Gallery for Evolutionary Computation and Artificial Intelligence Researches: Where Do We Come from and Where Shall We Go

Abstract

D’où venons-nous? Que sommes-nous? Où allons-nous? (Where do we come from? What are we? Where are we going?) is the title of a famous painting by Paul Gauguin created in his Tahiti days. The title raises very important but hard questions of our life. Contrary to the heavy questions, I started GEAR research group in 1990 in a very easy way, when he moved from a computer system engineer to a professor of Graduate School of Systems Management at Tsukuba University. Since then, as described in this book in total, we are concerning on the research and development for evolutionary computation and artificial intelligence researches. The name GEAR comes from the acronym. This short paper describes the history of experimental methods in social and management sciences referring to our experience and discusses the importance of agent-based approaches as interdisciplinary and integrated fields on complex systems. Yes, I and my colleagues are also facing with very heavy questions: Can we or should we simulate them? Where are agent-based approaches going? The paper concludes with the statements on where shall we go in the complex world in the near future.

1.1 The Shape of Jazz to Come: Introduction

Ornette Coleman, a jazz alto sax player, started “free jazz” activities in the end of 1950s. On those days, they thought that Ornette played only nonsense without any jazz theories and that audiences could not welcome the music. However, the jazz history says Ornette was an excellent player, who was able to vividly play Charlie Parker’s adlib music as if he was alive. Also, his activities have gradually become popular in the jazz literature (Segell 2005).

Same as the history of jazz, it is often the case that they could not understand the essential meanings in some scientific researches such as physics or mathematics: the birth of quantum mechanics, non-Euclidean geometry, and recent econo-physics are typical examples of such fields. First, they usually think bland new theories and nonsense ideas; they neglect them, and then, however, they gradually accept new ones because of the appearance of new evidences. I believe that experiment-based approaches in management and social sciences (EBMS) should follow such traditions of new scientific and practical activities. In this short article, I would like to draw the shape of EBMS to come referring to the titles of (fairly) popular (jazz) music.

1.2 A Child Is Born: When Experimental Methods Started

“A Child is Born” is composed and played by Thad Jones, a jazz trumpeter.

In the literature, the experimental approach to management or social sciences started from the pioneering work of Cyert and March in the 1960s (Cyert and March 1963). Also, the garbage can model by Cohen et al. (1972) is well-known. The existence of such early works is very remarkable, as both Fortran and Lisp, traditional programming languages for numerical and symbolic computation, were developed in 1960. However, because of the limits of computer performance and the immaturity of programming techniques, for researchers in noncomputer science-related fields, especially social scientists, the approach has not been successful in those days.

The second leap of EBMS was found in the early 1990s. In those days, researchers in distributed artificial intelligence fields have become interested in agent-based approaches. Agent-based modeling (ABM) or agent-based simulation (ABS) was a new modeling paradigm in those days (Axelrod 1994; Epstein 2006). It focuses from global phenomena to individuals in the model and tries to observe how individuals with individual characteristics or “agents” will behave as a group. The strength of ABM/ABS is that it stands between the case studies and mathematical models. It enables us to validate social theories by executing programs, along with description of the subject and strict theoretical development.

Interestingly, the movement of agent-based approach to social simulation or experimental management science spontaneously emerged worldwide (Terano 2007). In European region, they started SIMSOC (SIMulating SOCiety) meetings, which are followed by the activities of ESSA (European Social Simulation Association) and JASSS (J. Artificial Societies and Social Simulation). In the North American region, COT (Computational Organization Theory) workshops were started at AAAI and Informs Conferences, and then CASOS (Computational Analysis of Social and Organizational Systems), CMOT (Computational and Mathematical Organization Theory) Journal, NAACSOS (North American Association for Computational Social and Organization Sciences), and then CSSSA (Computational Social Science Society of the Americas) activities followed.

In Japan, we organized PATRA (Poly-Agent Theory and ReseArch) group and then continue to have the series of AESCS (Agent-Based Approaches in Economic and Social Complex Systems) workshops hosted by PAAA (Pacific-Asian Association for Agent-based Approach in Social Systems Sciences).

In summary, therefore, we have already had a long history on EMBS.

1.3 Now or Never: Why Not Now

“Now or Never” is composed and played by Hiromi Uehara, a jazz pianist.

Now, we must develop, extend, and then establish EBMS research in front of a large number of audiences because of the following two reasons:

• 1. Recent rapid progress of computer and network technologies makes us possible to easily implement computer-based simulation models. Such models help us to carry out EBMS with both machine agents and human subjects. Also, using such models, we are able to operationalize the concepts and ways of thinking of traditional management sciences. By the word operationalize, we mean that (i) social and organizational systems are observed by human experiments and computer simulations, and (ii) with both machine- and human-readable documentations, they are comprehensively and consistently understood for human experts and students related to management sciences.

• 2. Because of the recent crises of economic conditions worldwide and the tragedy of the great earthquake in Japan, we must deeply understand the mechanisms of human societies. We must develop the new principles of design and implementation of societies. Contrary to physical sciences domains, there are no first principles in management science domains. Therefore, the experimental approach is indispensable to uncover the secrets of human societies.

1.4 Pick Up the Pieces: What Are Elements

“Pick up the Pieces” is mainly played by the Candy Dulfer’s fusion music group. She is both a singer and a sax player.

To build a new EMBS architecture, we have already had various pieces or tools and techniques in our laboratory. Some of them are listed in the other papers in this book. The most important ideas of agent-based modeling in EBMS are first, in agent-based modeling, micro-macro links between agent interactions and environmental conditions shown in Fig. 1.1, and second, the architecture to uncover the interactions of micro-, mezzo-, and macroscopic levels among agents shown in Fig. 1.2.

Fig. 1.1

Principle of agent-based modeling

Fig. 1.2

Structures and formulation of management problems

In Fig. 1.1, an agent mean a model of human, firm, or objects, as a software component, which is equipped with internal states, decision rules, and information exchange mechanisms. As results of microlevel interactions of agents, macro level observable phenomena emerge. Furthermore, as each agent is able to observe such macro level phenomena, it might change its decisions and behaviors. Then complex micro-macro links among them may occur. The phenomena are observed in economic behaviors, social network behaviors, and group decisions in recent complex real-world problems. In ABM/ABS, behaviors and statuses of individual agents are coded into programs by researchers. They also implement information and analytical systems in the environment, so the model itself may be very simple. Even when the number or variety of agents increases, the complexity of simulation descriptions itself will not increase very much. (Axelrod 1997) has emphasizes that the goal of agent-based modeling is to enrich our understanding of fundamental processes that may appear in a variety of applications.

In Fig. 1.2, we introduce a mezzo-scopic structure between the microscopic (members/customers) and the macroscopic (market/industry) level. The problems on the business processes and organizational administration have such difficulties as (a) the problems have complexity with numerous factors in hierarchical structures, and (b) each structural behavior strongly depends on the member’s awareness and decisions. Such complex systems have been often described from the micro-macro loop viewpoint, because the business/organizational problems exist in the mezzo-scopic level in which they have no enough scale differences to neglect their uniqueness nor heterogeneity of their customers/employees in the firm. On the other hand, although recent researches on service sciences and/or the econo-physics adopt the outcomes from experimental economics or behavioral economics, they tend to only explain macro level phenomena from the microlevel customers or investors as the homogeneous set of agents or particles.

Figure 1.2 also illustrates these difficulties from the viewpoints of the interactions between micro-, mezzo-, and macroscopic levels. Arrow “A” indicates that the microlevel (members/customers) numerous factors affect the mezzo-level (corporation/organization) states. Arrow “B” shows the mezzo-level influence on the microlevel actors’ awareness and decisions.

Introducing both the diversity of microlevel agent’s awareness/decisions without off-scaling and an intermediate level structure enables to explore the emergence of organizational deviation and kaizen in corporation management. Actual business and administrative processes include both “A” and “B” inter-level interactions. The existence of these two interaction levels bring the low reproducibility of the business problem, we mentioned ahead that single experiment is not effective to explore the problems. Therefore, we need to apply appropriate experiment-based approaches to each “A” and “B.” We believe that, at first, we present the bottom-up simulation with the orthogonal design of experiments as the approach for “A” and that, then next, the combination methodology of gaming/narrative approach and the orthogonal design is also presented for “B.”

Our current pieces for EBMS include (i) agent-based simulation techniques to explore vast parameter spaces with evolutionary algorithms and grid computers (Yang et al. 2012), (ii) the combination of the organizational bottom-up simulations and the orthogonal designs of experiments, (iii) a new experimental method to measure the awareness via the integration of business games and the manga/narrative business cases, (iv) visualizing techniques for human decisions and behaviors of business processes, (v) conjoint analysis techniques with personae and/or organizational profiles, (vi) abstraction techniques of the agents’ semantic networks, and (vii) virtual scenarios and case set generation techniques based on the concept of design of experiments. About the detailed explanations on these pieces, please refer to the papers elsewhere.

1.5 Place to Be: Where Should We Focus On

“Place to be” is composed and played by Hiromi Uehara, a jazz pianist.

Using the pieces, we are attacking several critical problems on social, organizational, and/or economic fields. The recent lists are found in our GEAR website (GEAR 2018). They are categorized into the following items:

• 1. Research and development of advanced knowledge systems, which include data mining, marketing, education, social networking, recommendation, and manufacturing task domain problems

• 2. Application of agent-based social system sciences, which include organization, business, history, education, and financial task domains

• 3. Integration of gaming and case methods, which include marketing, business competition, finance, manga cases, human/computer participating gaming task domains

• 4. Theories for EBMS, which include double-structured networks, chaos controls, behavioral finances, and games

As found in information in GEAR website, we have already published over 1000 articles in both English and Japanese languages with about 320 coauthors. The network structure of the coauthor network is shown in Fig. 1.3, which shows, of course, the scale-free property in complex network theory.

Fig. 1.3

Coauthor network of articles in GEAR research group in 2018

These task domains are on the boundaries of traditional academic fields such as economics, organizational sciences, statistical physics, operations research, artificial intelligence, computer science, and system science. To address EBMS, we must focus not only on the principled approaches of traditional experimental methods discussed elsewhere but also agent technologies, which are characterized by their internal states, problem-solving and decision-making functionalities, and interaction capabilities.

1.6 Act Your Age: Which Way to Go

“Act Your Age” is composed and played by Gordon Goodwin with his Big Phat Band. He is a jazz pianist and a sax player.

To act our age in EBMS fields properly, I would like to emphasize the following three points:

• 1. Architecture (Hamano 2008) and code (Lessig 2006): The term architecture usually means a building, a typical artificial object. However, some of artificial systems we have built often show emergent properties. The Internet and social network sites are typical examples. In these artificial systems, they often become autonomous and out of control. The term code means such implicit rules determined by laws, cultures, and customs. They determine our both conscious and unconscious behaviors and/or beliefs. The concepts of architecture and code are critical to develop EBMS, because our experiments in social systems are deeply affected by the concepts.

• 2. Control and harness (Axelrod and Cohen 2000): About complex adaptive systems, the book states that they are not controllable and they should be harnessed. By harness, they mean to deal with things by natural forces; thus, compared with control, harness is a very calm concept. To harness (social and organizational) systems, Axelrod and Cohen emphasize the principle of evolution of life: copying, recombination, and selection. Their statements are very conformable to EBMS, because our experimental strategies heavily depend on evolutionary computation techniques.

• 3. Body points for acupuncture and moxibustion in oriental medicine: In oriental medicine, acupuncture points are important concepts to transfer therapy knowledge to others. Without the concepts, they are hard to make therapies of acupuncture and moxibustion, because the treatment itself requires very tacit knowledge and experience. They say the concepts of body points were invented once upon a time. To make clear the name and place of body points, the treatment techniques are considered to be distributed. Our pieces of EBMS have similar properties. Using our pieces, we are able to translate and transmit the results of EBMS with clear explanations to ordinary people in every field.

1.7 Adios Nonino: Concluding Remarks

“Adios Nonino” is composed and played by Astor Piazzolla, a Latin musician. The tune was composed, when the father of Astor passed away in order to pray for his death. The tune is, however, very beautiful with little feeling of the sadness on his death.

In this short paper, I have discussed the basic principles and key ideas of EBMS. EBMS is a just started new field of both scientific and practical activities. I believe, however, EBMS will be a major field in the near future to deal with real-world problems in our societies. Interestingly, Segell emphasizes the importance of the network of herds in a new community in the music world in his book (Segell 2005). To make progress in EBMS, I would like to ask a favor for your participation with your herds.

As take-home bring message, I would like to conclude the paper in the following statements:

• Agent simulation is a lie that helps us see reality

  (Original: Art is a lie that helps us see reality by Pablo Picasso).

• Something may be obvious once you know agent simulation

  (Original: Everything is obvious once you know the answers from the book title of Duncan J. Watts).