Cover not available
Article published In:
Interaction Studies
Vol. 22:2 (2021) ► pp.212–243

Interaction history as a source of compositionality in emergent communication

| University of Sussex | University of Warsaw
| University of Warsaw
| Polish Academy of Sciences
| Polish Academy of Sciences | University of Oxford
| University of Warsaw

In this paper, we explore interaction history as a particular source of pressure for achieving emergent compositional communication in multi-agent systems. We propose a training regime implementing template transfer, the idea of carrying over learned biases across contexts. In the presented method, a sender-receiver dyad is first trained with a disentangled pair of objectives, and then the receiver is transferred to train a new sender with a standard objective. Unlike other methods (e.g. the obverter algorithm), the template transfer approach does not require imposing inductive biases on the architecture of the agents. We experimentally show the emergence of compositional communication using topographical similarity, zero-shot generalization and context-independence as evaluation metrics. The presented approach is connected to an important line of work in semiotics and developmental psycholinguistics: it supports a conjecture that compositional communication is scaffolded on simpler communication protocols.

Keywords: emergent communication, path-dependence, compositionality, Lewis signaling games, generalization
Article outline
  • 1.Introduction
  • 2.Related work
    • Inductive biases for compositional communication
    • Obverter approach
    • Population-based training
    • Multi-task training
    • Evolutionary origins of grammar
    • Generalized signaling games
    • Motivation for our approach
  • 3.Method
    • 3.1Experimental setup
      • 3.1.1Dataset
      • 3.1.2Object naming game
    • 3.2Architecture of the agents
      • General setup
      • Vision module
      • Sender
      • Receiver
      • Hyperparameters
    • 3.3Template transfer approach
  • 4.Experiments and results
    • 4.1Measuring compositionality
      • Test set accuracy
      • Context-independence
      • Topographical similarity
    • 4.2Baselines
      • Random baseline
      • Same architecture
      • Obverter baseline
    • 4.3Results
    • 4.4Visualizing the activations of the receiver
  • 5.Discussion
    • Sources of compositionality
    • Evolutionary game-theoretic interpretation
    • Semiotic interpretation
    • Cognitive interpretation
    • Developmental interpretation
  • 6.Conclusions
    • Limitations
  • Acknowledgements
  • Notes
  • References
Available under the Creative Commons Attribution (CC BY) 4.0 license.

For any use beyond this license, please contact the publisher at [email protected].

Published online: 28 February 2022
DOI logo
https://doi.org/10.1075/is.21020.kor
References
Barrett, J., Skyrms, B., and Cochran, C.
(2018) Hierarchical Models for the Evolution of Compositional Language. Institute for Mathematical Behavioral Sciences Technical Report MBS 18-03.Google Scholar
Barrett, J. A. and Skyrms, B.
(2017) Self-assembling Games. The British Journal for the Philosophy of Science, 68(2):329–353. DOI logoGoogle Scholar
Barrett, J. A., Skyrms, B., and Mohseni, A.
(2019) Self-Assembling Networks. The British Journal for the Philosophy of Science, 70(1):301–325. DOI logoGoogle Scholar
Batali, J.
(1998) Computational simulations of the emergence of grammar. Approach to the Evolution of Language, pages 405–426.Google Scholar
Bogin, B., Geva, M., and Berant, J.
(2018) Emergence of Communication in an Interactive World with Consistent Speakers. arXiv:1809.00549 [cs]. arXiv: 1809.00549.Google Scholar
Bouchacourt, D. and Baroni, M.
(2018) How agents see things: On visual representations in an emergent language game. arXiv:1808.10696 [cs]. arXiv: 1808.10696. DOI logoGoogle Scholar
Brighton, H.
(2002) Compositional Syntax From Cultural Transmission. Artificial Life, 8(1):25–54. DOI logoGoogle Scholar
Brighton, H. and Kirby, S.
(2006) Understanding Linguistic Evolution by Visualizing the Emergence of Topographic Mappings. Artificial Life, 12(2):229–242. DOI logoGoogle Scholar
Brown, P. F., Della-Pietra, S. A., Della-Pietra, V. J., and Mercer, R. L.
(1993) The mathematics of statistical machine translation. Computational Linguistics, 19(2):263–313.Google Scholar
Bruner, J. S.
(1981) Intention in the structure of action and interaction. Advances in Infancy Research, 11:41–56.Google Scholar
(1983) Child’s talk: learning to use language. W.W. Norton, New York, 1st ed edition.Google Scholar
Cangelosi, A.
(2001) Evolution of communication and language using signals, symbols, and words. IEEE Transactions on Evolutionary Computation, 5(2):93–101. DOI logoGoogle Scholar
Chaabouni, R., Kharitonov, E., Bouchacourt, D., Dupoux, E., and Baroni, M.
(2020) Compositionality and Generalization In Emergent Languages. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 4427–4442, Online. Association for Computational Linguistics. DOI logoGoogle Scholar
Choi, E., Lazaridou, A., and de Freitas, N.
(2018) Compositional Obverter Communication Learning From Raw Visual Input. ICLR 2018. arXiv: 1804.02341.Google Scholar
Cogswell, M., Lu, J., Lee, S., Parikh, D., and Batra, D.
(2019) Emergence of Compositional Language with Deep Generational Transmission. ArXiv, abs/1904.09067.Google Scholar
Cowley, S. J.
editor (2011) Distributed language. Number v. 34 in Benjamins current topics. John Benjamins Pub. Co, Amsterdam; Philadelphia. OCLC: ocn741355729. DOI logoGoogle Scholar
Das, A., Kottur, S., Moura, J. M. F., Lee, S., and Batra, D.
(2017) Learning Cooperative Visual Dialog Agents with Deep Reinforcement Learning. 2017 IEEE International Conference on Computer Vision (ICCV), Venice, 2017 1. arXiv: 1703.06585. DOI logoGoogle Scholar
De Beule, J. and Bergen, B. K.
(2006) On the emergence of compositionality. In The Evolution of Language, pages 35–42, Rome, Italy. World scientific. DOI logoGoogle Scholar
Deacon, T. W.
(1998) The symbolic species: the co-evolution of language and the brain. Norton, New York, NY, norton paperback edition. OCLC: 254499872.Google Scholar
Elman, J. L.
(1990) Finding Structure in Time. Cognitive Science, 14(2):179–211. DOI logoGoogle Scholar
Foerster, J. N., Assael, Y. M., de Freitas, N., and Whiteson, S.
(2016) Learning to Communicate with Deep Multi-Agent Reinforcement Learning. NIPS’16 Proceedings of the 30th International Conference on Neural Information Processing Systems. arXiv: 1605.06676.Google Scholar
Gong, T., Shuai, L., and Zhang, M.
(2014) Modelling language evolution: Examples and predictions. Physics of Life Reviews, 11(2):280–302. DOI logoGoogle Scholar
Goodfellow, I., Bengio, Y., and Courville, A.
(2016) Deep learning. Adaptive computation and machine learning. The MIT Press, Cambridge, Massachusetts.Google Scholar
Grouchy, P., D’Eleuterio, G. M. T., Christiansen, M. H., and Lipson, H.
(2016) On The Evolutionary Origin of Symbolic Communication. Scientific Reports, 6:34615. DOI logoGoogle Scholar
Jaques, N., Lazaridou, A., Hughes, E., Gulcehre, C., Ortega, P. A., Strouse, D. J., Leibo, J. Z., and de Freitas, N.
(2018) Social Influence as Intrinsic Motivation for Multi-Agent Deep Reinforcement Learning. arXiv:1810.08647 [cs, stat]. arXiv: 1810.08647.Google Scholar
Kharitonov, E. and Baroni, M.
(2020) Emergent Language Generalization and Acquisition Speed are not tied to Compositionality. arXiv:2004.03420 [cs]. arXiv: 2004.03420. DOI logoGoogle Scholar
Kharitonov, E., Chaabouni, R., Bouchacourt, D., and Baroni, M.
(2019) EGG: a toolkit for research on Emergence of lanGuage in Games. arXiv:1907.00852 [cs]. arXiv: 1907.00852. DOI logoGoogle Scholar
Kingma, D. P. and Ba, J.
(2014) Adam: A Method for Stochastic Optimization. arXiv:1412.6980 [cs]. arXiv: 1412.6980.Google Scholar
Kirby, S.
(2001) Spontaneous evolution of linguistic structure-an iterated learning model of the emergence of regularity and irregularity. IEEE Transactions on Evolutionary Computation, 5(2):102–110. DOI logoGoogle Scholar
Kirby, S., Cornish, H., and Smith, K.
(2008) Cumulative cultural evolution in the laboratory: An experimental approach to the origins of structure in human language. Proceedings of the National Academy of Sciences, 105(31):10681–10686. DOI logoGoogle Scholar
Korbak, T., Zubek, J., and Raczaszek-Leonardi, J.
(2020) Measuring non-trivial compositionality in emergent communication. arXiv preprint arXiv:2010.15058.Google Scholar
Kottur, S., Moura, J. M. F., Lee, S., and Batra, D.
(2017) Natural Language Does Not Emerge ‘Naturally’ in Multi-Agent Dialog. arXiv:1706.08502 [cs]. arXiv: 1706.08502. DOI logoGoogle Scholar
Kriegeskorte, N.
(2008) Representational similarity analysis – connecting the branches of systems neuroscience. Frontiers in Systems Neuroscience. DOI logoGoogle Scholar
Kuciński, L., Korbak, T., Kolodziej, P., and Miloś, P.
(2021) Catalytic role of noise and necessity of inductive biases in the emergence of compositional communication. In Advances in Neural Information Processing Systems.Google Scholar
Lake, B. M., Ullman, T. D., Tenenbaum, J. B., and Gershman, S. J.
(2016) Building Machines That Learn and Think Like People. arXiv:1604.00289 [cs, stat]. arXiv: 1604.00289.Google Scholar
Lazaridou, A., Hermann, K. M., Tuyls, K., and Clark, S.
(2018) Emergence of Linguistic Communication from Referential Games with Symbolic and Pixel Input. arXiv:1804.03984 [cs]. arXiv: 1804.03984.Google Scholar
Lazaridou, A., Peysakhovich, A., and Baroni, M.
(2016) Multi-Agent Cooperation and the Emergence of (Natural) Language. arXiv:1612.07182 [cs]. arXiv: 1612.07182.Google Scholar
LeCun, Y., Bottou, L., Bengio, Y., and Haffner, P.
(1998) Gradient-based learning applied to document recognition. In PROCEEDINGS OF THE IEEE, pages 2278–2324. DOI logoGoogle Scholar
Levenshtein, V. I.
(1966) Binary Codes Capable of Correcting Deletions, Insertions and Reversals. Soviet Physics Doklady, 101:707.Google Scholar
Lewis, D. K.
(1969) Convention: a philosophical study. Blackwell, Oxford, nachdr. edition. OCLC: 837747718.Google Scholar
Li, F. and Bowling, M.
(2019) Ease-of-Teaching and Language Structure from Emergent Communication. arXiv:1906.02403 [cs]. arXiv: 1906.02403.Google Scholar
Lowe, R., Foerster, J., Boureau, Y.-L., Pineau, J., and Dauphin, Y.
(2019) On the Pitfalls of Measuring Emergent Communication. arXiv:1903.05168 [cs, stat]. arXiv: 1903.05168.Google Scholar
Nomikou, I., Koke, M., and Rohlfing, K. J.
(2017) Verbs in Mothers’ Input to Six-Month-Olds: Synchrony between Presentation, Meaning, and Actions Is Related to Later Verb Acquisition. Brain Sciences, 7(12):52. DOI logoGoogle Scholar
Nowak, M. A. and Krakauer, D. C.
(1999) The evolution of language. Proceedings of the National Academy of Sciences, 96(14):8028–8033. Publisher: National Academy of Sciences Section: Biological Sciences. DOI logoGoogle Scholar
Oliphant, M. and Batali, J.
(1997) Learning and the Emergence of Coordinated Communication. Center for Research on Language Newsletter, 111.Google Scholar
Paszke, A., Gross, S., Chintala, S., Chanan, G., Yang, E., DeVito, Z., Lin, Z., Desmaison, A., Antiga, L., and Lerer, A.
(2017) Automatic differentiation in PyTorch.Google Scholar
Peirce, C. S.
(1998) Collected papers of Charles Sanders Peirce. Thoemmes Press, Bristol, England. OCLC: ocm39692049.Google Scholar
Raczaszek-Leonardi, J.
(2012) Language as a System of Replicable Constraints. In LAWS, LANGUAGE and LIFE, volume 71, pages 295–333. Springer Netherlands, Dordrecht. DOI logoGoogle Scholar
Raczaszek-Leonardi, J., Nomikou, I., Rohlfing, K. J., and Deacon, T. W.
(2018) Language Development From an Ecological Perspective: Ecologically Valid Ways to Abstract Symbols. Ecological Psychology, 30(1):39–73. DOI logoGoogle Scholar
Rainey, H. J., Zuberbühler, K., and Slater, P. J. B.
(2004) Hornbills can distinguish between primate alarm calls. Proceedings of the Royal Society of London. Series B: Biological Sciences, 271(1540):755–759. DOI logoGoogle Scholar
Rumelhart, D. E., Hinton, G. E., and Williams, R. J.
(1986) Learning representations by back-propagating errors. Nature, 323(6088):533–536. DOI logoGoogle Scholar
Ryan, M. and Rand, S.
(1993) Sexual selection and signal evolution: the ghost of biases past. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences, 340(1292):187–195. DOI logoGoogle Scholar
Skyrms, B.
(2010) Signals: evolution, learning, & information. Oxford University Press, Oxford; New York. OCLC: ocn477256653. DOI logoGoogle Scholar
Steinert-Threlkeld, S.
(2020) Towards the Emergence of Non-trivial Compositionaliy. Philosophy of Science. DOI logoGoogle Scholar
Stern, D. N.
(1974) The Goal and Structure of Mother-Infant Play. Journal of the American Academy of Child Psychiatry, 13(3):402–421. DOI logoGoogle Scholar
Tomasello, M., Carpenter, M., Call, J., Behne, T., and Moll, H.
(2005) Understanding and sharing intentions: The origins of cultural cognition. Behavioral and Brain Sciences, 28(5):675–691. DOI logoGoogle Scholar
Wittgenstein, L.
(1953) Philosophical Investigations. Basil Blackwell, Oxford.Google Scholar
Yamashita, Y. and Tani, J.
(2008) Emergence of Functional Hierarchy in a Multiple Timescale Neural Network Model: A Humanoid Robot Experiment. PLoS Computational Biology, 4(11):e1000220. DOI logoGoogle Scholar