Abstract
We are concerned here with how structural properties of language may come to reflect features of the world in which it evolves. As a concrete example, we will consider how a simple term language might evolve to support the principle of indifference over state descriptions in that language. The point is not that one is justified in applying the principle of indifference to state descriptions in natural language. Instead, it is that one should expect a language that has evolved in the context of facilitating successful action to reflect probabilistic features of the world in which it evolved.
Similar content being viewed by others
Notes
There is a tradition of such explanations in the field. Skyrms (2010a) shows how communication may evolve without natural salience, but not that it actually did so; Barrett (2007) and Purves (2018) show that it is possible for partitioning into linguistic kinds to evolve and to be ideally successful without latching onto natural kinds, but not that linguistic kinds in fact fail to track natural kinds; and Roth and Erev (1995), Erev and Roth (1998), and Barrett and Zollman (2009) show that dynamics like forgetting and experimenting may promote the evolution of language, but not that they actually have been causal factors in the evolution of language.
See Argiento et al. (2009) for details.
See Barrett (2014b) for a discussion of such games more generally.
See Barrett (2014a) for a detailed description of this particular sender-predictor game, its behavior, a discussion of the problem of priors, and a proposed dissolution of the problem.
See also Bertrand (1889).
Another well-known problem with the principle of indifference is that it leads to complications and inconsistencies on infinite domains (Keynes 1921). This will not concern us here, given that all of our examples are finite.
This is sometimes referred to as the maximum entropy principle. It is closely tied to the principle of indifference, but for reasons discussed in the last section and later in the paper, we do not take maximum entropy to provide any justification whatsoever for adopting the principle of indifference. See Jaynes (1957), and Williamson (2010) for contrary views on this issue.
Every partition that takes advantage of both signals, that is, every partition except (0, 10) and (10, 0)—which are dynamically unstable—has the same expected payoff as long as the sender and receiver coordinate upon the partition and regardless of what weight the receiver puts on each action conditional upon the relevant signal. See LaCroix (2019) for details.
The average expected payoff is calculated from the evolved dispositions of the agents (that is, from their actual urn contents) at the end of the run given the unbiased (or biased as in the next section) states of nature.
The fastest run achieves this payoff within 1801 plays and the slowest after 25995 plays. 0.90 runs are within 5% of their final expected payoff prior to the first 7500 plays.
The cumulative success rate is a measure of success that takes account of the history of the game. It is calculated by dividing the number of plays that led to a success by the total number of plays in that run. When the players are successful, early failures are washed out as the number of plays increases.
To provide a bit more detail, the null hypothesis is that the combinatorial distribution and the empirical distribution are identical for the \(10\times 2\times 10\) signaling game. The K–S test gives the statistic \(D=0.0843\), which is the supremum of the set of distances between the empirical distribution function we observe and the expected distribution function from the combinatorial measure. This corresponds to a p value of 0.0017; hence one might reject the null hypothesis with high confidence.
0.90 runs are within 5% of their final expected payoff prior to the first 9500 plays.
A K–S test of the empirical distribution against a distribution sampled from the combinatorial expectation yields the statistic \(D = 0.1912\), with a p value \(p < 0.0001\), suggesting again that the combinatorial expectations do not by themselves explain the empirical results.
0.90 runs are within 5% of their final expected payoff prior to the first 2200 plays.
0.90 runs are within 5% of their final expected payoff prior to the first 10000 plays.
A (5, 3, 1) partition occurs 0.073, a (4, 4, 1) partition occurs 0.019, a (5, 2, 2) partition occurs 0.066, a (4, 3, 2) partition occurs 0.030, and a (3, 3, 3) partition occurs 0.005. In each case, the distinct permutations of these partitions occur with roughly equal frequency.
For example, the data presented in the original studies of Seyfarth et al. (1980a, b) on vervet monkey alarm-call systems indicated substantial variation in responses and that the responses are probabilistic rather than deterministic. Furthermore, vervets do not just vocalize for alarm calls. They also call when they find food, in aggressive confrontations, and during sexual activity, among others. Vervets additionally vocalize via grunting in a variety of circumstances: (a) when a submissive meets a dominant individual, (b) when a dominant meets a submissive individual, (c) when one vervet goes out into an open area, and (d) when a vervet comes across an out-group conspecific (Cheney and Seyfarth 1982, 1990). Schlenker et al. (2016) provide a linguistic analysis of approximately 40 years of data from experimental primatology, which displays several complexities of monkey communication systems. This highlights but some of the myriad ways in which the world is significantly more complex than the model which we use to represent it.
See Barrett and Zollman (2009) for several learning dynamics that typically behave very differently than simple reinforcement learning. As a quick and very simple example, the learning dynamics win-stay/lose-shift would not even evolve a stable language in the present games, let alone one that supported the principle of indifference.
There is an extensive literature in cognitive science that seeks to quantify how language might reflect the structure of the world. See Lupyan and Dale (2016) for an overview and discussion of the linguistic niche hypothesis.
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, 373–390.
Barrett, J. (2006). Numerical simulations of the lewis signaling game: Learning strategies, pooling equilibria, and evolution of grammar. Technical Report. UC Irvine, Institute for Mathematical Behavioral Science. Preprint MBS 06-09.
Barrett, J. (2007). Dynamic partitioning and the conventionality of kinds. Philosophy of Science, 74, 527–546.
Barrett, J. (2009). Faithful description and the incommensurability of evolved languages. Philosophical Studies, 147(1), 123–137.
Barrett, J. (2014a). Description and the problem of priors. Erkenntnis, 79(6), 1343–1353.
Barrett, J. (2014b). On the coevolution of theory and language and the nature of successful inquiry. Erkenntnis, 79(Suppl 4), 821–834.
Barrett, J., & 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.
Bertrand, J. (1889). Calcul des probabilités. Paris: Gauthier-Villars et Fils.
Bradley, D. (2019). Naturalness as a constraint on priors. Mind, Forthcoming.
Cheney, D., & Seyfarth, R. (1990). How monkeys see the world: Inside the mind of another species. Chicago: University of Chicago Press.
Cheney, D. L., & Seyfarth, R. M. (1982). How vervet monkeys perceive their grunts: Field playback experiments. Animal Behaviour, 30(3), 739–751.
Erev, I., & Roth, A. E. (1998). Predicting how people play games: Reinforcement learning in experimental games with unique, mixed strategy equilibria. The American Economic Review, 88(4), 848–881.
Hempel, C. G. (1965). Aspects of scientific explanation, and other essays in the philosophy of science. New York: Free Press.
Hu, Y., Skyrms, B., & Tarrès, P. (2011). Reinforcement Learning in Signaling Game. arXiv preprint arXiv:1103.5818.
Huemer, M. (2009). Explanationist aid for the theory of inductive logic. British Journal for the Philosophy of Science, 60(2), 345–375.
Jaynes, E. T. (1957). Information theory and statistical mechanics. The Physical Review, 106(4), 620–630.
Kemp, C., Gaby, A., & Regier, T. (2019). Season naming and the local environment (pp. 539–545). In CogSci: The annual meeting of the cognitive science society.
Keynes, J. M. (1921). A treatise on probability. London: Macmillan.
LaCroix, T. (2019). Communicative bottlenecks lead to maximal information transfer. Journal of Experimental and Theoretical Artificial Intelligence. Forthcoming.
Lewis, D. (2002/1969). Convention: A philosophical study. Blackwell, Oxford.
Lupyan, G., & Dale, R. (2010). Language structure is partly determined by social structure. PloS One, 5(1), e8559.
Lupyan, G., & Dale, R. (2016). Why are there different languages? The role of adaptation in linguistic diversity. Trends in Cognitive Sciences, 20(9), 649–660.
Neth, S. (2019). Bayesian naturalness. Unpublished Manuscript. April, 2019. PDF File.
O’Connor, C. (2017). Games and kinds. The British Journal for the Philosophy of Science, 70(3), 719–745.
Purves, G. M. (2018). Conventionality and causality in lewis-type evolutionary prediction games. The British Journal for the Philosophy of Science, axy076. Forthcoming.
Rescorla, R. A., & Wagner, A. R. (1972). A theory of Pavlovian conditions: Variations in the effectiveness of reinforcement and nonreinforcement. In A. H. Black & W. F. Prokasy (Eds.), Classical conditioning II: Current research and theory (pp. 64–99). New York: Appleton-Century-Crofts.
Resnik, D. B. (1991). How-possibly explanations in biology. Acta Biotheoretica, 39(2), 141–149.
Roth, A., & Erev, I. (1995). Learning in extensive form games: Experimental data and simple dynamical models in the intermediate term. Games and Economic Behavior, 8, 164–212.
Schlenker, P., Chemla, E., Schel, A. M., Fuller, J., Gautier, J.-P., Kuhn, J., et al. (2016). Formal monkey linguistics. Theoretical Linguistics, 42(1–2), 1–90.
Schultz, W. (2004). Neural coding of basic reward terms of animal learning theory, game theory, micro-economics and behavioural ecology. Current Opinion in Neurobiology, 14(2), 139–147.
Schultz, W., Dayan, P., & Montague, P. R. (1997). A neural substrate of prediction and reward. Science, 275, 1593–1599.
Seyfarth, R. M., Cheney, D. L., & Marler, P. (1980a). Monkey responses to three different alarm calls: Evidence of predator classification and semantic communication. Science, 210, 801–803.
Seyfarth, R. M., Cheney, D. L., & Marler, P. (1980b). Vervet monkey alarm calls: Semantic communication in a free-ranging primate. Animal Behaviour, 28(4), 1070–1094.
Shannon, C. (1948). A mathematical theory of communication. The Bell System Mathematical Journal, 27, 379–423.
Shannon, C., & Weaver, W. (1949). The mathematical theory of communication. Urbana, Chicago: University of Illinois Press.
Skyrms, B. (2006). Signals. Philosophy of Science, 75(5), 489–500.
Skyrms, B. (2010a). Signals: Evolution, learning, and information. Oxford: Oxford University Press.
Skyrms, B. (2010b). The flow of information in signaling games. Philosophical Studies, 147(1), 155–165.
Sutton, R. S., & Barto, A. G. (1998). Reinforcement learning: An introduction. Cambridge: MIT Press.
van Fraassen, B. (1989). Laws and symmetry. Oxford: Clarendon Press.
Williamson, J. (2010). In defence of objective Bayesianism. Oxford: Oxford University Press.
Acknowledgements
Both authors would like to thank the reviewers for comments on an earlier draft of the paper. They would also like to thank Brian Skyrms for insightful conversations on the nature of information in signaling games. LaCroix would like to thank Charles Reiss and the audience at CRSP (Winter 2020) in Montréal. LaCroix would additionally like to thank Mila - Québec Artificial Intelligence Institute and, in particular, Yoshua Bengio for providing generous space and resources.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Barrett, J.A., LaCroix, T. Epistemology and the Structure of Language. Erkenn 87, 953–967 (2022). https://doi.org/10.1007/s10670-020-00225-4
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10670-020-00225-4