From interaction to information Sample Clauses

From interaction to information. Once ▇▇,t has been chosen the active agent engages in an information gathering interaction (IGI) with each agent in Ii,t. An IGI is an abstraction for a process between two agents that results in an increase in information for at least one of the agents. An IGI is used to represent many things, such as message passing in distributed systems, or disease propagation in contagion models. The important facet of an interaction is that the agents have some information about each others state after the interaction. Consider the distributed commit problem from [▇▇▇▇▇, 1997]. Each agent sends messages to other agents in the system. Each message can be considered an IGI between the two agents, as it increases the information of one agent. In particular, one of the agents now has single agent state information about the other senders state. In the linguistic agreement case a language game is a type of IGI. As agents interact they receive information about the language of the other agent. IGIs can be organized around two properties, their accuracy and their effort. An IGI can result in inaccurate information. For instance, following the distributed commit problem example, if the sender agent exhibited Byzantine behavior – that is, unpredictable behavior – then the message that 3So technically, in this case time (as in number of instances of the GAP process) to convergence scales with N 2. Our work views time as the number of instances of this process before convergence. is sent might not actually represent the state of the sender. The following are some types of inaccuracies that can occur with IGIs. This type of issue occurs when each state is a complex entity, such as a language. For instance, consider the Naming game setting described in Section 2.7 where the state of each agent is a mapping between a set of words and meanings. On each interaction the speaker/hearer agents only learn about the meanings of a subset of each others words – each agent only has a partial view of the other agents language. As a model of human linguistic interaction this restriction on communication makes sense – humans do not, and fundamentally cannot, communicate their entire language to each other. The case of partial information is very important and is described in more detail in Chapter 5. All of these situations can modify the information agents receive. An IGI represents some type of interaction and thus it requires resources to execute. We use the term effort as a dimensionless q...