Skip to main content
Log in

On the Coevolution of Theory and Language and the Nature of Successful Inquiry

  • Original Article
  • Published:
Erkenntnis Aims and scope Submit manuscript

Abstract

Insofar as empirical inquiry involves the coevolution of descriptive language and theoretical commitments, a satisfactory model of empirical knowledge should describe the coordinated evolution of both language and theory. But since we do not know what conceptual resources we might need to express our future theories or to provide our best future faithful descriptions of the world, we do not now know even what the space of future descriptive options might be. One strategy for addressing this shifting-resource problem is to track the predictive and linguistic dispositions of inquirers rather than to track their theories and conceptual resources directly. Sender-predictor games, a variant of Skyrms–Lewis sender-receiver games, provide very simple models for the coordinated coevolution of predictive and linguistic dispositions. Such models explain how it is possible for (1) predictive and descriptive dispositions of inquirers to coevolve, (2) term-wise incommensurable, but nevertheless descriptively faithful languages, to sequentially evolve, and (3) a sort of underdetermination to occur where inquirers might satisfy their descriptive and predictive aims by revising their linguistic dispositions, their theoretical dispositions, or both. Such models also provide an elementary characterization of what it might mean for descriptions of the world to be faithful and hence for empirical inquiry to be successful. In doing so they provide a relatively weak, but perfectly clear, endogenous account of epistemic norms.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6

Similar content being viewed by others

Notes

  1. The shifting-resource problem is closely related to the problem of how to handle the catch-all hypothesis in a Bayesian model of inquiry, which is, in turn, closely related to what Kyle Stanford calls the problem of unconceived alternatives (Stanford 2006). The shifting-resource problem is also related to the problems posed for rational inquiry by the threat of diachronic Kuhnian incommensurability (Kuhn 1996). The fundamental problem is that, insofar as we expect our descriptive resources to evolve in inquiry, we currently lack the resources even to express our best future theoretical options. Stanford takes this to pose a problem for scientific realism. The suggestion here is that the shifting-resource problem is a problem for anyone who wishes to provide a concrete model of inquiry insofar as such a model requires one to characterize future conceptual options.

  2. For discussions of standard Skyrms–Lewissender-receiver games and variants see Argiento et al. (2009), Barrett (2009, 2007a, b, 2006), Lewis (1969), Skyrms (2006, 2010). The sender-receiver and sender-predictor games described in the present paper were coded and run as C++ simulations on a notebook computer. For details on the implementation of these and similar games see Barrett (2006) in particular. The idea of a sender-predictor game came from conversations with Michael Dickson. Dickson has subsequently studied sender-predictor games that are significantly more subtle than those described here. Of particular interest are models where the laws of nature are not deterministic. In such models, assuming a simple learning dynamics for the agents that closely resembles the dynamics described below, the agents are typically able to successfully evolve a descriptive language and successfully predict the most likely evolution of nature.

  3. See Lewis (1969) for the original description of signaling games in the context of traditional game theory. See Argiento et al. (2009), Barrett (2009, 2006), Huttegger (2007), Skyrms (2010) for descriptions of the basic game and variations in the context of evolutionary game theory and for discussions of conditions under which successful signaling might evolve.

  4. If the dynamics of nature is deterministic, the sender-predictor game will be formally equivalent to a sender-receiver game with the same overall structure. That the same game can be interpreted two very different ways, as a game involving the evolution of language for successful coordinated action or as a game involving the evolution of language and successful prediction, is evidence for how closely intertwined the evolution of language and theory may be.

  5. On this model, for N max = 1,000 and 1,000 runs each with 106 plays, for example, the sender and receiver nearly always (0.993) evolve a set of nearly optimal (0.994) dispositions. The exact details of the learning dynamics do not matter much as similar results are obtained for a wide range of parameter values for this sort of bounded reinforcement learning with punishment. What matters for evolving successful descriptive and predictive dispositions here is that (1) there be a maximum number of each type of ball in each urn (so bad habits do not get too strongly ingrained), (2) no type of signal or predictive action ever go to extinction in any urn (so there is always a possible escape from suboptimal dispositions), and (3) both reinforcement and punishment, or weakening, of first-order disquisitions be possible (since without reinforcement, the inquirers would be unable to learn; and without punishment they would be unable to forget suboptimal dispositions and hence lack the means of escape). On such a learning dynamics it is possible for the inquirers’ first-order dispositions to randomly wander away from success, but they will quickly return and spend most of their time almost ideally successful in their actions. See Barrett and Zollman (2009) for a discussion on the role of forgetting in evolutionary games.

  6. The following two games arguably allow for just such a distinction. A yet richer game, in the relevant sense, would be one where the same term is sometimes used to promote present action and sometimes used to promote future action. If the two uses of the same term are assumed to be synonymous, such a game would allow one to determine who is describing prior states and who is predicting future states in a sharper sense than the games below.

  7. For parameters akin to those discussed in the last game, the senders and receiver typically (0.995) evolve a set of nearly optimal (0.994) linguistic and predictive dispositions. Again, such results hold for a wide range of parameter values on bounded reinforcement learning with punishment.

  8. The language that evolves in this game is similar to the simple compositional language used by putty-nosed monkeys. See Arnold and Zuberbühler (2006, 2008). For a discussion of the evolution of natural kind terms see Barrett (2007a).

  9. For parameters akin to those described in the last two games, the inquirers here typically (0.972) evolve a second set of nearly optimal (0.993) descriptive and predictive dispositions that answer to the new regularities of nature and/or changed second-order dispositions.

  10. In simulations of the game as described here, the receiver evolved to make predictions differently with probability 0.581 and the senders evolved new coordinated partitions that classify the states of nature differently with probability 0.419. This lack of symmetry is interesting. Here it may be the result of the fact that it is harder for the senders to evolve a new coordinated language than for the receiver to evolve a new way of making predictions.

  11. The sort of termwise incommensurability exhibited here is relative tame. Since the agents can, in full statements, individuate precisely the same states of nature, while they lack the ability to translate terms between languages, they can translate statements. More subtle types of incommensurability, however, may evolve in such games. See, for example, Barrett (2009) for a discussion of the sort of incommensurability that may be exhibited between languages corresponding to different partial pooling equilibria. In this case, there is neither term-wise not statement-wise commensurability between the evolved languages. Further, while the agents cannot individuate precisely the same states in such a case, their descriptions of the world may nevertheless be equally faithful insofar as they may allow for equally faithful overall precision in description.

  12. The pragmatic account of successful inquiry suggested here fits well with, for example, C. S. Peirce’s characterization of scientific inquiry. See “Some Consequences of Our Four Incapacities (1868),” “The Fixation of Belief (1877)” and “How to Make our Ideas Clear (1878)”, chapters 3, 7, and 8 of Houser and Kloesel (1992), respectively. Peirce’s theory of signs also allowed for the coevolution of language and theory. Indeed, sender-predictor games might be taken as providing a very simple dispositional model for his late theory of signs where the interpretant in his account is coded for in the dispositions of the inquirers. See Atkin (2010) for an introduction to Peirce’s semiotics.

  13. See Barrett (2008) for a discussion of descriptive nesting in inquiry.

  14. See Barrett (2008) for an extended discussion of this point.

  15. Tracking such explanatory dispositions would require a model that is significantly more subtle than anything considered in this paper. The evolution of explanatory dispositions might, for example, involve the evolution of dispositions to send signals that characterize the relationships between one’s own first- and second-order dispositions to signal and predict and that may instill similar relations in the dispositions of other inquirers.

References

  • Argiento, R., Pemantle, R., Skyrms, B., & Volkov, S. (2009). Learning to signal: Analysis of a micro-level reinforcement model. Stochastic Processes and Their Applications, 119(2), 373–390.

    Article  Google Scholar 

  • Arnold, K., & Zuberbühler, K. (2006). The alarm-calling system of adult male putty-nosed monkeys, Cercopithecus nictitans Martini. Animal Behaviour, 72, 643–653.

    Article  Google Scholar 

  • Arnold, K., & Zuberbühler, K. (2008). Meaningful call combinations in a non-human primate. Current Biology, 18(5), R202–R203.

    Article  Google Scholar 

  • Atkin, A. (2010). Peirce’s theory of signs. Stanford encyclopedia of philosophy. http://plato.stanford.edu/entries/peirce-semiotics/. Accessed 23 April 2012.

  • Barrett, J. A. (2006). Numerical simulations of the Lewis signaling game: Learning strategies, pooling equilibria, and the evolution of grammar. UC Irvine Institute for Mathematical Behavioral Sciences Technical Report. http://www.imbs.uci.edu/tr/abs/2006/mbs06_09.pdfMBS06-09. Accessed 23 April 2012.

  • Barrett, J. A. (2007a). Dynamic partitioning and the conventionality of kinds. Philosophy of Science, 74, 527–546.

    Article  Google Scholar 

  • Barrett, J. A. (2007b). The evolution of coding in signaling games. Theory and Decision, 67(2), 223–237.

    Article  Google Scholar 

  • Barrett, J. A. (2008). Approximate truth and descriptive nesting. Erkenntnis, 68(2), 213–224.

    Article  Google Scholar 

  • Barrett, J. A. (2009). Faithful description and the incommensurability of evolved languages. Philosophical Studies, 147(1), 123–137.

    Article  Google Scholar 

  • Barrett, J. A., & Zollman, K. (2009). The role of forgetting in the evolution and learning of language. Journal of Experimental and Theoretical Artificial Intelligence, 21(4), 293–309.

    Article  Google Scholar 

  • Houser, N., & Kloesel, C. (Eds.). (1992). The essential peirce: selected philosophical writings, volume 1 (1867–1893). Bloomington and Indianapolis: Indiana University Press.

  • Huttegger, S. (2007). Evolution and the explanation of meaning. Philosophy of Science, 74, 1–27.

    Article  Google Scholar 

  • Kuhn, T. S. (1996). The structure of scientific revolutions. Chicago: University of Chicago Press.

    Book  Google Scholar 

  • Lewis, D. (1969). Convention. Cambridge, MA: Harvard University Press.

    Google Scholar 

  • Skyrms, B. (2006). Signals. Philosophy of Science, 75(5), 489–500.

    Google Scholar 

  • Skyrms, B. (2010). Signals: evolution, learning, and information. New York: Oxford University Press.

    Book  Google Scholar 

  • Stanford, K. (2006). Exceeding our grasp: Science, history, and the problem of unconceived alternatives. Oxford: Oxford University Press.

    Book  Google Scholar 

Download references

Acknowledgments

I would like to thank Brian Skyrms, Michael Dickson, Elliott Wagner, Simon Huttegger, and Tucker Lentz for helpful discussions. I would also like to thank Jim Weatherall, Cailin O’Connor, Alistair Isaac, and the referees for helpful comments on an earlier draft of this paper.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Jeffrey A. Barrett.

Rights and permissions

Reprints and permissions

About this article

Cite this article

Barrett, J.A. On the Coevolution of Theory and Language and the Nature of Successful Inquiry. Erkenn 79 (Suppl 4), 821–834 (2014). https://doi.org/10.1007/s10670-013-9466-z

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10670-013-9466-z

Keywords

Navigation