Language Evolution and Computation Bibliography

Natural language, communication or related mental phenomena must surely be a prominent candidate for an evolutionary explanation. This paper discusses a primitive model of language generation by evolution and learning among a population of artificial organisms whose brains are realized by a model of associative memory with a neural network structure. The goal of our study is to acquire general knowledge of the theory that relates the mechanisms to the evolutionary process such as language generation, and to develop the evolutionary systems which have facilities for still more intelligent information processing.

I present a simple game, the Basic Action-Response game, which allows investigation of the claim that signals must be costly to be reliable. The Basic Action-Response game is the simplest communication game possible, by investigating its parameters we are able to define clearly ``conflict'', ``handicap'', ``communication'' and other relevant concepts. I explore the conditions on the magnitude of the stabilizing cost and handicap that must hold in order to maintain the evolutionary stability of signalling. It will be demonstrated that stable communication need not make use of costly signals at ESS, not even ``on average'', and that ``negative handicaps'' can be stable as long as the stabilizing cost is large enough.

Conflicts of interest arise between signaller and receiver in most kinds of biological communication. Some authors have argued that this conflict is likely to give rise to deceit and exploitation, as receivers lag behind in the coevolutionary 'arms race' with signallers. Others have argued that such manipulation is likely to be short-lived and that receivers can avoid being deceived by paying attention to signals that are costly and hence 'unfakeable.' These two views have been hard to reconcile. Here, we present results from simulations of signal evolution using artificial neural networks, which demonstrate that honesty can coexist with a degree of exploitation. Signal cost ensures that receivers are able to obtain some honest information, but is unable to prevent exploitative signalling strategies from gaining short-term benefits. Although any one receiver bias that is open to exploitation will subsist for only a short period of time once signallers begin to take advantage of it, new preferences of this kind are constantly regenerated through selection and random drift. Hidden preferences and sensory exploitation are thus likely to have an enduring influence on the evolution of honest, costly signals. At the same time, honesty and cost are prerequisites for the evolution of exploitation. When signalling is cost-free, selection cannot act to maintain honesty, and receivers rapidly evolve to ignore signals. This leads to a reduction in the extent of hidden preference, and a consequent loss of potential for exploitation.

Language is a shared set of conventions for mapping meanings to utterances. This paper explores self-organization as the primary mechanism for the formation of a vocabulary. It reports on a computational experiment in which a group of distributed agents develop ways to identify each other using names or spatial descriptions. It is also shown that the proposed mechanism copes with the acquisition of an existing vocabulary by new agents entering the community and with an expansion of the set of meanings.

An account is offered to change over time in English verb morphology, based on a connectionist approach to how morphological knowledge is acquired and used. A technique is first described that was developed for modeling historical change in connectionist networks, and that technique is applied to model English verb inflection as it developed from the highly complex past tense system of Old English towards that of the modern language, with one predominant ``regular'' inflection and a small number of irregular forms. The model relies on the fact that certain input-output mappings are easier than others to learn in a connectionist network. Highly frequent patterns, or those that share phonological regularities with a number of others, are learned more quickly and with lower error than low-frequency, highly irregular patterns. A network is taught a data set representative of the verb classes of Old English, but learning is stopped before reaching asymptote, and the output of this network is used as the teacher of a new net. As a result, the errors in the first network were passed on to become part of the data set of the second. Those patterns that are hardest to learn led to the most errors, and over time are ``regularized'' to fit a more dominant pattern. The results of the networks simulations were highly consistent with the major historical developments. These results are predicted from well-understood aspects of network dynamics, which therefore provide a rationale for the shape of the attested changes.

Models of animal signalling stress that among unrelated individuals the transfer of honest information normally requires that signals are costly, and costly in a way related to the true information revealed by the signal. In the absence of such a cost, `cheats', that lie about their states or needs, are able to evolve and exploit the preferences of receivers. We show here that spatial constraints imposed on the interactions between signallers and receivers favour honest signalling even in the absence of any costs: `islands' of honesty coexist in `seas' of dishonesty. The extent to which honest or dishonest strategies are favoured, is shown to depend upon the relative payoffs from signalling and receiving. As the receiving component of fitness becomes greater than the signalling component of fitness, as might be true in `life-dinner' type interactions, honesty is increasingly favoured. In addition, in spatial populations, honesty can be favoured locally even when the mean global payoffs to honesty are lower than the mean payoffs to dishonesty. Our model provides a general framework for analysing signals in spatially structured populations and might therefore apply to signalling in both natural and cultural situations.

Much animal communication takes place via symbolic codes, where each symbol's meaning is fixed by convention only and not by intrinsic meaning. It is unclear how understanding can arise among individuals utilizing such arbitrary codes, and specifically, whether evolution unaided by individual learning is sufficient to produce such understanding. Using a genetic algorithm implemented on a computer, I demonstrate that a significant though imperfect level of understanding can be achieved by organisms through evolution alone. The population as a whole settles on one particular scheme of coding/decoding information (there are no separate dialects). Several features of such evolving systems are explored and it is shown that the system as a whole is stable against perturbation along many different kinds of ecological parameters.

In this paper, we elaborate upon the framework by considering more explicitly the problem of creating shared symbolic structure. A lexicon is (among other things) a set of public structures for denoting or implicating shared meanings. In other words, the existence of a lexicon requires the sharing of forms and meanings -- and mappings between these -- among members of an interacting population of agents. Leaving aside cosmic and theological events which could create such an outcome, how could a lexicon come to be? The solution provided here is based upon a convergence of agents' schemes for classifying visual phenomena. All agents have a capacity for such classification, but convergence upon a singular scheme is shaped by the constraints for consensus when employing the scheme in interaction with other agents.