Portal de Periódicos da CAPES

The Role of Imitation in Generating a Shared Communication System

2012; Wiley; Volume: 34; Issue: 34 Linguagem: Inglês

1551-6709

Junya Morita, Takeshi Konno, Takashi Hashimoto,

Multi-Agent Systems and Negotiation

The Role of Imitation in Generating a Shared Communication System Junya Morita, Takeshi Konno, Takashi Hashimoto {j-morita,t-konno,hash}@jaist.ac.jp School of Knowledge Science, Japan Advanced Institute of Science and Technology 1-1 Asahidai, Nomi, Ishikawa 923-1292, Japan that a type of imitation called “role-reversal”, in which the child aligns herself with the adult speaker, is essential for pro- ducing a communicative symbols. However, there is an important difference between lan- guage acquisition and forming a new communication system. In language acquisition, there is a clear distinction between a learner and an instructor (or a demonstrator). On the other hand, a new communication system usually emerges from a situation where no predefined roles exist. Although several studies concerning imitation exist, only few studies have dealt with this situation. We argue that the model-based approach is the best method to explore the role of imitation in an interactive situation. From this perspective, we present a computational model, in which two agents autonomously assign roles to themselves. Agents also posses general learning mechanisms such as rein- forcement learning and instance-based learning to form a new communication system. By incorporating imitative learning into these general learning mechanisms, we investigate the role of imitation in the process of forming a new communica- tion system. Furthermore, we compare the constructed model against a human experiment. The results of this comparison reveal the cognitive mechanism involved in the formation of a human communication system. Before presenting our model, we will provide an overview of our previous experiment. Abstract What types of learning or reasoning are involved in form- ing a new communication system? To answer this question, this paper presents a computational model for forming a new communication system. The model was developed with ACT- R (Adaptive Control of Thought-Rational). In the model, two agents autonomously assign their roles to themselves. Agents also posses general learning mechanisms implemented in ACT-R. By incorporating imitative learning into these gen- eral learning mechanisms, this paper studies the role of imita- tion in the process of forming a new communication system. Finally, we compared the proposed model against a human ex- periment. The results of the simulation indicate that through imitation, after a short period of interaction, an isomorphic system is created. The result of the simulation also suggests the existence of imitation in the process of forming a human communication system. Keywords: Communication; Imitation; ACT-R Introduction People try to communicate with others even when they do not have a common language. They also understand inten- tions of others through repeated interactions. Apparently, hu- mans have the ability to develop a new communication sys- tem where only a few common ground rules are shared in advance. How can a new communication system be devel- oped? What types of learning or reasoning are involved in this process? Addressing these questions will contribute not only to understand the origins of our communication but also to predict changes in our communication in this era of glob- alization. Some researchers have examined these questions by de- signing communication environments in the laboratory (for a review Scott-Phillips & Kirby, 2010). For example, Galantucci (2005) conducted an experiment to observe the formation of communication systems in which a pair of par- ticipants communicated through a medium that restricted the use of standard communication means such as utterances and letters. He observed the process of forming a new commu- nication system, and discussed that implicit information was conveyed through routine behavior, and reported that a tem- poral order of messages was built into communication sys- tems. However, this study cannot answer the questions above because he did not identify the cognitive mechanism involved in this process. The present study focuses on imitation as the type of reasoning in forming a new communication system. Im- itation has been investigated in various fields of cognitive science (Barnes & Thagard, 1997; Gergely, Bekkering, & Kir´aly, 2002; Tomasello, 1999; Thagard, 2001). For exam- ple, Tomasello (1999) described the role of imitation in lan- guage acquisition by the infant/child. He especially argued Experiment The present study simulates the experiment reported in Konno, Morita, and Hashimoto (in press), where we mod- ified and used a coordination game taken from Galantucci (2005). As in Galantucci’s study, the game environment con- tained two characters, each controlled by a players, and four intercommunicating rooms. The game was composed of sev- eral repeated rounds. At the beginning of a round, charac- ters were randomly placed in two different rooms. Players were unaware of the location of each other and aimed to bring their characters to the same room. The characters could not move to rooms that were located diagonally. Owing to this constraint, players need to communicate before moving their characters. Figure 1 presents the flow of each round consisting of three steps: step 1 for exchanging messages; step 2 for moving characters; and step 3 for confirming the result of the move- ment. Among these steps, step 1 is the most crucial for the success of this task. In this step, the two players construct their own messages composed of two figures such as “ using six available figures: , , , , , and . The meanings and usage of the figures were not shared among