©Copyright JASSS

JASSS logo ----

M. Afzal Upal (2005)

Simulating the Emergence of New Religious Movements

Journal of Artificial Societies and Social Simulation vol. 8, no. 1

To cite articles published in the Journal of Artificial Societies and Social Simulation, reference the above information and include paragraph numbers if necessary

Received: 27-Jul-2004 Accepted: 22-Nov-2004 Published: 31-Jan-2005

* Abstract

Not unlike other social sciences, study of religion in general and study of new religious movements (NRMs) in particular, has suffered from a problem of having too many inter-related free variables and a few data points available to constrain their values. This paper suggests cognitively inspired computer modeling as a technique for exploring, refining and testing theories of religion. Although computer simulation has become a relatively accepted technique for studying social theories, it has rarely been used to study religion. To illustrate this point I describe in detail the Agent-based Information Entrepreneur Model (AIM), a computer model of the recently proposed cognitive theory of new religious movements.
Cognitive Science of Religion, Multiagent Systems, Rational Choice Theory, New Religious Movement Emergence

* Background

Contrary to the expectations of a vast majority of early social scientists, religion refuses to go away. Instead the world seems to be more religious than ever. New religious movements have played a crucial role in the revival of religion. Understanding NRMs is important not just for understanding religion (both old and new) but also for learning about culture at large. As Bainbridge (1985) famously said, "cult is culture writ small." Understanding NRM can also tell us about how norms in the larger culture are formed, maintained, and spread. Unfortunately, we know little about the dynamic cognitive and social processes that underlie the emergence, maintenance, and spread of new religious movements. The situation has changed little since 1979 when Bainbridge and Stark lamented that most work on the emergence of new religious movements was 'nearly devoid of theory' (Bainbridge and Stark 1979). This is not just because of the political debates regarding the social value of NRMs that have consumed the attention of many NRM scholars since the inception of the discipline as the study of shin shukyo (new religions) in Japan in the 1960s (Lewis 2004). A fundamental problem preventing the development of a scientific theory of the emergence of new religious movements is that such movements are often inspired by individuals who are not available in large enough numbers to allow controlled experiments for the extraction of statistically reliable patterns in their behavior (Fusfeld 1992; Upal 2005). In such situations, agent-based social simulation (Bainbridge et al. 1994) can be an extremely powerful tool for
This paper argues that an agent-based simulation of NRM theories is a useful tool that has gone largely unexplored (see Bainbridge 1995 and Doran 1998 for exceptions). Agent-based social simulation allows us to (a) precisely measure the prevalence of a belief in a population of agents, and (b) perform controlled experiments where we can change the values of various parameters and look inside the heads of our agents to see how these changes affect their beliefs. I illustrate the benefits of computer modeling for studying religion with the help of the Agent-based Information Entrepreneur Model (AIM) designed to model the cognitive theory of new religious movements (Upal 2005).
The cognitive theory focuses on new religious movements that are inspired by charismatic (Weber 1922) individuals. It posits that NRM founders are information entrepreneurs (IEs) i.e., creators, marketers, and sellers of innovative solutions to perceived social problems. The social problems range from foreign cultural invasion perceived to be a problem by nationalists of various hues to the greed and self-indulgence of the 'Me' generation perceived to be a problem by many in the Moral Majority (Snowball 1991). Information entrepreneurs are utility maximizing rational agents who obtain various social advantages such as reputation enhancement and increased respect from other utility maximizing rational agents who buy their solutions. Next, I briefly review salient aspects of the cognitive theory of NRMs and describe the computer model. This is followed by a description of the experiments that have been done to test the theory and a discussion of the results obtained.

* Cognitive Theory of New Religious Movements (CTNRMS)

The cognitive approach to new religious movements (Upal 2005) is part of a new paradigm for the study of religion that assumes that religion is, "a natural product of aggregated ordinary cognitive processes" (Barrett 2000) and that religious agents are, "rational agents" acting to "maximize their perceived utility" (Iannaccone 1998). CTNRMs builds on Stark and Bainbridge's (1987) pioneering entrepreneurship theory in emphasizing the role of charismatic (Weber 1922) individuals such as Joseph Smith, Sun Myung Moon, Mary Baker Edy, Shabatai Tzvi, and Mirza Ghulam Ahmed who inspire new religious movements. Upal (2005) argues that such leaders belong to a class of people known as information entrepreneurs (IEs). Information entrepreneurs include religious as well as secular socio-political leaders who: The rewards that IEs receive for their activities include: For NRM founders, there is also the additional pleasure of serving God, who they believe is guiding them.
Human culture abounds with ideas?from scripts to schemata, for example?that provide us with causal explanations for a state of the world that we find ourselves in. We acquire these ideas from agents we interact with as well as from embedded cultural institutions (Day 2003; Norman 1993; Simon 1996). Importantly, however, the process of concept acquisition involves more than simply copying everything one hears or reads onto a blank mental slate. It involves understanding and integrating the ideas into one's existing set of beliefs by adjusting the ideas and/or refining one's existing knowledge base. Even though each human being is a unique organism with unique background knowledge and faced with unique circumstances, people living in similar environments have similar enough beliefs that only minor adjustments and/or refinements to the acquired ideas are needed.
In times of rapid social or environmental change (brought about, for instance, by military, religious, political and/or natural upheavals when social controls relax and allow freer market of ideas (Iannaccone 1998)), however, more people become disenchanted with their dominant social myths and feel freer to change their beliefs. This provides business opportunities for information entrepreneurs such as NRM founders. New schools of thought are formed in religion, science, and politics by the efforts of individual information entrepreneurs. In this way, social institutions like religious systems are susceptible to transformation (Whitehouse 1995, 2000). NRM founders are leaders of the new schools of thought. In many cases NRM founders are originators of new ideas. In other cases they are the ones who are better able to integrate previously prevalent ideas into a new cohesive package and sell the package to their peers better than others can.
Social agents are embedded in their environment in more ways than one. Not only does the environment provide them with problems to work on and define boundaries of the solution space they can explore but it also provides them continuous feedback during the solution generation process. Similar to other entrepreneurs, for information entrepreneurs the behavior of the first customers is crucial in determining whether the new venture sinks or swims. Once an information entrepreneur succeeds in attracting a critical number of loyal customers (Marwell and Oliver 1993), his/her teachings attain the role of norms in the new group.
While the cognitive theory of new religious movements has been applied to explain the emergence of NRMs such as the Ahmadiyya Movement of Mirza Ghulam Ahmed (Walter 1918; Lavan 1974; Friedmann 2003) testing it and other NRM theories has proved challenging (Upal 2005). Computer modeling has increasingly emerged as a technique of choice in other social sciences faced with the similar problems. I believe that computer modeling has great potential for clarifying and testing theories of religion. This potential remains largely unexplored despite a few exceptional attempts (Bainbridge 1995; Doran 1998). To illustrate the potential of computer modeling, I have developed an agent-based computer model of the cognitive theory of new religious movements.

* The Agent-based Information Entrepreneur Model (AIM)

The AIM system consists of N agents that need health points to survive. Health of an agent decays with time at the rate of one point per simulation instant or round. When an agent's health level drops below the health-threshold, the agent dies and a new agent is born to take its place. Agents gain points by playing a version of Keynes' beauty contest[1] (Keynes 1936) called guess-the-average-number game in which each agent can select a number between 0 and 100 (Hoffmann 2002). An agent receives a boost to its health based on how close its estimate is to the average of all the numbers selected according to a closeness function fc. The ability to predict the target number is crucial to the survival of an ordinary agent. Agents who are extra-ordinarily successful at solving the problems they face (i.e., they have either earned point above the IE-points-threshold and/or the strength of their strongest rule is above the IE-rule-strength-threshold) become information entrepreneurs at any point during the game after a number of practice-rounds. The current version of the AIM model assumes that IEs stop playing the number guessing game on their becoming information entrepreneurs. This is meant to model the extra-ordinary changes that take place when people decide to dedicate themselves to starting a new movement. Most devote themselves fully to calling others to their mission and have little time for holding ordinary jobs (Lewis 2004). In the current version of AIM, IE agents receive all their points by selling solution(s) to guess-the-average-number problem to ordinary agents. An ordinary agent may decide to accept the solution offered by an IE agent if the receiving agent's existing knowledge is so ineffective in predicting the average number that strength of one of its rules drops below the rule-usefulness-threshold. This is meant to models that people are more susceptible to accepting a new solutions if their current knowledge is inadequate in solving their problems. If an ordinary agent accepts the solution offered to it by an IE agent, it must pay a small amount of health points called homage in return to that IE agent. The homage points are deducted from the health level of the ordinary agent and added to the health level of the IE. Once, an ordinary agent adopts a solution as its active rule to be used for solving the problems, it must pay contributions in the form of a tithe to the IE whose solution it has adopted every time it uses that solution. This is meant to model the rational choice theory axiom that producers of religious products must continually receive payments from their consumers to stay in business (Iannaccone 1998).
Each agent's health value is initialized to 400 and its knowledge-base is initialized to contain three of the following seven rules.
  1. select a number between 1 and 15
  2. select a number between 16 and 30
  3. select a number between 31 and 45
  4. select a number between 46 and 55
  5. select a number between 56 and 70
  6. select a number between 71 and 85
  7. select a number between 86 and 100.
Each rule starts of with an initial strength of 100. At each time step, the rule with the maximum strength is activated for prediction and a random number generated in the range offered by the active rule. Once all ordinary agents have predicted a number, the average is calculated and each agent updates the strength of its active rule based on the distance from the average number. This is intended to capture the view that norms require agreement among a population of agents to come into existence and survive and that once the norms become accepted, they impact individual behavior by rewarding those who behave in accordance with the norms while punishing those who do not (Finnermore and Sikkink 1998; Hoffmann 2002). Each agent increases the strength of its rule by rule-increment if a rule's prediction is within the acceptable range of the average; otherwise the strength is decreased by the same amount. Periodically (i.e., every nth round), an agent may replace one of its non-performing rules (i.e., the rules whose strength falls below the rule-usefulness-threshold) by a random rule or by a rule offered by an information entrepreneur. When an IE's rule is adopted by a critical mass of agents as an active rule, the group acquires the ability to have its own norms. This is simulated by allowing group members to use the average of their own group (as opposed to the average of the entire agent population) as the target number to update their rule strengths, once the number of agents using an agent's rule reaches critical-mass-threshold. When the leader of a group with critical mass dies, the group stays together and the most successful group member (i.e., the member with the most number of health points and rule strength) is selected as the new IE. Members of those groups that have fewer followers disperse into the non-affiliated population upon the death of their IE.
The size of the agent population is maintained by creating a new agent upon the death of an agent. When an ordinary non-affiliated agent dies a new agent with a random set of three rules is created. However, when an affiliated agent dies, the new agent is born with the group rule as its active rule. The current version of AIM is implemented in Java and has been run using JRE version 1.4.2. It can be freely downloaded from http://www.eecs.utoledo.edu/~aupal/iam/ies/

* Experiments and Results

Some of the most persistent controversies in NRM research (Lewis 2004) swirl around questions such as: I designed three experiments to study the behavior of a population which has IEs offering their strongest and weakest rule to other agents. In Experiment I, all IEs offer their best solution to other ordinary agents. In Experiment II, all IEs offer the worst solution in their repertoire of solutions to other agents. In Experiment III, some IEs offer the best solutions while others offer the worst solutions. Hundreds of experiments were run with various parameter values and a set of values was selected (number-of-agents = 100, num-practice-rounds = 10, initial-health-level=400, homage-points = 1, tithe = 0.01, accuracy = 5, critical-mass-threshold =10, IE-health-threshold = 396, IE-rule-threshold = 105, useful-rule-threshold = 100, critical-health-threshold = 100) to run three controlled experiments.

Experiment I

In this set of experiments, all IE agents offered their most successful rule as a solution to the other agents. The results shown in Figures 1-6 are typical of the results obtained. We drew the following conclusions from these trials.
Figure 1. The initial state of an Experiment I at time = 0. Each round black dot corresponds to one non-affiliated agent. The spatial positions of the agents do not mean anything as the current version of RBS does not take into account an agent's spatial position
Figure 2. The state of an Experiment I world at time = 11. Agent 5 (shown by a red star) has emerged as an IE offering Rule 3
Figure 3. The state of an Experiment I world at time = 50. Thirty eight agents have made the rule offered by the IE as their active rule and paid homage and tithe to IE 5
Figure 4. The state of an Experiment I world at time = 453. Two more agents, Agent-100 (shown by a green star) and Agent-101 (shown by a blue star) have emerged as information entrepreneurs
Figure 5. The state of a Experiment I world at time = 500. The new IEs are only able to attract a small number of followers (3 for Agent-100 and 5 for Agent 101)
Figure 6. The state of an Experiment I world at time = 1000. Both Agents 100 and 101 have died and their few followers disbursed back into the non-affiliated group, while Agent 5 is holding onto its 49 followers (by far the largest grouping including the 40 non-affiliated members). A number of new IEs emerged starting around time 700 but none has been able to gather more than a couple of followers

Experiment II

In this experiment we changed the rules of the game so that all IE agents offered their weakest rules to ordinary agents. This small change resulted in a number of changes in the aggregate societal behavior as illustrated by the sample run shown in Figures 7-10.
Figure 7. The state of an Experiment II world at time = 32. Agent 0 (shown by a red star) and Agent 31 (shown by a blue star) both declare their entrepreneurship at time = 4. IE 0 offers Rule 7 and IE 31 offers Rule 6. Agent 0 gets to propagate its knowledge first and has attracted 48 agents by round 32
Figure 8. The state of an Experiment II world at time=44. The strongest-rule IE initially attracts only the non-affiliated agents but starts to attract some of the former followers of the weakest-rule IE 0. as time proceeds
Figure 9. Things remain fairly static until Round 387 when one of the unaffiliated agents Agent-97 (shown by a yellow 4-corner star) declares its weakest-rule entrepreneurship offering Rule 1
Figure 10. An intense battle for followers ensues between the three agents with a number of inter-sect conversions. The simulation ends with no non-affiliated agents remaining and each entrepreneur having established a norm community

Experiment III.

This experiment had a heterogeneous population of IE agents as shown in the sample run shown in Figures 11-14. Some IE agents offered their strongest rule, while the others sold their weakest rule. An agent had a 50% chance of becoming a strongest-rule IE and a 50% chance of becoming a weakest-rule IE.
Figure 11. The state of an Experiment III world at time = 23. Two agents Agent 4 (shown by a red star) and Agent 35 (shown by a blue star) have declared their entrepreneurship in Round 5. IE 4 is a weakest-rule IE offering Rule 1 while IE 35 is a strongest-rule IE offering Rule 3
Figure 12. By Round 51, IE 4 has lost 4 members to IE 35 who also attracts another 26 non-affiliated members
Figure 13. Agent 100 (shown as a yellow star) declares its entrepreneurship in Round 349. It is a weakest rule IE also offering Rule 1. By Round 850, it has attracted only four followers when Agent 121 (shown as a purple star) also declares its entrepreneurship. Agent 121 offers Rule 3. By this time IE 35 has attracted another 30 followers while IE 0 has lost 4 more followers
Figure 14. By the end of the simulation, IE 4 has lost all its followers while IE 35 has gained 16 more followers. Three more non-affiliated agents become strong-rule entrepreneurs all offering Rule 3 but none attracts any followers. Only 18 non-affiliated agents remain

* Discussion

One way to analyze consequences of these experiments for the real world NRMs is to explore the similarities between the resulting experimental worlds and the real world of NRMS. We can easily eliminate the world of Experiment I from contention because we know that in the real world of NRMs the rate of conversion from one movement to another is significantly higher than zero (Lewis 2004). But whether the real world looks like the worlds of Experiment II or the worlds of Experiment III is not clear at this point. This implies that at least some of the NRM founders sell solutions that are not good solutions to the original societal problems. The simulations show that paradoxically some of these solutions become useful solutions once an NRM takes root and is able to create its own norms as the target number to be predicted becomes the number being guessed by the group agents all of whom have the same active rule. In this way the very organization and isolation of the group members from the society at large results in their expectations becoming a self-fulfilling prophecy.

* Conclusions and Future Directions

This paper has advocated computer modeling as a promising method for testing theories of religion given the many problems associated with "ecological noise." The paper illustrates the potential of computer modeling by describing in detail the AIM system developed to explore the recently advanced cognitive theory of new religious movements (Upal 2005).
One of central issues that all social simulation researchers have to wrestle with is selecting the model's level of complexity. If the model is too complex, it loses its value as a model because it is either intractable and/or its results are just as complex to understand as that of the real phenomena that is being modeled. However, if the model is too simple, the effort spent in designing and implementing the computer model is often not worth it because the consequences of a simple model can be readily foreseen without constructing it. While AIM as a model of the cognitive theory of NRMs has several limitations, I believe that it is a good compromise between the complexity and simplicity because it models the most crucial elements of the theory.
Cognitive theory of NRMs postulates that self-deception plays a critical role in both causing the NRM founders to attribute their solutions to the divine and in causing ordinary agents to maintain their belief in the NRMs community norms. Talbott (1995) has argued that it is rational for a utility maximizing agent A to believe in a proposition p, regardless of whether p is actually true or not, if the agent prefers a world in which the proposition Q is true to a world in which Q is false, where Q="A believes that p is true". This can explains perhaps why some NRM leaders discount the possibility that they themselves are creators of the solutions they claim to be of divine origin. CTNRMs argues that this may be because NRM founders prefer to live in world in which they believe that their beloved God provides them with solutions to life's problems to a world in which they do not believe that their God solves their problems. Similarly, once an IE's teachings gain the status of group norms, those who follow them are rewarded while those members of the new society who refuse to follow them are punished. In an environment, in which even minor inadvertent infringements from the normal behavior may be detected by others (Schank 1932; Erikson 1966), it becomes beneficial for an agent to believe Q = "I believe the norm to be beneficial" regardless of whether the agent actually believes the norm to be beneficial or not. Thus self deception also plays a crucial role in maintaining the new group norms.
Future versions of AIM may also involve modifying the game, for instance, to change the target number to be guessed from average to (p * average) where o < p < 1. In such a game, agents that better anticipate the guesses of other agents are likely to be more successful than agents who do not take anticipate the guesses of other agents. This will allow us to study the impact that difference in the cognitive processing levels has on the emergence and maintenance of NRMs.
Plans are also in the works for further experiments to study how allowing agents to have relationships with one another and allowing an agent's relationships to affect its beliefs impacts the structure of the social groupings that emerge. I believe that there is a lot more work that can be done within the AIM simulation framework to understand the dynamics of new religious movements.

* Acknowledgements

This paper has benefited immensely from my discussions with the members of the 'I-75 Cognition and Culture Group'; Ryan Tweney and Jason Slone. In particular, Jason provided crucial feedback and encouragement without which this paper would not have been possible. Rashid Erfani designed earlier versions of the program that lead to the AIM System. I am also grateful to the anonymous reviewers for their helpful suggestions.

* Notes

1Unlike Keynes' original beauty contest and number guessing games based on it, such as Nagel (1995), our version is not a winner-take-all game. Also unlike Nagel's number guessing game which awards all winnings to the player whose guess is closest to (p × average) with 0 < p < 1 to distinguish between various levels of reasoning capabilities, the current version of our game (with p=1) is not designed to do that. In our version the Nash equilibrium strategy is to guess 50.

* References

BAINBRIDGE W S (1985) Cultural Genetics, Religious Movements, Stark, R. (ed) New York: Paragon.

BAINBRIDGE W S (1995) Neural Network Models of Religious Belief , Sociological Perspectives, 38: 483-495.

BAINBRIDGE W S, Brent E, Carley K, Heise D, Macy M, Markovsky B, Skvoretz J (1994) Artificial Social Intelligence, Annual Review of Sociology 20: 407-436.

BAINGBRIDGE W S, Stark R, (1979) Cult Formation: Three Compatible Models, Sociological Analysis, 40: 285-295.

BARRETT J L (2000) Exploring the natural foundation of religion, Trends in Cognitive Science, 4 (1): 29-34.

DAY M (2003) Religion, Off-line Cognition, and the Virtues of Embeddedness. Paper delivered at the American Academy of Religion Conference, Atlanta, GA: November 23.

DORAN J (1998) Simulating collective misbelief, Journal of Artificial Societies and Social Simulation, 1(1).

ERIKSON K T (1966) Wayward Puritans: A Study in the Sociology of Deviance. New York: John Wiley & Sons.

FINNERMORE M and Sikkink K (1998) International Norm Dynamics and Political Change. International Organization 50 (4): 887-918.

FRIEDMANN Y (2003) Prophecy Continuous: Aspects of Ahmadi Religious Thought and Its Medieval Background. New Delhi: Oxford University Press.

FUSFELD, W. (1992) Review of Prophecy Continuous, International Journal of Middle East Studies, 24 (2) 347-348, 1992.

GOPNIK A (1998) Explanation as orgasm, Minds and Machines, 8:101-118.

HOFFMANN M (2002) Entrepreneurs and the Emergence and Evolution of Social Norms. in Proceeding of Agent-Based Simulation 3 Conference, ed. Christoph Urban, 32-37. Ghent, Belgium: SCS-Europe.

IANNACCONE L (1998) Introduction to Economics of Religion, Journal of Economic Literature, 36: 1465-1496.

KEYNES K M, The General Theory of Employment, Interest, and Money. London: Macmillan.

LAVAN S (1974) The Ahmadiyah Movement: A History and Perspective. New Delhi: Manohar Book Service.

LEIK R. K. and B. F. Meeker 1995. Computer Simulation for Exploring Theories: Models of Interpersonal Co-operation and Competition. Sociological Perspectives, 38: 463-482.

LEWIS J R (2004) The Oxford Handbook of New Religious Movements, Newbury Park, CA: Sage.

MARWELL M and Oliver P (1993) The Critical Mass in Collective Action: Studies in Rationality and Social Change, Cambridge, MA: Cambridge University Press.

NAGEL R (1995) Unraveling in Guessing Games: An Experimental Study, American Economic Review, 85 (5): 1313-1326.

NORMAN D (1993) Things that make us smart: Defending human attributes in the age of the machine. New York: Addison-Wesley Publishing Book.

SCHANK, R C (1986) Explanation Patterns : Understanding Mechanically and Creatively, Hillsdale, NJ: Lawrence Erlbaum Associates.

SCHANK R L (1932) A study of community and its group institutions conceived of as behavior of individuals. Psychological Monographs 43 (2): 1- 133.

SIMON H (1996) The Sciences of the Artificial. Cambridge, MA: MIT Press.

SNOWBALL D (1991) Continuity and Change in the Rhetoric of the Moral Majority, New York: Praeger.

STARK R. and Bainbridge W S (1987) A Theory of Religion. New York: Peter Lang.

TALBOTT W J (1995) Intentional self-deception in a single coherent self, Philosophy and Phenomenological Research, LV(1) 27-74.

UPAL, M A (2005) Towards cognitive science of new religious movements, forthcoming in Journal of Cognition and Culture.

WALTER H A 1918. The Ahmadiya Movement. London: Oxford University Press.

WEBER M (1993, first published in 1922) The Sociology of Religion, Boston: Beacon Press.

WHITEHOUSE H (1995) Inside the Cult: Religious Innovation and Transmission in Papua New Guinea. Oxford: Oxford University Press.

WHITEHOUSE H (2000) Arguments and Icons: Divergent Modes of Religiosity. Oxford: Oxford University Press.


ButtonReturn to Contents of this issue

© Copyright Journal of Artificial Societies and Social Simulation, [2005]