The evolution of ambiguity

Roland Mühlenbernd, Management Department, Ca'Foscari University of Venice, Italy

I present game-theoretic communication models that describe language use in the form of signaling systems. In such models, the utility/fitness of a signaling system is defined by its communicative success. It can be shown that a signaling system is optimal when it forms a one-to-one mapping between expressions and meanings. However, human language is not optimal in this respect, not least because it is ambiguous to a large extent, and ambiguity forms a one-to-many mapping between expressions and meanings. It can be shown that under basic conditions, ambiguous signaling systems are not evolutionarily stable and therefore cannot survive under evolutionary dynamics. I will present a number of formal studies that point out conditions which make ambiguity evolutionarily stable. Finally, I will suggest some ideas on how to test those conditions experimentally by applying tools from Experimental Economics.

Monotonicity: between simplicity and expressiveness

Dariusz Kalociński, University of Warsaw, Poland

Monotonicity is an abstract property of meaning that has been attested in several semantic domains, including adjectives, quantifiers and modals. So far it has been shown that monotonicity facilitates learning (Chemla et al., 2019, Steinert-Threlkeld & Szymanik, forthcoming) and arises via artificial iterated learning with pragmatic agents biased towards simplicity and expressiveness (Carcassi et al., forthcoming). I will attempt to explain monotonicity assuming a domain-general economy principle according to which humans tend to optimize communicative and cognitive costs associated with a language (Kemp & Regier, 2012). The results suggest that the tradeoff between these two factors might be sufficient to explain monotonicity. Moreover, the generality of this argument indicates that monotonicity might arise at various timescales for which such optimization is viable. We backup this conclusion with initial simulations based on a recent model of meaning coordination (Kalociński et al., 2018).

Zipf’s law of abbreviation, compression and efficiency of coding in primate vocal communication

Stuart Semple, University of Roehampton, UK

Zipf’s law of abbreviation predicts a negative relationship between word length and frequency of use, and holds in a wide range of languages. This law has been mathematically linked to compression, and it has been argued that this phenomenon reflects a universal pressure for coding efficiency. Recent studies have tested whether patterns consistent with this law are also seen in the vocal communication systems of primate species. Agreement with the law has been found in the vocal repertoire of Formosan macaques, in a subset of vocalisations of marmosets, but not in the vocalisations of uakaris. In this talk, I will present results of tests of Zipf’s law of abbreviation in vocalisations of a wide range of primate species not previously considered in this respect. In addition, using new methods to quantify the efficiency of these vocal communication systems, I will present results of additional tests for the occurrence of compression.

Testing shared language as a source of trust in an investment game

Magdalena Schwarz, Theresa Matzinger & Nikolaus Ritt, University of Vienna, Austria

Languages are ideal social markers. Salient and hard to fake, they help us identify group members [1, 2]. We hypothesize that sharing a language increases trust. We test this in an established paradigm for measuring trust, the investment game [3]: the first player receives some money and may entrust a share to the second player. The shared sum is tripled and a part of that may be returned to the first player. In our pilot study, participants played the investment game against what they believed was a player in another room, whom they heard speak in a local dialect. We had two groups of participants. The first group shared a similar dialect with the other player; the second one did not. We predict the first group to entrust their partner with a greater share of money than the second group. This would support an evolutionary link between language and trust.

Rational Griceans are vague

Michael Franke, Institute for Cognitive Science, University of Osnabrück, Germany

I present a model of meaning evolution in which pragmatic agents (formalized in the Rational Speech Act tradition) communicate with antonym pairs of gradable adjectives, and their respective negations. Agents can have different degrees of rationality. Adjectives can have variable degrees of vagueness. I show that evolutionary pressure for efficient communication selects rational Gricean language use paired with a mild degree of semantic vagueness.

A cost-benefit approach in everyday language use

Slawomir Wacewicz, Przemyslaw Zywiczynski, Centre for Language Evolution Studies, Nicolaus Copernicus University in Toruń, Poland

We sketch out a framework for reclassifying some language use traditionally subsumed under Linguistic Politeness (LP) as strategic language use. Although saying “thank you” or “sorry” costs virtually no time or effort, it can often be used in lieu of money as a virtual payment or compensation, leading to questions about the nature of its purchasing power. Goffman (1967) and Brown and Levinson (1987) suggest that polite language has purchasing power because it converts into a valuable but nebulous social resource “Face”. Following recent work, and especially Chaudhry and Loewenstein (2019), we present a more specific analysis of that social resource in terms of two components: one related to a person’s position in the hierarchy (“Prestige”) and the other to a person’s trustworthiness in observing the norm of reciprocity (“Obligation”). We discuss implications for theories of cooperation.

Sponsoring Institutions