Psikhologicheskie Issledovaniya • ISSN 2075-7999
peer-reviewed • open access journal


Steiner P. Dialogism and Game Theory [Full text]

Russian version: Штайнер П. Диалогизм и теория игр
University of Pennsylvania, USA

About author
Suggested citation

The idea of dialogism is considered in historical aspect as an alternative to individualistic/atomistic and collectivistic/organic approaches. The relation of M.M. Bakhtin’s dialogism theory and von Neumann-Morgenstern’s game theory is analyzed. These theories differences concerning their structuredness level, determinancy, methodological basis (humanities and natural sciences), relation to language (as a universal environment or mean of calculation) and morals are revealed. It is shown that when explaining interpersonal interaction in social space these two theories match up in cultural field though move in different directions complementing each other more than similar ones.

Keywords: dialogism, game theory, semantics, logic, interaction


As Mike Holquist observed in his acclaimed book on the subject, dialogism is a concept that can be tackled from many different angles. “At the highest level of (quite hair-raising) abstraction,” it might be understood, he suggested, as a new and distinct epistemological principle [Holquist, 2002, p. 15]. To appreciate the significance of this principle in Western thought let me place it, first of all, in its historical context.

According to Ernest Gellner’s sweeping generalization about intellectual history, two theories of knowledge – profoundly alien to each other – informed European thought at the dawn of the last century: the individualistic / atomistic one on the one hand, and the collectivistic or organic on the other [Gellner, 1998, p. 3–39]. The difference between the two can be briefly summarized as follows: the former conceives of cognition as an empirical process in which an isolated individual (in a Robinson Crusoe-like manner) aggregates distinct units of external information into a rational whole. In the realm of social reality, Gellner argues, this episteme coincides with the liberal outlook that treats men and women as free agents who associate with others in a contractual manner guided solely by private choice unencumbered by any exogenous force, whether social norm or political power. The antipode of this knowledge-paradigm denies the atomistic nature of our existence. For collectivists / organists, humans are not discrete particles who in a single-handed cognitive effort grasp and comprehend the world around them, but always already integral components of a cultural whole. What matters is the social fabric holding everybody together rather than the beheld subjects. Understood thus, each and every individual consciousness is – to its innermost layers – constituted by what it shares with other members of a group: religion, history, language. Who we are and how we think is a function of these collective mental representations knowing no authorship or ownership. Consequently, if this category still makes sense, an individual can be nothing more than a passive impersonator of such cultural systems, their avatar in the lingo of the IT techies.

Furthermore, the word “culture” is not simple. It has a dual thrust, two distinct implications. Every inclusion presupposes some exclusion. Culture fuses human beings into indissoluble communities – ethnic groups, social classes, nations. But at the same time, it opens unbridgeable gaps between insiders and outsiders, those who belong and those who do not. If the individualistic paradigm is all-inclusive since it regards all rational beings as essentially the same, its collectivistic rival is particularistic, dividing humankind into the people of culture vs. the barbarians. And the two shall never meet save on the battlefield.

Dialogism, insofar as I understand this notion, offers a way out of the epistemological dilemma presented by these two incompatible epistemological paradigms. It salvages the concept of subject condemned to death by the organicists without, at the same time, relapsing into the extremes of the subjective individualism propagated by the atomists. From its particular vantage point, beings are autonomous and self-interested entities yet not necessarily windowless monads unconnected to a social whole. This is so because human nature is above all interactive: our selfhood cannot be fully constituted without the mediation of others. We are but agents in an ongoing transaction through which we traffic all cultural fungibles-mates, commodities, words. Our actions might be self-seeking but, insofar as they must reflect (whether directly or indirectly) the intentions of the partners / rivals with whom we engage, our behavior cannot be totally self-confined. Inevitably it must transcend the limits of a single consciousness, containing within itself prefigured reactions of the co-participants as well as their projected secondary and tertiary feedbacks, etc. Outside this unfolding reflexive loop, so the argument goes, the quintessence of the human existence cannot be grasped at all.

In the field of literary studies dialogism is ineluctably linked to the work of Mikhail Bakhtin and the circle of his friends who expanded the concept of dialogue far beyond its strictly technical sense, rendering it a cognitive metaphor for the humanities and social sciences. But pioneering as they may be, Bakhtin et al. are not the only 20th century scholars who forged an interactive model as an epistemological paradigm in their respective disciplines. Since 1944 when John von Neumann and Oskar Morgenstern published their epoch-making book, game theory has emerged from mathematics as an ever expanding inter-disciplinary matrix across an ever-growing array of scholarly fields [von Neumann, Morgenstern, 1944].

Within this specific historical context the Bakhtinians and game theorists might appear as comrades-in-arms. Outside of it, however, their affinity seems far from obvious. In what will follow I will attempt to compare in the most elementary fashion these two intellectual strains to see whether there are any points of contact between them besides a most general thrust toward what I termed “dialogism.” My exercise, I must confess, has proven quite frustrating for a number of reasons, most notably because of the extreme diffusion of game theory. According to the most recent survey of the field I was able to find, published in 2003, “while games comprise widely used method in a broad intellectual realm (including, but not limited to philosophy, logic, mathematics, cognitive science, artificial intelligence, computation, linguistics, physics, economics), each discipline advocates its own methodology and unified understanding is lacking [Pietarinen, 2003, p. 317].” For reasons of convenience as well as my own limitations, I drew game theoretical material from the somewhat narrower realm of social sciences.

The second source of my distress has been the great disciplinary incongruence between the two subjects of my comparison. Bakhtin himself was above all a student of fiction and, though his contributions exceed literary criticism narrowly conceived, I’ve yet to meet an American economist recognizing his name. Game theory, on the other hand, was developed originally as an alternative to natural-science oriented mathematics, well suited for describing inert matter but not economic interactions between two minds trying to psych each other out. As such, it relies on a high degree of formalization and appears rather clumsy when related to entities fuzzy to the extreme, like novelistic discourse.

Given the incongruity of their respective disciplinary orientations, it is not surprising to notice how Bakhtinians and game theorists differ in their view of language. Using the oft quoted distinction drawn by Hintikka (whose own contribution to game-theoretical semantics can be only acknowledged here), one might couch this difference in terms of two antithetical conceptions of language “as a universal medium” and “as calculus [Hintikka, Hintikka, 1986, p. 1].”[1] For Bakhtin et al. language is the primary modeling system informing our consciousness in each and every one of its aspects. We can never escape it, for only through language do we establish – within the equally linguistic vortex of social relations – our very personal identity. Game theorists, on the other hand, view natural language as a calculus, an implementation of logical rules that can be transformed, according to specified procedures, into any analogous notational system (whether verbal on nonverbal) independent of its actual user. A host of consequences follow when one of these two stances is adopted over the other. Let me mention just two of them.

Those upholding the calculus view tend to maintain rigorously the distinction between object- and meta-languages. The possibility of stepping out from one’s own language is not only possible but even desirable. An external meta-view provides an opportunity to leave behind the confusions with which a home language might be marred, to focus on its lacunas and contradictions invisible from an initial perspective, to map it better, improve or even replace it at least partially. Bakhtinians, subscribing to the universal medium position, regard the object-/meta-language split untenable. According to them, insofar as all discourse is but dialogical – an endless chain of replies to replies – every utterance is meta-meta-linguistic, always reflecting in its structure what was already said before somewhere, somehow. But this dialogue unfolds, the Russian thinkers maintained, strictly within the bounds of language, however heterological this entity might be. The translation of a message into any other rational system of representation (e. g. logical symbols) would entail such a significant semantic deficit as to render such an enterprise truly meaningless.

The inability to look at one’s language from without entails another problem dividing the two camps: the relationship between words and things. Calculus-position proponents like game theorists rely on a meta-linguistic concept of truth that derives from a certain correspondence between a proposition and non-verbal reality. This reliance enables them to match their hypotheses observable facts and modify the former accordingly. Thus, we experimentally might learn that the majority of players in actual interactive games of a more complex type do not select optimal strategies as game theorists define them, and this might temper our enthusiasm about the theory’s ultimate applicability. Nevertheless, we can decide this issue on the basis of empirical evidence. The dialogism of the Bakhtin circle, because of its particular conceptualization of language, is not fallible in this way. The universe of discourse generates its referents or, put differently, meaning comes from correlating one utterance to another rather than to extra-linguistic designata.

This stance, however, doesn’t lead the Russian “meta-semioticians” to an unbridled linguistic relativism as one might expect. Instead of a “reality-check” the slippage of signs is limited by what could be called a “morality-check” – the ethical dimension of language – the ultimate answerability of a responsible subject engaged in a dialogue. I cannot dwell on this subject at any length here. But their respective attitudes to ethics provide me with another point of comparison between Bakhtin and the game theorist. While the former’s emphasis on moral values as the motivating force of human action seems well recognized[2] (see, e.g., [Hirschkop, 1999]), the latter’s view is more complex. Let me illustrate what I mean by perhaps the most well-known example of a non-cooperative two-person game known as the prisoner’s dilemma [transparency: see at the end]. I take the label “non-cooperative” as vaguely similar to what Bakhtin called “the word with a loophole” in the sense of “preserving for oneself the possibility of changing the last, terminal meaning of one’s words [Bakhtin, 1979, p. 271].” Put simply, two criminals are absolutely unreliable beings, unable, therefore, to enter into any binding agreement. Thus, even though collaboration (a reciprocal non-confessing) would provide the best payoff for the two players, the advice that game theory offers to the agonizing prisoners is to the contrary. Do not try to be a mister nice guy, confess on the spot before the other does and let him / her rot in the slammer for ten years while you are out in three months or, if he / she confesses too, which is the most likely behavior, get at least a reduced sentence.

This example suggests a radical incompatibility between Bakhtin and game theory in terms of ethics. In the Hobbesian world projected by the latter no altruistic behavior (even if obviously beneficial for the two parties involved) is rewarded and a defection is the must. Yet, the game-theory view of morality is not as unilateral as it might seem. To make the point let me bring in Bakhtin’s unrestricted concept of dialogue understood as an exchange with neither a beginning nor an end, always unfolding, never ever finalizable. Game theory, I must stress, is well aware of the temporal dimension circumscribing any human interaction and it carefully differentiates single-round games from those repeated indefinitely. To wit, the prisoner’s dilemma can be played as a one-shot affair but also as a series of such individual games. If this happens, however, the nature of the interaction changes completely. Now, in making their respective decisions, the players can take into account the rival’s previous moves and through emitting signals about their own intentions attempt to condition his / her future responses. A different strategy, the game theorists tell us, is optimal in this situation: cooperation rather than defection.

The logic of this advice is easy to grasp. We are facing what different branches of social science call “trust, “reputation building,” or “social capital.” In competing for a while the dilemma stricken prisoners not only realize that long-term mutual benefits outweigh short-term individual gains, but also learn the appropriate communicative mechanisms through which thieves’ honor is generated. Out of the blue, asocial criminals start bonding. From the chaos and anarchy of a dog-eat-dog universe a spontaneous and self-sustainable moral order emerges in which people are obliging because they expect the same from others. And the most successful strategy to induce such a norm of reciprocity in erstwhile rivals, Robert Axelrod’s experiments with the indefinitely repeatable prisoner’s dilemma game, is, strangely enough, a variation on Matthew’s maxim “thou shalt love thy neighbour as thyself” (19:19). In a series of tournaments populated by game theorists and other social scientists, a “TIT FOR TAT” strategy – do not do to others what you do not wish upon yourself – proved to be the winner. This strategy is successful because, according to Axelrod, it satisfies the following three conditions: “1. The possibility of encountering TIT FOR TAT is salient. 2. Once encountered, TIT FOR TAT is easy to recognize. 3. Once recognized, TIT FOR TAT’s nonexploitability is easy to appreciate [Axelrod, 1984, p. 53-54].”

Does this self-through-other reflexivity as the source of ethics bring game theory close to Bakhtin? Most likely not, because the two seem to be talking at cross-purposes. The unrestricted nature of dialogue is for Bakhtinians the gateway to freedom. The meaning of a word is as infinite as is the number of the contexts it can enter – actually or potentially – and cannot be ever fully determined. For game theorists, on the other hand, the indefinite reiteration of the same game establishes conventions limiting the agent’s choices of moves and increasing the predictability of the outcome. This conclusion, however, should not come as a surprise to anyone. To quote one of the commentators on the subject, “in game theory… determinism is more than byproduct; it is an ambition [Varoufakis, 1991, p. 185].”

Given this glaring contradiction in their respective research agendas should I stop my comparison of Bakhtinians and game theorists here and declare its total futility? This, I believe would be a bit premature. Parallelism can be either positive or negative and it is the latter to which I will try to draw your attention. There is no doubt that finite two-person zero-sum games of perfect information (e.g., checkers) are strictly determined as was proven even before game theory was formulated [Morton, 1970, p. 13]. But discovering exact strategies for simple games is, in my opinion, the least attractive aspect of the theory. What I find intriguing about it is the fact that in its scientific rigor game theory was able to undermine the very achievement of its goal mentioned above: the belief in a deterministic explanation of social phenomena. And it is not just their complexity – the variety of possible alternatives – which makes such phenomena incomputable. Debunked is the entire rationalist tradition that has been with us since the 18th century, the reliance on Reason as the sole explanatory principle of human actions in general and rational choice theory in particular.

Let me explain what I mean by returning (for the last time) to the prisoner’s dilemma exhibit. Rational strategy in this game, so the theory tells us, is one which enables a player with well ordered preferences to maximize his / her payoffs (i.e., being jailed for the shortest time). But what is reasonable, I illustrated earlier, depends entirely on the repeatability of the game. And anything short of an indefinite iteration can be reduced logically (through backward induction) to a single-shot game. That is, if the dilemma-solving exercise consists of only 100 rounds, in our final move defection is the safest bet for no longer do we need not to worry about any effect this decision will have on the future behavior of our rival. The same, however, holds also for round 99 for if our competitor correctly figures out that our last move will be non-cooperative, he will defect one step ahead of us. This reasoning can be logically extended to steps 98, 97 and back to step 1.

I introduced this little wrinkle just to point out yet another problem with rational strategy: time preference. It is obvious that, because of their uncertainty, infinitely postponed benefits appear much less likely than those consumed immediately and will be discounted. This consideration might influence significantly any desire for long-term cooperation. “In the long run” Keynes is credited to have rebutted von Hayek’s blind faith in the slow but inevitable self-adjustment of all social systems, “we are all dead.” To sum up, in the presence of equally compelling multiple criteria rational choice calculation is indeterminate and the decision whether to defect or not is inevitably governed by other preferences.

This theoretical insight is further instantiated by empirical evidence. Our actual choices, needless to say, are virtually never calculated according to strategies that would pass as rational among game theorists. What they call “reason” is clearly a prescriptive rather than a descriptive concept. Humans of the real world tend to solve social problems not on the basis of logical reasoning but of collective experience. The instrumental rationality with which the theory started is, of necessity, replaced by a bounded rationality that stresses the role of social convention in our decision-making. But such convention seems clearly exogenous to “the game” taken in a strictly logical sense and is determined by factors any unreconstructed rationalist would surely deem irrational.

What is the lesson to be learned from this failure of game theory to deliver what it promised? As two students of the field put it succinctly in their “critical introduction”: If game theory does make a … substantial contribution, then we believe that it is a negative one… We believe that [it] reveals the limits of ‘rational choice’ and of the (neoclassical) economy approach to life [Hargreaves et al., 1995, p. 2].” In this way, I would like to conclude, the game theoretical path crosses that of the Bakhtin group, though traveling the opposite way. By pushing the edge of the envelope – submitting its assumptions to rigorous testing-game theory exhausted and closed one venue for explaining interactive human behavior in a deterministic fashion. And albeit its principles and methods are clearly alien (if not inimical) to the Bakhtinians’ understanding of dialogism one definitely should, as Pavel Medvedev – a member of this group – once remarked about Russian Formalists, “appreciate a good enemy much more than a bad ally [Medvedev, 1928, p. 232].”


Prisoner’s Dilemma
(attributed to A.W.Tucker)

Two suspects are taken into custody and separated. The district attorney is certain that they are guilty of a specific crime, but he does not have adequate evidence to convict them at trial. He points out to each prisoner that each has two alternatives: to confess to the crime the police are sure they have done, or not to confess. If they both do not confess, then the district attorney states he will book them on some very minor trumped-up charge such as petty larceny and illegal possession of a weapon, and they will both receive minor punishment; if they both confess they will be prosecuted, but he will recommend less than the most severe sentence; but if one confesses and the other does not, then the confessor will receive lenient treatment for turning state’s evidence whereas the latter will get “the book” slapped at him. In terms of years in a penitentiary the strategic problem might reduce to:

Adapted from R. Duncan Luce & Howard Raiffa. Games and Decisions: Introduction and Critical Survey. New York, 1957. P. 95.


Axelrod R.M. The Evolution of Cooperation. New York, 1984.

Bakhtin M.M. Problemy poetiki Dostoevskogo. 4 izd. M.: Sovetskaya Rossiya, 1979. 320 s. [in Russian]

Gellner E. Language and Solitude: Wittgenstein, Malinowski and the Habsburg Dilemma. Cambridge, 1998.

Hargreaves S.P. et al. Game Theory: A Critical Introduction. London, 1995.

Hintikka M.B., Hintikka J. Investigating Wittgenstein. Oxford, 1986.

Hirschkop K. Mikhail Baktin: An Aestehtic for Democracy. Oxford, 1999.

Holquist M. Dialogism: Bakhtin and His World. 2nd ed. London, 2002.

Kusch M. Language as Calculus vs. Language as Universal Medium: Study in Husserl, Heidegger and Gadamer. Dordrecht: Kluwer Academic Publishers, 1989.

Luce R.D., Raiffa H. Games and Decisions: Introduction and Critical Survey. New York, 1957.

Medvedev P.N. Formal'nyi metod v literaturovedenii: Kriticheskoe vvedenie v sotsiologicheskuyu poetiku, Leningrad: Priboi, 1928. 232 s. [in Russian]

Morton D. Game Theory: A Nontechnical Introduction. New York, 1970.

Pietarinen A.-V. Games as Formal Tools versus Games as Explanations in Logic and Science // Foundations of Science. 2003. N 8. 317–364.

Varoufakis Y. Rational Conflict. Oxford: Blackwell, 1991.

von Neumann J., Morgenstern O. Theory of Games and Economic Behavior. Princeton, 1944.

[1] Merrill B. Hintikka and Jaakko Hintikka. Investigating Wittgenstein. Oxford, 1986. P. 1. For further elaboration For further elaboration of this topic, see also Martin Kusch. Language as Calculus vs. Language as Universal Medium: Study in Husserl, Heidegger and Gadamer. Dordrecht, Kluwer Academic Publishers, 1989.

[2] See, e.g., Ken Hirschkop. Mikhail Baktin: An Aestehtic for Democracy. Oxford, 1999.

Received 20 May 2009. Available online 18 February 2010.

About author

Steiner, Peter. Ph.D., Professor, Department of Slavic Languages, University of Pennsylvania, 3451 Walnut Street, Philadelphia, PA 19104, USA.
E-mail: Этот адрес электронной почты защищен от спам-ботов. У вас должен быть включен JavaScript для просмотра.

Suggested citation

APA Style
Steiner, P. (2010). Dialogism and Game Theory. Psikhologicheskie Issledovaniya, 1(9). Retrieved from 0421000116/0008. [in English, abstr. in Russian].

Russian State Standard GOST P 7.0.5-2008
Steiner P. Dialogism and Game Theory [Electronic resource] // Psikhologicheskie Issledovaniya. 2010. N 1(9). URL: (date of access: 0421000116/0008. [in English, abstr. in Russian]

Back to top >>

Related Articles