网范文：“Multi-Agent Connectionist Approach to Communication ”多智能体联结主义提出了一个个体反复网络通信,在个人周期性网络模拟信息吸收的过程中,同时代理通信的信仰和观点传播，以及个人之间的联系网络。这篇社会范文讲述了多代理联结主义的交流与传播。在模型中,信任是由与代理接收一致性的信仰,并导致个体之间的连接网络变化,称为信任权重。从而激活扩散，英语论文范文，类似于标准的联结主义的过程,虽然信任权重采取特定的功能。
具体地说,它们会导致选择性不可靠信息的传播,从而过滤掉,交际不良的内在机制处理，英语论文题目，主要涉及有说服力的沟通和极化现象,刻板印象和谣言的传播,和缺乏分享独特的信息集团决策。认知并不局限于单个代理的思想,但涉及到与其他思想的交互。因此,全面理解人类思维，要求了解其社会发展。下面的范文进行详述。
Abstract
A multi-agent connectionist model is proposed that consists of a collection of individual recurrent networks that communicate with each other, and as such is a network of networks. The individual recurrent networks simulate the process of information uptake, integration and memorization within individual agents, while the communication of beliefs and opinions between agents is propagated along connections between the individual networks. A crucial aspect in belief updating based on information from other agents is the trust in the information provided. In the model, trust is determined by the consistency with the receiving agents’ existing beliefs, and results in changes of the connections between individual networks, called trust weights. Thus activation spreading and weight change between individual networks is analogous to standard connectionist processes, although trust weights take a specific function. Specifically, they lead to a selective propagation and thus filtering out of less reliable information, and they implement Grice’s (1975) maxims of quality and quantity in communication. The unique contribution of communicative mechanisms beyond intra-personal processing of individual networks was explored in simulations of key phenomena involving persuasive communication and polarization, lexical acquisition, spreading of stereotypes and rumors, and a lack of sharing unique information in group decisions.
Cognition is not limited to the mind of an individual agent, but involves interactions with other minds. A full understanding of human thinking thus requires insight into its social development. Sociologists and developmental psychologists have long noted that most of our knowledge of reality is the result of communication and social construction rather than of individual observation. Moreover, important real-world problems are often solved by a group of communicating individuals who pool their individual expertise while forming a collective decision that is supposedly better than what they might have achieved individually. This phenomenon may be labeled collective intelligence (Levy, 1997; Heylighen, 1999) or distributed cognition (Hutchins, 1995). To understand such group-level information processing, we must consider the distributed organization constituted by different individuals with different forms of knowledge and experience together with the social network that connects them and that determines which information is exchanged with whom.
In addition to qualitative observations and conceptualizations, the phenomenon of distributed cognition has been studied in a quantitative, operational manner, although from two very different traditions: social psychology and multi-agent simulations. Social psychologists have typically studied group level cognition using laboratory experiments, and their research has documented various fundamental shortcomings of collective intelligence. We often fall prey to biases and simplistic stereotypes about groups, and many of these distortions are emergent properties of the cognitive dynamics in interacting minds. Examples are conformity and polarization which move a group as a whole towards more extreme opinions (Ebbesen & Bowers, 1974; Mackie & Cooper, 1984; Isenberg, 1986), communication within groups which reinforce stereotypes (e.g., Lyons & Kashima, 2017), and the lack of sharing of unique information so that intellectual resources of a group are underused (Larson et al., 1996, 1998; Stasser, 1999; Wittenbaum & Bowman, 2017). Although this research provides empirical evidence about real people to convincingly test particular hypotheses, it is often limited to small groups with minimal structures performing tightly controlled tasks of limited duration. In contrast, the approach of multi-agent systems originates in distributed artificial intelligence (Weiss, 1999) and the modeling of complex, adaptive systems, such as animal swarms (Bonabeau, Dorigo & Theraulaz, 1999) and human societies (Epstein & Axtell, 1996; Nowak, Szamrej & Latané, 1990). In these systems, agents interact in order to reach their individual or group objectives, which may be conflicting. However, the combination of their local, individual actions produces emergent behavior at the collective level. In contrast to experimental psychology, this approach has been used to investigate complex situations, such as the emergence of cooperation and culture, and the self-organization of the different norms and institutions that govern the interactions between individuals in an economy. It manages to tackle such complex problems by means of computer simulations in which an unlimited number of software agents interact according to whatever simple or complex protocols the researcher has programmed them to obey. Although the complexity of situations that can be studied is much greater than in an experimental paradigm, there is no guarantee that the results produced by the simulation say anything meaningful about real human interaction.
Moreover, many of the earlier simulations to represent agent interaction are too rigid and simplistic to be psychologically plausible. The behavior of a typical software agent is strictly rule-based: If a particular condition appears in the agent's environment, the agent will respond with a particular, preprogrammed action. More sophisticated cognitive architectures may allow agents to learn, that is, adapt their responses to previous experience, but the learning will typically remain within a symbolic, rule-based level. Perhaps the most crucial limitation of many models is that the individual agents lack their own psychological interpretation and representation of the environment. As Sun (2017, p. 6.) deplored, multi-agent systems need “…better understanding and better models of individual cognition”. Another shortcoming of many multi-agent models is their rigid representation of relations between agents. In the most common models, cellular automata, each agent (“automaton”) occupies a cell within a geometrical array of cells, typically in the form of a checkerboard. Agents will then interact with all the agents in their geometric neighborhood. This is clearly a very unrealistic representation of the social world, in which individuals interact differentially with other individuals depending on their previous experiences with those others. Other types of simulations, such as social network models, are more realistic with respect to human relationships, but still tend to be rigid in the sense that the agents cannot change the strength of their relationship with a particular other agent (for a full discussion, see section on Alternative Models).
Conclusion
The proposed multi-agent TRUST connectionist model combines all elements of a standard recurrent model of impression formation that incorporates processes of information uptake, integration and memorization, with additional elements reflecting communication between individuals. Although one of many possible implementations, our model reflects the theoretical notion that information processing within and between individuals can be characterized by analogous, yet unique processing principles. One of the unique features of the model is the assumption that acquired cognitive trust in the information provided by communicators is an essential social and psychological requirement of communication. This was implemented through the inclusion of trust weights, which change depending on the consistency of the incoming information with the receiving agents’ existing beliefs and past experiences. Trust weights lead to a selective filtering out of less reliable data and selective propagation of novel information, and so bias information transmission. From this implementation of cognitive trust emerged Grice’s (1975) maxims of quality and quantity in human communication. In particular, the maxim of quality was implemented by outgoing trust weights which led to an increased acceptance of stereotypical ideas when communicators share similar backgrounds, while the maxim of quantity was simulated by attenuation in the expression of familiar beliefs (as determined by receiving trust weights) which led to a gradual decreased transmission of stereotypical utterances. By spanning a diversity of findings, the present simulations demonstrate the broad applicability and integrative character of our approach.
This may lead to novel insights for well-known social-psychological phenomena and may point to potential theoretical similarities in less familiar domains. It may help us to understand why communicating information among the members of a group sometimes makes their collective cognition and judgments less reliable. One element that distinguishes the present model from earlier approaches, but that is common to most connectionist modeling, is the dynamic nature of our system. It conceives communication as a coordinated process that transforms the beliefs of the agents as they communicate. Through these belief changes it has a memory of the social history of the interacting agents. Thus, communication is at the same time a simple transmission of information about the internal state of the talking agent, as well as a coordination of existing opinions and emergence of novel beliefs on which the conversants converge and so lead to a “common ground”. By combining all those elements in a social-distributed network of individual networks, the unique contribution of our model is that it extends distributed processing, which is sometimes seen as a defining characteristic of connectionism, into the social dimension. This is how social psychology can contribute to the further development of connectionism as tool of theory construction (cf. Smith, 1996).()
网站原创范文除特殊说明外一切图文作品权归所有；未经官方授权谢绝任何用途转载或刊发于媒体。如发生侵犯作品权现象，保留一切法学追诉权。()
更多范文欢迎访问我们主页 当然有需求可以和我们 联系交流。-X()
|
