, which defines the desired or aimed state. * Operators, which are functions or operations that make the desired changes between states. IX There are three main strategies for state space search; which can be stated as follows; * Data driven & goal driven search, * Depth first & breadth first search, * Heuristic search. In order to solve artificial intelligence based complex problems, the use of the search tecniques are essential but not enough. The knowledge that is required to solve the problem, must be represented in some way. The technique used to represent knowledge must have the property of getting new knowledge by making inferences from the current knowledge use. Knowledge representation in artificial intelligence can be classified as; * Logical representation, * Propositional logic, * First order predicate logic, * Procedural representation, * Network representation, * Semantic networks, * Conceptual networks, * Structural representation, * Frames, * Scripts. Generally, mathematical logic is used for logical representation. Although they are used widely in artificial inteligence's applications -especially in theorem proofs-; expressions whose correctness differ time to time and/or person to person can not be represented by the use of them. In procedural logic, the knowledge required to solve the problem is represented as a rule set. This kind of representation is used especialy in expert systems. Expert systems' are sometimes named as rule based systems since in their knowledge base, there are rules such as "if /then" expressions. The graph theory is widely used in network representation, since concepts are represented as nodes and relations between them are represented as arcs. Semantic networks which is a branch of the network representation, have the ability of giving the required knowledge by using smaller computer memory when compared with predicate logic. But since they do not have heuristic knowledge, it becomes impossible to do heuristic search by using semantic networks. This is also the same for the second branch of the network representation, namely the conceptual networks which are used especially in natural langauage processing. By changing every node in the networks to a complex data structure, the representation tecnique named as structural representation is formed. But although it has a more complex structure, dealing with details only when it's desired, is an important property of frames. The different between frames and scripts is that; scripts have an event sequence, not a static explanation group. Cognitive maps which are used as a knowledge representation tecnique in artificial intelligence, are network representation-like tecniques. A cognitive map, is a representation of the causal beliefs or assertions of an individual or a group. In a cognitive map, the concepts that a person uses are represented as nodes, whereas the causal links between concept are represented as arcs. The real power of cognitive maps is, its graph form representation by what one can see all the concepts and relationships at one glance. The relationships that are represented as arrows have certain values. There are eight possible values of the arrows which go from the cause variable to the effect variable. The most basic values are positive, negative and zero. The positive relationship between A and B means that an increase in A will cause an increase in B; where the negative relationship between A and B means that An increase in A will cause a decrease in B. Zero valued arrows mean that there is no relationship between A and B. The eight possible values which are listed below are sipmly all of the logical combination of positive, negative and zero. The methods used to derive a cognitive map should satisfy 4 quite demanding requirements as follows; * The methods should be un obstrusive. * The derivation should not require advance specification of the concepts as a particular decision maker may use in his cognitive map. * The derived cognitive map should be closely tied to an evaluation theory of decision making; so that it can be used to advice and even criticize the decision maker. * The method for deriving the cognitive map should be valid; which İ3 to say that the cognitive map shoul be an accurate representation of the assertions used by the decision maker. There are 3 methods for the derivation of a cognitive map; * Documentary coding method, * Questionnaire method, * Open ended probing interview. The first method which is deriving a cognitive map from existing documents has the advantage of being both unobstrusive and fully able to employ the concepts used by the decision maker himself. The questionnaire method's advantages are that it allows the aggregation of individual's opinions and forms a big data base. And the third method's advantage is its allowing the researcher to interact actively with the source of the data. After the derivation of the cognitive maps, they can be analysed in two ways: * The structural analysis * The content analyis si In structural analysis, a cognitive map is analysed by the following features; * Frequency of utility/goal/policy/peripheral variables, * Degree of path balance, * Degree to which policy choices are consistent with the maps, * Frequency of the cycles, * Density of the map, * Variable frequency. In structural analsis, a cognitive map is analysed by the following features; * Definition-speculation score, * Goal increasing score, * Matrix distance, * Distance ratio. In this thesis, the cognitive maps of three political elites; * Tansu Çiller, * Mümtaz Soysal and * Mesut Yılmaz, are analysed according to both content and structurel analyis. When the structural analysis features are examined, it is seen that,as the political responsibility increase, * the frequency of the goal variables, * the ratio of peripheral variables to policy variables, * the density of the map, * the variable frequency, increases also. The frequency of cycles is only 1 out of three cognitive map and it can be said that all of three cognitive maps have path balance. In the content analysis, the goal increasing score hasn't been analysed since all the political elites think that only their party's policies can increase the Turkey's utiliy. The results that are prepared according to the other three content analysis measures; are given below; Individual Definition-Speculation Score Çiller 0.71 Soysal 0.923 Yılmaz 0.62 Binary Comparisons Matrix Distance Distance Ratio Çiller-Yılmaz 99 0.0309 Yılmaz-Soysal 102 0.0327 Çiller-Soysal 119 0.0372 Since the definition score is the measure of individual's self-confidence, it can be said that Soysal is the most self- confident person where Yılmaz is the least. Matrix distance and distance ratio scores are the measures of the similarity of Binary analysed cognitive maps. According to the scores given above, the cognitive maps of Çiller and Yılmaz are the most similar cognitive maps with the matrix distance score of 99, whereas the cognitive maps of Çiller and Soysal are the least similar with the matrix distance score of 119. su The cognitive maps of the politicians are analysed according to the simulation tecnique of Nozicka & Borinam & Shapiro (1976) as to find the three politicians' political preferences which are thought to have the power to maximize Turkey's Utility. At the end of this analysis, Çiller and Yılmaz are found to have common political preferences which is an expected result.