As an adolescent, I had been much taken up with the non-theistic possibilities of Hindu philosophy -- especially, the solipsism of Advaita (the philosophy of "non-dualism") and the concept of action-without-desire of the Gita. I could never figure out how this philosophy actually fit in a real life, especially the real life of an atheist. The core problem for Advaita, like that of Buddhism, is the relation between the individual (the Atman, frequently translated as the "Self") and the rest of the world, referred to as the Brahman, translated as the Universal Self. The problem apparently arises because we are conscious of our selves and see the relation between the self-conscious being and the rest of the world in a variety of ways that can be termed transactions. Being conscious of self, we are conscious of other beings that participate in these transactions. The concepts of Dharma (that which is lawful) and Karma (actions, and the consequences of actions) serve to organize the rules governing these transactions.

As is usual in human concept invention, these concepts evolved from practice. The relevant attribute is that humans are multiple autonomous beings that cooperate to perform complex activities in a social context. Learning to perform complex activities requires recognition of the differing roles in those activities and some mechanism for allocating credit (or blame) between roles and the actions performed by those roles. Some of these activities are "successful", other are "failures". However, most activities are complex and the process of determining how credit (I will use this word to include "blame" as well) is to be allocated among the constituent actions is a difficult one. Many complex activities have actions that are performed by single individuals. Actions may also be associated with other unique objectifiable components of the external world, such as trees, animals, or places, etc.. The cognitive processes involved in learning assign some portion of the credit for an action to one or more of these objectifiable entities. The process of assigning credit for an action and the objectifying of worldly entities go hand-in-hand.

In the human social environment, the learning process is also social. That is, the objectified entities are associated with signs or symbolic tokens that can be used for communication. Without the construction of such communicable tokens there can be no learning that transcends the individual.

But the construction of a token is not a simple act. The token must be accepted by the community. By accepting a token, the community accepts the paradigm implicit in the token that some entity is an object and is worthy of being allocated credit. The token represents the entity and can potentially be subject to certain kinds of symbolic manipulation. The process is fundamentally social and political, in addition to being cognitive.

Being a political, social, and cognitive construct, a newly constructed token goes through a period of severe stress-testing. The more complex the construct, the more complex the stresses it is subjected to. Also, being political and social, tokens that represent fundamental transformations are frequently ignored (only occasionally, actively rejected) by the society at large, but comes to occupy a niche, usually a sub-community that has use for the construct. As the values and paradigms of that sub-community spread, some of that community’s tokens come to be more commonly accepted, and become part of the common understanding.

At the individual level, the assignment of credit to participating objects can be modeled as a largely involuntary function of our brain. But when a society begins to try to assign credit for the performance of complex activities, the process is no longer simple. Among other things, the symbol system must develop the concept of an "agent" that plays a "role" in the activity. Under suitable conditions, the role of "director" may evolve. Under suitable conditions, the director role-player may have to assign credit for its own performance. That is credit-assignment is no longer an autonomic activity of the individual cognitive systems involved. In such a context, directors may have to objectify their own role, in order to evaluate themselves successfully. This is a first step in the development of self-consciousness. Once the role of "director of complex activities" develops in a society, that society has a necessary component for the development of self-consciousness in individuals.

My experience as a computer scientist specializing in cognitive science and "artificial intelligence" made me aware that much intelligent behavior can be accomplished without self-consciousness. Consciousness of self is not a necessary feature of even the most complex behavior of systems consisting of multiple, autonomous entities. Some complex computer programs have been constructed that illustrate the point; however, we can also observe the complex behavior of social insects as an example of this point. The need for consciousness of self arises when the entities must engage in two-way equal communications to accomplish explicitly stated long-term goals. With short-term goals the statement of a solvable problem often implies the solution and the coordination needed to resolve it, and explicit communication may not be necessary. That is to say when the role of "director" is formulated, the complementary role of "directed agent" is also formulated. The director and the directed cooperate in establishing goals and responsibilities for achieving them. The concepts of "self" and "other" encoded in the external language of communication are useful for this cooperation. They may not be necessary for the directed agent, but may be necessary for communication between a niche sub-community of directors.

Thus the need for social credit assignment leads to the development of directors. Directors need to objectify themselves to perform credit assignment successfully. Directors then need a language with the objectified concept of self for communication with other directors. Note that the objectified concept of the "other" is not as complex. The objectified concept of "you", i.e., a non-other who is nevertheless not in telepathic communication with one is the most complex of all. That is to say, the concept of first, second, and third persons probably evolved in the order third, first, and second.

If the autonomous entities (that is, director and directed) function in hierarchically organized and organically interconnected units, the concept of self and other is not simple or static. A director may be a directed agent in a different context, and vice versa. Thus when the concept of self was first formalized for use within a linguistic niche, it might have been difficult to comprehend. If the society was already hierarchical, the concept would be seen as paradoxical. If the society was not hierarchical when the concept first came into being, it would have made sense only to a small subgroup. When the society then became hierarchical (as all complex societies tend to become), the concept would have been a major source of cognitive distress and schizophrenia within the group of directors. And for the directed, the concept would have been largely useless -- it would at best be a mysterious way by which the directors referred to themselves (the royal "we", so to speak, would have been incomprehensible to "them").

So, the concept of self implies and requires a major paradigm shift in the language of the group that developed the concept. Depending on the nature of relations between the first such group and other social groups, the concept might spread very fast between the directors of these groups. But the precise meaning of the concept would differ in some ways from society to society, since the director role would have developed in differing ways. (This is simplistic, but it could have developed in some societies through the role of warlord, in others from the role of trader, in yet others from the role of judge or mediator).

But just because the concept spread did not mean that it and its consequences were better understood even among the affected individuals. Julian Jaynes, in his book The Origin of Consciousness in the breakdown of the Bicameral mind identifies a phenomenon that he terms "the breakdown of the bicameral mind" as the source of the concept of self-consciousness. He hypothesizes a stage of human consciousness in which the mind was conceptualized as consisting of two "rooms" or components (hence "bicameral"). One room was the ordinarily conscious mind, but the other one appeared to the first as a divine voice that gave instructions that must be obeyed. My belief is that the situation is reversed -- the increasing use of language with constructs for self and other, and the spread of this language beyond the niche of "directors", resulted in the breakdown of a simple director-directed agent model for explaining and assigning credit for complex operations. A king required to explain some action performed by him might say that he was "directed" to do the action. This would be comprehensible to his people. In rationalizing an action after the fact, he might name a god or natural force as the director. This occurs simply because the initial version of the concept of self may not have been self-referential, and a self-referential use would have been incomprehensible. But as the usage of "self" as a construct spread, its obvious self-referential use would have lead to a breakdown in the old modes of rationalization and the development of new self-referential modes.

Computer programming languages and systems occasionally use self-referential concepts. Designers of "practical" systems tend to avoid self-referentiality -- it tends to be expensive to implement. As the systems grow in complexity, self-referentiality is re-invented. This re-invention of self-referentiality has been a feature of computer systems almost from the very beginning. As Russell and Whitehead showed in Principia Mathematica self-referentiality gives rise to many paradoxes, mostly unresolvable. As Turing and Godel showed later, self-referentiality leads to non-computable functions and undecidable theorems. Predictably, the re-invention of self-referentiality in a computer language or system leads to a re-recognition that this is too expensive, again. This is followed by the invention of restricted concepts or features that obviate the need to use self-referentiality. To add to the irony, every re-invention of self-referentiality has been heralded as a major conceptual advance and the subsequent abandonment has been decried as a major regressive blunder necessitated by pragmatism.

As the use of the concept spread, its use in self-referential ways would have been an obvious extension. The problems of that extension, even if not apparent to casual users would have been apparent to the original niche users as well as philosophers with time on their hands. As the concept of self began to be used by non-directors in ways that had not been considered appropriate by the developers of the concept, considerable conflict would have been generated. The problems would have spanned the entire range from individual, to political, and social uses of the term. I speculate that this change occurred in most Old World languages sometime between 1500 BC and 800 AD. That is the "royal we" became "plain old I" and everybody was confused.

The confusion can be stated in terms of the classical questions regarding the "self" -- What statements can the "self" make about itself and how valid were they? What was this "self" and what was its relation to the "other" or "others"? What were the responsibilities of the "self"? How was the "self" to be held responsible for behavior? How was the "self" to act when its actions clearly affected the "other" in various ways, some good, some bad? And so on. Much of religious thought consists of answers to these questions. For instance, the Mosaic Laws can be considered a set of laws that largely define the relation between the self and the community, and thus define the "self". The above questions only arise after a concept of self is externalized and articulated for use in the common language.

The time period over which an initial set of answers to these questions developed is probably coterminous with the development of written and spoken languages for command. Thus we may have to search for the beginnings in the mammoth hunters of an earlier ice age (who may have used their ability to coordinate complex activities to destroy entire species). The process has not terminated yet, but a significant milestone probably occurred between 9000 B.C.E. (when the first chiefdoms may have appeared in the Old World and a need for director-directed agent language may have arisen) and 500 B.C.E. (by which time the concept of self, paradoxes and all, became part of the common language). The "coincidence" identified by Julian Jaynes, that the breakdown of the bicameral mind took only about 700 years (almost instantaneous in this time scale) and was apparently universal to the Old World would then be due to the well-developed trade links between the major Old World civilizations, that helped spread the concept. Thus we can tie together the development of self-conscious languages, the issues or personal responsibility dealt with in the Mosaic Ten Commandments, and the development of religions that claimed universality because they dealt with the relation between individuals and everything else. These are also the issues dealt with in the Bhagavad-Gita, an essential part of the Mahabharata.

The reality is that we still do not fully understand the consequences of the invention of the concept of the "self". The problem is that the concept is useful in some ways but its implications, if drawn, cause problems. When we raise questions about the rights of individuals versus the rights of society or the individual’s duty to society, we are raising questions that would have no meaning without the concept of self. When Mark Twain comments that he would never be a member of a club that would have him as a member, he is expressing something about the human condition that would not have been expressible to a pre-self-conscious human, but it is also a comment about the essential paradox that self-consciousness brings with it. Which questions and implications of the concept are legitimate and which ones are not is still controversial. We have elaborated on the concept, adding layers of hierarchy and control to mediate between the self and everybody else.

The concept of self is not necessary for long-term intelligent behavior or for long-term species preservation (both production and reproduction can proceed without externally expressed consciousness). It is not even necessary for the maintenance of complex, organic social structures, or for the maintenance of family and kinship groups (witness insects such as bees and ants, and pack animals like dogs and wolves). When the concept, of the self is used for thinking about these things, controversy, confusion, and conflict reign

My version of the Gita takes off from this perspective. In the original, Arjuna, the war-leader of the Pandavas must blow his conch to signal the start of the war. But at the last moment he balks and questions the rightness of his actions in fighting this war. His charioteer, Krishna, counsels him. Krishna defines the concept of desire-free action. That is, Arjuna should do his duty and conduct the war, without expectations. Arjuna is a warrior, his duty is to fight wars that he has decided to fight. But Arjuna is the director of the war for his side -- his blowing of the conch will be the signal for the fighting to start, and thus for everybody else to perform their duty. So the question remains -- who determines Arjuna’s duty and why should Arjuna accept that determination. Since it is Arjuna and his brothers who constitute one side in the war, this is a self-referential question. In the Gita, Krishna ducks this issue by revealing to Arjuna his (Krishna’s) divine nature -- both his horrific aspect as eater of worlds, and his benign aspect as protector of creation. Arjuna should start the war because God, the ultimate director, is telling him that it is right and proper.

The above is a simple rendering of the Gita. But it highlights the aspect that I wish to focus on for my rewrite -- that the issue of responsibility that Arjuna raises has to do with the concept of self and its personal connection to the horrors to come. Krishna’s answer throws the burden of responsibility off the warrior’s back onto a nebulous higher authority. Then he admonishes Arjuna to maintain detachment from the consequences of his actions.

But there is a simpler self-referential answer that Krishna first tries out on Arjuna and that Arjuna rejects -- that Arjuna is acting as part of his family to obtain the rights that are due to them. Arjuna weighs his rights against the slaughter and rejects the equality. But Arjuna is committing a fundamental categorical error -- he is weighing the desirability of long-term, familial goals against the undesirability of the short-term effects on him as an individual of the means to achieve the goals. The undesirability of the effects are distressing to him personally; the loss of the long-term goals would affect his family and his future descendants. The two are not commensurable. But Krishna does not take this approach. There is a good reason for it, as it introduces an infinite regress, since ultimately the family consists of individuals who make decisions by themselves. Ultimately, Arjuna would have had to justify his war as being good for humanity.

So Krishna’s real answer is that Arjuna should not even try to compare two incommensurables. The long-term goals are Arjuna’s duty, and that’s all there is to it. Arjuna needed counseling to help him handle the distress (or pleasure) caused by the short-term effects that might prevent him from accomplishing his long-term goals and Krishna provided that counseling.

In the long run, we are all dead. Thus the only viable long-term goals are those of social institutions. Krishna’s counsel to Arjuna has to be understood as: In accomplishing the goals of your social institutions (i.e., your duty), you must not let your own self’s desires, whether pleasure or pain, get in the way.

So, my version of the Gita is as follows: The Pandavas have setup an ambush in which they expect to kill their grandfather Bhishma, the chief of the Kaurava army. This is the first engagement of the war in which one of the immediate family is likely to die. Arjuna is to execute the ambush with Krishna. Arjuna is nervous and upset and begins to talk about giving up the battle. Krishna advises Arjuna that in the pursuit of long-term social goals, the shorter-term goals and satisfaction of oneself are not material. Successfully accomplishing long-term goals requires detachment from the pleasure or pain associated with short-term actions. Thus, killing his relatives is necessary for achieving what he and his brothers had set to achieve and the intrusion of his personal likes and dislikes could only hinder their joint effort. The pain that he felt from killing a friend could distract him from the next task to be accomplished; so could the pleasure he got from killing an enemy.

In my version, Arjuna raises the issue of the infinite regress of goals. Krishna rejects the argument by limiting the number of levels of regress to the level at which the Society of Poets (the made-up organization of which Krishna is the head at this point in the story) functions to forestall crisis. Krishna’s "divine nature" lies in his leadership of that organization and its grand plans. By revealing some of these plans to Arjuna, Krishna convinces Arjuna that he is doing the right thing.

This advice is not particularly different from the Bolshevik point of view during the Russian Revolution, or for that matter, the behavior expected of managers in a capitalist corporation (in my first job for an large American company, one of my first faux pas was to disagree publicly about a corporate decision that had already been made. I was severely admonished with the words that junior employees would take my dissent to mean that a decision had not been reached and as a senior employee it was my duty to publicly fall in behind the corporate decision). This is also how the Nazis expected their bureaucrats to behave, or how the US Marine Corps and other military units around the world expect their troops to act. That is to say, once a corporate decision has been made, everybody is expected to fall in line.

This relationship is what a director expects from the directed. Anything else makes it impossible to manage complex activities and assign credit with any confidence. It is one of many solutions to the problem of how to handle personal responsibility. The modern consensus is that there is something fundamentally wrong with it; however, we still do not know how to deal with the problem in practice.

Krishna explains to Arjuna that duty is ultimately self-determined; and that understanding this self-referentiality was something that only Kshatriyas could expect to understand, because they acted as directors and had understood the need for self-referentiality to explain their own actions. Brahmins were technicians, architects, scribes, and other specialists and understood how to perform in response to direction, but had not understood the paradigm of directing. Vaisyas (traders and other middlemen) understood the concept of exchange, but their motivation for exchange was externally determined by their customers. The rest of the community were lead by their leaders. Only the leaders, the natural Kshatriyas, understood what it took to identify complex social goals and organize to achieve them. Without a self-referential language, Kshatriyas could not justify their deeds to themselves, but had to invoke gods (i.e., higher authorities) to explain their directions. With a self-referential language the self could be compartmentalized and could function in multiple roles and one role could determine the duties of another role, thus eliding the problems posed by self-referentiality.

In the original Gita, Krishna introduces the concept of "Yoga", i.e., discipline or yoke. He uses it to explain different ways by which an individual may understand the relation between the individual self and the Universal Self. In my version, Krishna discusses the discipline needed to manage a self-referential and self-directed life. These include the ability to compartmentalize one’s life, the ability to perform without second thoughts actions that have been pre-determined, and the ability to avoid suffering as a result of such a split in consciousness. This is the Yoga of self-knowledge (Atma-Vidya-Yoga? Swa-Jnana-Yoga?).

I do not expect that my revision of the Gita will have the poetic qualities or elegance of the original. But it is close to the original in some sense and deals with the same problems.