Social Learning with Case-Based Decisions - Research Proposal Example

Social Learning with Case-Based Decisions Introduction From an evolutionary perspective, both individual and social learning can be viewed as forms of phenotypic plasticity. Both modes of learning are developmental processes that cause organisms to acquire different behaviors in different environments. Phenotypic plasticity may be adaptive in temporally or spatially varying environments if the use of environmental cues enables organisms to acquire behavior that is adaptive in each local habitat. For example, by sampling novel foods and learning to avoid noxious food types, a cosmopolitan species like the rat can acquire an appropriate diet in a wide range of environments. Mechanisms of phenotypic plasticity may also have fitness costs. By sampling novel foods, the rat incurs risks that could be avoided by an animal with rigid genetically specified food preferences (Wright, 1995). The ways in which individual learning and social learning allow organisms to adapt to different environments are, however, quite different. Behavioral variants acquired by individual learning are not transmitted from one generation to the next. This means that each individual's behavior develops independently based on the interaction of genetically inherited learning mechanisms and the local environment. Generic variation underlying learning mechanisms may evolve, but the behavioral variants acquired by learning do not. Individual learning is adaptive if it bestows some advantage on the individual. In contrast, behaviors acquired by the imitative and observational forms of social learning are transmitted from one individual to another and thus from one generation to the next. From an evolutionary biologist's perspective social learning is interesting because it mixes aspects of a system of inheritance with aspects of ordinary phenotypic flexibility, creating a system for the inheritance of acquired variation. To understand the conditions under which social learning is adaptive we must understand how individual learning and social learning interact to determine the evolutionary dynamics of the behavioral variants themselves as well as the genes that underlie learning processes. The evolutionary properties of the inheritance of acquired variation have received relatively little theoretical attention. This inattention may be due to the fact that evolutionary biologists have supposed that the inheritance of acquired variation is rare in nature, essentially restricted to human culture and a few unusual animal systems, such as the songs of some birds. Those biologists who have imagined that social learning is common in animals besides humans have not always taken proper account of the difficulty of demonstrating true imitation in the face of several processes that can mimic its effects. With a few exceptions recent theoretical work on cultural transmission has concentrated on explaining human culture rather than on the more general properties of social learning (Blonski, 1999). Under what circumstances should natural selection favor a growth of reliance on social learning at the expense of individual learning? The answer to this question is important because it seems likely that social learning originally evolved in species with extensive individual learning abilities. Our focus on the adaptive value of social learning does not imply that selection is the only important evolutionary process, or that all behavior is adaptive. We do believe, however, that understanding the conditions under which social learning is adaptive is an important first step in understanding its evolution and the conditions under which one would expect to find social learning in nature. At first glance, it may seem that social learning will always be the superior form of phenotypic plasticity. Acquiring adaptive behavior by conditioning and other forms of individual learning is often an inefficient process. Learning trials divert time and energy from other fitness-enhancing activities, they may entail serious risks, and there may be substantial chance of not acquiring locally adaptive behavior. It thus seems much more efficient to acquire behaviors by social learning. Studies of humans suggest that social learning can be both rapid and accurate. It is plausible that by simply copying the behavior of others, individuals can acquire locally adaptive behaviors without incurring the costs associated with individual learning. This argument is problematical, however. It certainly makes sense to imitate others if the most common behavior among available models is adaptive in the local environment. The problem is that as individuals come increasingly to rely on social learning, models exhibiting locally adaptive behaviors might become uncommon. To see that this is the case, consider the very simple example in which there are two kinds of individuals in a population-- learners who acquire their behavior by a process of individual learning that results in adaptive behavior, and imitators who depend completely on imitation. As long as imitators are rare, they are likely to copy the adaptive behavior of learners. Assuming that imitation is less costly than individual learning, imitation will be more adaptive. However, as imitators become more common, they are more and more likely to acquire their behavior by copying another imitator, who may have also copied an imitator and so on. In a variable environment, the most common behavior may not be the most adaptive behavior, and individual learning may be more adaptive than imitation (Laland, 1996). The reason that elementary general models are useful, despite their simplicity and unrealism, is that even the simplest evolutionary processes are hard to understand. Thus, simple models serve as an essential supplement to intuition, which is often misleading. In the case at hand, several quantitative variables, such as accuracy of individual learning, costs of achieving a given level of accuracy, and patterns of environmental variation, interact to affect the mixture of social and individual learning that selection would favor. Furthermore, the optimal mix of social and individual learning is affected by population-level properties of social learning; because behaviors can be spread from individual to individual by social learning, long-run outcomes over many generations are relevant to the problem. It is not trivial to keep all these interacting parts of the problem straight. Simple models can serve as a check on less formal methods of deductive reasoning, as a basis for constructing more realistic models, and as an unambiguous standard of comparison for purposes of discussion. Biological Development and Social Learning The process of biological evolution is based upon a complex function of mutation and genetic recombination under the influence of environmental selection. This phylogenetic process displays two modes --adaptive specialization and adaptive generalization. The former acts to improve the adaptation of stereotyped organism behavior. The latter acts to improve the adaptability of the organism. The one may narrow the operational environment of the species while the other tends to broaden it. Both modes of adaptation are manifestations of phylogenesis, and adaptive generalization usually opens the way for a whole new round of adaptive specializations (Schlag, & Pollock, (1999). The process of phylogenesis displays the characteristics of a learning system. It describes the process through which the behavior of a biotype becomes transformed by virtue of the biological population's internal capacity to generate new behavioral ideas through interaction with the environment. This learning process exhibits both the capacity for behavioral refinement and behavioral innovations. The innovative transformations are of special interest to us because they are manifestations of a learning system performing as a developmental system. We note that the adaptive generalization acts to increase the complexity and improve the rational organization of organisms and this is the morphological counterpart to the development of more adaptable behavior. Adaptability is fostered by a series of improvements in environmental tolerance, organism mobility, and discriminatory capacities of the organism. The latter, as manifest in the development of the nervous system and the brain, has been especially important in the evolutionary history of adaptable behavior. We observe that the phylogenetic process, as conventionally described, falls short of internalizing all of the creative elements of selfreorganization or behavioral reprogramming. However, we can see that, at the level of the biosystem as a whole, many of the environmental changes that form the shaping edge of the creative dialogue are themselves a product of biological evolution. We can visualize them as components in an internal feedback response that makes of the evolutionary process an endogenous learning system (Wright, 1995) This aspect of the process cannot be satisfactorily detailed because it has not been adequately attended to by the life scientists. However, the nature of the process is sufficiently well known to characterize it as a stochastic process that does not support positive prediction. It turns out, therefore, that it does not provide a fully adequate model for describing social change, nor does it form an adequate base for the conduct of social prediction and planning. Nevertheless, it may illustrate some aspects of this process and, as a base for comparison, serve to highlight some of the unique characteristics of the social process. Social Development and Social Learning We Have Come To The Point that we recognize that economic and social development implies changes in modes of behavior. This, in turn, implies that development is essentially a learning process. Learning processes can be conceived as taking the form of programmed learning or creative learning. In the interest of investigating model concepts that can be characterized as creative learning, we have investigated the stochastic learning model represented by the modern synthetic theory of biological evolution. That review makes plain that there is another creative learning model that we characterize as social learning. In order to give this model a more complete articulation, let us examine the principal features of the process of social learning and the way in which it has emerged (Hunt, 1962). Consider, first, the way in which the process of social learning came into being. This process has both an individual aspect and a group aspect. The Learning System at the Level of the Organism The most striking thing about evolutionary history is the fact that the operation of phylogenesis in its generalizing mode has created improvements in organism adaptability until it has generated learning organisms. This gives special point to one of the large-scale features of evolution already emphasized. We observed earlier that instead of modifying the genetic base of behavior by pruning and reshaping a stereotyped pattern, adaptive generalization, by promoting adaptability, provides the organism with the power to modify its own behavior during its life cycle without sole recourse to another round of selective transformation of the genotype. We observed that this power was particularly served by the development of a nervous system and brain that permits discriminating behavioral responses, and that this power has reached its highest manifestation in man. These traits persisted and developed because, once invented, they enormously enhanced the survival characteristics of the organisms and populations so favored and became inscribed in the genetic base of the species. Expressed another way, the behavioral reprogramming of organism behavior that is phylogenesis gradually evolved a program (genotype) that provides the organism with the power to reprogram itself--to act as a true learning system at the organism level. Phylogenesis operating as a learning system produced a learning subsystem--the learning organism--that operates at a different level and by more direct means. The development of this new biological capacity did not create a way for the new adaptive behavior (acquired by the organism in the course of its lifelong encounter with the environment) to be passed on to its progeny through the genetic material. Thus, during the earlier phases of the evolution of this discriminating process, there was no means for accumulating acquired behavior and each organism had to "rediscover" the world for itself (Estes, 1984). As these learning powers became enhanced in later species, we could see the emergence of a new learning dynamic. As organisms acquired the power to perceive their environments, interpret those perceptions, and generate a feedback response, they found that their environments included other members of their own species engaged in a learning response to the environment. In time, the power of mental abstraction arrived at the point where the behavioral responses of others could be perceived and interpreted in a way that permitted behavioral mimicry. This opened the door to the accumulation of acquired behavior in a population not subject to the mortality constraints of organisms. In short, the learning capacity of the organism became socialized into a more general learning system that operates once again at the level of the population rather than the individual. But the process at work here obviously exhibits a different dynamic form than that of phylogenesis. The advantages of shared learning for behavioral adaptation and survival assured the reinforcement and development of this mode of learning. It reached its peak in man, in whom the power of abstraction is raised to a level supporting formal symbolic modes of communication or sharing of acquired experience. In the human species the learning organism reaches the point where learning becomes largely socialized because the dominant aspect of the individual organism's learning environment is the presence of and the sharing with other human learning organisms (Scharfstein, & Stein, 1990). The development of socialized learning opens the way for an important change in the way learning systems operate. The phylogenetic process always operated through genetic differentiation. Under the influence of variations in environmental ranges genetic pools became progressively differentiated into subsets (the process of speciation). In a few instances the rudiments of genetic diffusion were present. This is a process that brings about the transformation of biological systems by the diffusion of genes between species through introgression or hybridization. Where this occurs it leads to a convergence of species characteristics rather than the divergent characteristics of adaptive radiation. This mechanism has played a greater role in the evolution of plants than animals, but its role for the most part has been extremely limited in both. It is obvious however, that the process of socialized learning places great reliance upon the process of information diffusion with its attendant convergent qualities. Is Social Learning an Adaptive Specialization? Learning, like other behavioral or structural traits, may vary in its usefulness according to the particular environmental problems an animal faces. In psychology, Rozin and Kalat (1971) were the first to propose explicitly that some learning abilities could be seen as adaptive specializations molded by natural selection to cope with particular ecological demands. Three major assumptions underlie this view: (1) learning is not a single, general, set of rules for the modification of behavior, but an assemblage of discrete abilities that may be oriented in different directions in different contexts; (2) because different species face different ecological contexts, learning abilities can be expected to vary across species; and (3) the origin of ecologically correlated learning differences is divergent natural selection. Adaptive specialization is part of a wider, ecological, program for the study of learning. Several logical and mathematical models for the evolution of learning also fall within this ecological program, as does recent comparative work on spatial memory in birds and mammals and flower exploitation skills in hymenoptera. Like any other learning ability, social learning can also be seen as an adaptive specialization to particular environmental demands. The degree of social flocking will be found to show a positive correlation with the ability to learn avoidance and other responses through empathic processes. In all three cases, interspecific differences in social learning appear to support the ecological view. Mandrills are quicker at social learning of an avoidance response than are baboons, which are, in turn, more rapid than are vervet monkeys; these differences are in the same direction as species differences in gregariousness. Opportunistic great tits learn avoidance discrimination more easily in social conditions than conservative greenfinches. Great tits and blackbirds, more opportunistic than their respective congenerics the marsh tit and songthrush, also socially learn a new food searching behavior more rapidly. At first glance, the comparative literature therefore seems to suggest that social learning is an adaptive specialization to opportunist and gregarious lifestyles. Social Learning Theory Social learning theory is an objectivist theory that emphasizes the application of universal (scientific) principles. This perspective emphasizes that the practitioner need not apply (social learning) behavioral theory differentially, but can achieve effective interventions if the theory's practice assumptions are sensitively and accurately undertaken. Other behaviorists, such as those espousing cognitive-behavioral theory, may emphasize other learning mechanisms (Blechman, 1984) and may contend that the behavioral approach is not as value-neutral as is presented. The application of the principles of social learning theory to the field of social work is almost three decades old. The earliest behavioral article authored by a social worker appeared in 1965, and the earliest book was published in 1967. Since that time the field of behavioral social work has expanded enormously. Social learning theory is now a preferred orientation for a large minority of social workers, and the approach is well represented in professional curricula, textbooks, and journals. The social learning theory foundation of behavioral social work has produced an approach to understanding human behavioral development and diversity of expression that is characterized by an empirical approach to re- search, largely (although not exclusively) quantitative in orientations and committed to understanding objective relationships between human beings and their psychosocial environments. This empirical orientation has proven to be professionally productive: Well over 50% of the controlled outcome studies with positive results that have appeared in the social work literature have been based on social learning theory (Scharfstein, and Stein, 1990). That the behavioral approach is represented in professional training and practice and is relatively well documented as efficacious for a wide variety of problems addressed by social workers seems reasonably established. The Role of Biological Factors The biological differences that imperfectly distinguish various cultural and ethnic groups obviously arise from genetic factors, but a given person's genetic endowment is itself the function of many generations of natural shaping by the environments in which human beings have lived. Indeed, recent developments in the field of sociobiology suggest that aspects of behavior, not just physical features, may be genetically mediated. Extensions of the influence of biological factors are also being made into the realm of those behaviors giving rise to what we call ethnicity. Although it is clear that nurture has a role as well as nature in accounting for cultural diversity, from the perspective of social learning theory, most cultural differences are not intrinsic to people but to the different environments in which they live and from which they learn. 1. People's actions are lawful and shaped by personal history and biology; all persons, therefore, deserve equal respect regardless of behavior. This is true regardless of race, sex, age, ethnicity, class, religious or sexual preferences, educational level or disability. 2. Cultural practices emphasizing mutual respect and cooperation generally produce greater overall satisfaction and reinforcement than do those based on oppression, competition or coercion; therefore acting for the common welfare is generally valuable (Scharfstein, & Stein, 1990). Although clearly derived from social learning theory, these elements of Mattaini's Behavior Code would seem highly relevant to social work practice in general, and to practice with culturally diverse client groups in particular. Definitions and Mechanisms of Behavior Change One definition of behavioral social work is the following: Behavioral social work is the informed use, by professional social workers, of interventive techniques based upon empirically-derived learning theories that include but are not limited to, operant conditioning, respondent conditioning, and observational learning. Behavioral social workers may or may not subscribe to the philosophy of behaviorism. The foundations of social learning theory were originally developed outside the field of social work, primarily through the work of experimental psychologists, but the last three decades have seen ample evidence that the principles of learning established in the experimental laboratory with other organisms are isomorphic with those operating with human beings . That humans present vastly more complex elaborations of behavior is undeniable, but it seems clear that to a very large extent our comportment develops through the three learning mechanisms described in the above definition. Respondent conditioning is the most fundamental learning mechanism shared by humans and other animals. Shown to occur in all animal species ever tested, including single-cell organisms, even individual nerve and muscle cells are capable of respondent learning. Pavlov and his associates demonstrated how a neutral stimulus, if presented one or more times immediately prior to the occurrence of an event (unconditioned stimulus) that produced an innate reflexive response (unconditioned response), the previously neutral stimulus could come to elicit a similar response. After such conditioning trials the neutral stimulus that now elicits such responses is called a conditioned stimulus (CS) and the largely involuntary response to the CS is called a conditioned response. The types of bodily responses involved in such conditioning processes are called respondent behaviors, and the learning process is called respondent conditioning. It appears that significant components of human sexual behavior, emotional reactions, and even glandular secretions can be modifiable through the processes of respondent conditioning. The capacity for such learning in human is present at birth. The second major form of learning addressed by social learning theory (actually the dominant one) is called operant conditioning, which examines how the consequences that have followed a behavior in the past come to influence its present occurrence. This is a completely different mechanism than respondent conditioning, which focuses on how stimuli that occur before a behavior come to elicit relatively simple reflexive acts. In the operant model developed by Skinner and others, the consequences that follow a behavior can take four major forms: something good is presented (positive reinforcement), something bad is taken away (negative reinforcement), something bad is presented (positive punishment), or something good is taken away (negative punishment). Reinforcement (both positive and negative) tends to strengthen behavior, whereas punishment tends to weaken it. Building upon these conceptually simple operations, a vast edifice of operant theory has been established, strongly supported by empirical research, which addresses most of those human activities we call voluntary behavior (the term operant is derived from the fact that our behavior can be said to operate upon our psychosocial environment). Virtually all animal species have been shown to acquire new behavior or to modify existing repertoires through operant learning processes, a capacity that seems to be present in humans even at birth (Laland, 1996). Learning via imitation of others (also called modeling) is a third important mechanism of learning. The capacity for imitation appears to be present in humans at birth and has been demonstrated in other animals. The development and maintenance of complex imitative repertoires seems strongly mediated by operant factors and is held to account for much of an individual's acquisition of culturally and gender-specific behaviors. The innate capacity for imitation is itself strongly affected by the consequences of one's own modeled behavior. A child who imitates a peer and obtains reinforcement for this is simultaneously strengthening two repertoires: the particular behavior that was imitated, and the generic likelihood to model others. Over the time course of childhood development, most individuals develop strong imitative skills, skills that are durable in part because of the sometimes inconsistent occurrence of being rewarded for imitation. By adolescence and often earlier, generalized imitation will persist for quite some time even in the absence of reinforcing consequences. The focus of social learning theory on these three processes for developing and changing behavior does not deny the possible relevance of other factors. Rather it represents a parsimonious determination to explore as fully as possible the possible role of environmentally mediated learning prior to developing accounts based on other mechanisms. This is similar to other fields, such as genetics (Hunt, 1962). Geneticists examine the influence of chromosomal factors in their accounts of the development of the human body and its actions. They largely ignore operant factors occurring during a person's life, possible psychodynamic influences, drug ingestion, and so forth. Nevertheless, genetics is seen as a credible field whose findings are accorded respect. To have a clearly defined focus is scientifically acceptable, and the focus of behavioral social work is on how the environment gives rise to behavior via learning processes. No other explanatory principles are necessarily deemed to be unimportant. This focus upon learning mechanisms that are based on the transactions occurring among persons and their environments is obviously congruent with central themes adopted by the field of social work and represents one manner of operationally defining the historic person-in-environment perspective of our field. Basic Assumptions It is not well recognized that behavioral social work's learning theory foundations have given rise to an approach to practice that is almost exclusively oriented toward the person-in-environment perspective that has long characterized the field of social work (Estes, 1984). A behavioral analytic perspective on human development is a viable alternative to traditional stage theories. Rather than postulating a hierarchy of stages in cognitive-moral-behavioral-development, behavior analysis attempts to explain the emergence of differential human capabilities and expressions in terms of the operation of various forms of empirically supported learning theories. These mechanisms of learning are said to apply to all cultures and races and to men and women equally. The psychosocial environments in which they operate may considerably vary, but the fundamental mechanisms are understood to remain the same. Practice Assumptions Behavioral social work does not strongly rely upon traditional theories of human development in attempting to understand culturally diverse clients. This is justified because the validity of traditional stage theories is increasingly coming into doubt, and because of the growing evidence that they may fail to capture cultural diversity issues related to development. Similarly, traditional practice wisdom and contemporary research about culturally diverse groups is being questioned. For example, it appears that most applied research on African-American women has involved limited or nonrepresentative samples. Most studies that examine racial differences highlight between-group differences, ignoring the fact that within-group variation is typically very large, and that various racial groups more often than not share far more features in common than differences. A similar caveat has emerged from meta-analyses of research on purportedly significant gender differences in cognitive functioning. From a social learning perspective, a middle-class white adolescent may have more in common with a middle-class black teenager than with a poor white youth from Appalachia (Arthur, 1992). Social learning theorists do recognize that differences among clients exist, and that these differences need to be assessed and taken into account when practicing with persons from culturally diverse backgrounds. For the behavioral social worker, effective culturally diverse practice does not consist of assuming that skin color or ethnic heritage exerts a predictable influence on one's role as a client. Rather, it consists of some traditional social work practices such as conducting individualized assessments, attempting to leave behind preconceptions of what people should be like, based on their race or gender to take what clients bring to the practitioner in a non-condemnatory way to respect clients' views to try and conceptualize the problem in a manner consistent with social learning theory and to translate this conception to the client in a manner he or she can understand to propose, based upon empirical research if such exists, one or more courses of joint action to develop a means to assess client change in a manner that is acceptable to the client, and is at the same time reliable and valid to alter treatment plans as the client and/or data indicate modifications are needed and to recruit significant others into treatment as needed to help ensure the maintenance and generalization of meaningful change. The above is not meant to disregard those cultural differences that may exist between clients, but is intended to emphasize that such differences need to be revealed through individualized assessment, rather than assumed to exist by the social worker. Behavioral social workers make no pretence that the behavioral approach is value free or can be applied across the board with different persons, regardless of their individual likes and dislikes or learning histories. But it is sufficiently flexible that a skilled social worker, drawing upon an informed knowledge of social learning theory, is in a good position to devise an interventive program that takes into account such individual differences and that will be of benefit to the client. Critique and Limitations In discussing the real and potential limitations of utilizing social learning theory in social work practice with culturally diverse clients, the one limitation that strikes the author most forcibly is the extent to which this approach continues to fall prey to misconceptions about the nature of behavioral practice. A second limitation, shared with colleagues from other theoretical orientations perhaps, is the temptation to apply techniques in a rotelike manner. A third limitation is that the study of the principles of social learning theory and of the practice techniques comprising behavioral social work could consume an entire M.S.W. training program. This is a conceptually rich orientation, which has generated literally hundreds of interventive methods, each with its own degree of empirical support and applicability to clients from culturally diverse backgrounds. The design and content of most M.S.W. programs does not allow for mastery of this content. As a consequence, social workers are often not aware of the remarkable degree of empirical support that this approach to practice has garnered, and are poorly trained in the skillful delivery of behavioral interventions (Blonski, 1999). Social learning theory is itself an evolving body of knowledge. The traditional distinctions between respondent and operant conditioning have been called into doubt by some researchers, and it may emerge that learning via imitation is a complex form of operant conditioning as opposed to a conceptually distinct mechanism. Fifty years ago the field now known as social learning theory did not exist. It is to be hoped that 50 years hence the present formulations will have been replaced by a conceptual framework that captures even more of nature's truth and that yields ever more effective methods of social work practice. In the interim, it must be acknowledged that the present framework is an incomplete one and that many areas of social work practice, particular those dealing with culturally diverse client groups, lack a strong foundation of empirical support for the application of behavioral methods. Recognition of this fact should not obscure the impressive achievements that have been made to date in this area, but rather should serve as a stimulus for further conceptual and empirical progress. Social learning theory has much to offer social work in terms of helping members of our profession offer culturally sensitive and effective interventive services to diverse client groups. The theoretical foundation of behavioral social work is strongly supported by empirical research, and numerous replication studies across clients representing various cultures, races, ethnic backgrounds, nationalities, sexual orientations, and other aspects of diversity suggest that this approach produces positive benefits generalizable to various clients groups. A social learning theory perspective on diversity issues suggests that cultural/ethnic differences arise through the operation of common learning processes occurring within the context of differing psychosocial environments. No particular group is held up as normative standard, and an individual's behavior can only be properly understood when examined within the context in which it developed. By viewing client problems as arising from past and/or present environmental learning experiences and as a function of various physical, social, and psychological resources, an inherently nonpathological, respectful, and optimistic perspective arises from which to promote positive changes. Reference: Arthur, W. Brian (1992) 'On Learning and Adaptation in the Economy', Discussion Paper Number 854, Queen's University, May. Blonski, Matthias (1999) 'Social Learning with Case-Based Decisions', Journal of Economic Behaviour and Organisation, 38(1). Estes, W. K. (1984) 'Human Learning and Memory', in Marler, Peter and Terrace, Herbert S. (eds.) The Biology of Learning (Berlin: Springer-Verlag). Hunt, Earl B. (1962) Concept Learning: An Information Processing Problem (New York, NY: Wiley). Rozin, P. and J.W. Kalat. 1971. Specific hungers and poison avoidance as adaptive specializations of learning. Psychol. Rev. 78. Laland, Kevin (1996) 'Is Social Learning Always Adaptive?', Paper presented at ELSE Interdisciplinary Workshop. Scharfstein, David S. and Stein, Jeremy C. (1990) 'Herd Behaviour and Investment', American Economic Review, 80(3), Schlag, Karl H. and Pollock, Gregory B. (1999) 'Social Roles as an Effective Learning Mechanism', Rationality and Society, 11(4), Wright, W. A. (1995) 'Sequential Strategy for Learning Multi-Stage Multi-Agent Collaborative Games', Draft Paper. Read More