Roland Mühlenbernd, Slawomir Wacewicz and Przemyslaw Zywiczynski

Introduction
Our study sheds new light on a key feature in human language: ambiguity. A variety of forms of
ambiguity are an essential part of natural language, including lexical ambiguity, semantic and syntactic
ambiguity, or the related concept of vagueness (cf. Brochhagen 2017). Moreover, it can be shown that
ambiguity plays an important role not only in human language, but also in animal communication (cf.
Arnold & Zuberbühler 2006). To investigate factors that may support the emergence and stability of
ambiguity in communication systems, we have worked out the following three steps: (1.) We developed
a game-theoretic model of communication that enables a thorough evolutionary game theoretic (EGT)
analysis pointing to conditions that support the evolutionary stability of ambiguous communication in
populations of communicating individuals. (2.) Based on this game-theoretic model, we implemented a
multi-agent model of interacting and replicating agents to simulate evolutionary processes of emerging
communication systems. The simulation results are informative on the probabilities of emergence of
more vs less ambiguous communication systems. (3.) We conducted online-experiments where multiple
participants interact by repeatedly playing a communication game. We recorded the emergent communication
systems and contrasted them with the results from the formal and computational analyses.

Background
A very popular formal model for studying communication systems is the signaling game (Lewis 1969),
which models communication between a sender and a receiver. Sender and receiver strategies depict
encoding–decoding processes between information states and signals. By studying signaling games
under evolutionary dynamics, it is possible to show that unambiguous signaling strategies are evolutionarily
stable, whereas ambiguous strategies cannot survive under evolutionary dynamics (Wärneryd
1993), given basic conditions, such as aligned interests of interlocutors or no signal costs. However, it
can be shown that changing these conditions can make ambiguity evolutionarily stable (cf. Santana
2014, O’Connor 2015, Mühlenbernd 2020). The predictions about the emergence and stability of strategies
in signaling games are mostly a result of formal analysis and simulation studies (cf. Huttegger
2007, Skyrms 2010). In recent years, a new research trend emerged that appeared to be a promising
way to explore and validate such formal predictions: the study of communication games in the
laboratory, in particular by applying tools from Experimental Semiotics (cf. Galantucci 2017) and
Experimental Economics (cf. Bruner et al. 2019). However, to our knowledge no experimental work
exists that studies the evolution of ambiguity in communication systems.

Work and results
We study a minimal example of a game theoretic communication model that integrates the idea of
contextual cues helping the receiver to disambiguate between meanings. As a first step, an EGT
analysis reveals that there are three types of communication strategies that are evolutionarily stable:
signaling systems (SI), partially ambiguous systems (PA), fully ambiguous systems (FA). Second, a
computational analysis shows that of these systems, PA systems have the highest emergence frequency
under most conditions. And third, online experiments point to additional conditions under which
communication systems emerge that reproduce or deviate from formal/computational predictions.

References
Arnold & Zuberbühler (2006). Language evolution: semantic combinations in primate calls. Nature
441(7091): 303.
Brochhagen (2017). Signaling under uncertainty: Interpretative alignment without a common prior. The
British Journal for the Philosophy of Science, 71(2), 471–496.
Bruner, O’Connor & Rubin (2019). Experimental Economics for Philosophers. In: E. Fischer, M. Curtis
(eds.), Methodological Advances in Experimental Philosophy, Bloomsbury Academic.
Galantucci (2017). Experimental Semiotics. Oxford Research Encyclopedia of Linguistics.
Huttegger (2007). Evolution and the explanation of meaning. Philosophy of Science 74, 1-27.
Lewis (1969). Convention. A Philosophical Study. Wiley-Blackwell.
Mühlenbernd (2020). Evolutionary stability of ambiguity in context signaling games. Synthese, online
first.
O’Connor (2015). Ambiguity is kinda good sometimes. Philosophy of Science 82(1), 110-121.
Santana (2014). Ambiguity in cooperative signaling. Philosophy of Science 81(3), 398-422.
Skyrms (2010). Signals: Evolution, Learning and Information. Oxford University Press.
Wärneryd (1993). Cheap Talk, Coordination, and Evolutionary Stability. Games and Economic
Behavior 5(4), 532-546.