diff --git a/005_team_3_agent_design/005_team_3_agent_design.tex b/005_team_3_agent_design/005_team_3_agent_design.tex index 7eadb26..8899a0a 100644 --- a/005_team_3_agent_design/005_team_3_agent_design.tex +++ b/005_team_3_agent_design/005_team_3_agent_design.tex @@ -62,7 +62,7 @@ \subsection{Agent Generation} While the implementation of initial randomness alone allowed for interesting strategies to occur, we also wanted the agents to adapt their morality on the basis of their community and associations, as argued by Shao. This was made possible by having all actions an agent does affect its emotion variables. \par \section{Agent Communication}\label{sec:agent_communication} -Communication between agents is fundamental for the possibility of collaboration and self-organisation between agents. Crowd behaviour is considered to be generated by individuals, context-dependent and dynamic, as defended by the modern foundation of crowd behaviour \cite{0d9bc1ee81234780b2bb6ecd02762d56}. In this project, communication between agents is based on two concepts, treaties and direct messages, which are equally exploited by the Team 3 agent and combined with the conceptualization of friendship levels. +Communication between agents is fundamental for the possibility of collaboration and self-organisation between agents. Crowd behaviour is considered to be generated by individuals, context-dependent and dynamic, as defended by the modern foundation of crowd behaviour \cite{0d9bc1ee81234780b2bb6ecd02762d56}. In this project, communication between agents is based on two concepts, treaties and direct messages, which are equally exploited by the Team 3 agent and combined with the conceptualisation of friendship levels. \subsection{The Agent’s Social Network} The agent forms relationships with others surrounding it as they communicate with each other. The agent remembers everyone it has met and assigns a level of friendship to them depending on their interactions. By doing this, the agent forms a social network that will affect its decisions, something that the Theory of Planned Behaviour explores in the “subjective norm” aspect of the theory. In the TPB, all three main variables are interdependent and for this reason, the agent will have a better or worse opinion of someone it has just met depending on their morality at that point. \par diff --git a/009_team_7_agent_design/009_team_7_agent_design.tex b/009_team_7_agent_design/009_team_7_agent_design.tex index 437570c..ccff821 100644 --- a/009_team_7_agent_design/009_team_7_agent_design.tex +++ b/009_team_7_agent_design/009_team_7_agent_design.tex @@ -2,11 +2,11 @@ \chapter{Team 7 Agent Design}\label{team_7_agent_design} \section{Overview} \label{sec: Team 7 Overview} -This chapter outlines the design and findings of Team 7's agent. The primary goal of team 7 was to emulate human behaviour closely and to replicate the normal variation in personality types exhibited in a typical human population. This would then enable analysis into whether particular personality types exhibited particular behavioural patterns or are more conducive to self organisation and thus acheiving the goal of sustainability in a collective action problem. +This chapter outlines the design and findings of Team 7's agent. The primary goal of Team 7 was to emulate human behaviour closely and to replicate the normal variation in personality types exhibited in a typical human population. This would then enable analysis into whether particular personality types exhibited particular behavioural patterns or are more conducive to self organisation and thus acheiving the goal of sustainability in a collective action problem. \section{Agent Design} \label{sec: Agent Design} -There are 3 key elements that define the agent: its personality, behaviour and memory. The personality defines a given agent at its core, and for the purpose of this experiment is kept constant throughout a given simulation. An agents behaviour is influenced by its personality, memory (a record of its experiences) and the environment. The environment can include current floor, food available and interaction with other agents. The personality and behaviour then combine to influence the various decisions that the agent makes over the course of the simulation. +There are three key elements that define the agent: its personality, behaviour and memory. The personality defines a given agent at its core, and for the purpose of this experiment is kept constant throughout a given simulation. An agents behaviour is influenced by its personality, memory (a record of its experiences) and the environment. The environment can include current floor, food available and interaction with other agents. The personality and behaviour then combine to influence the various decisions that the agent makes over the course of the simulation. This basic agent framework is illustrated in \Cref{agent_flow}. \begin{figure}[H] @@ -26,7 +26,7 @@ \subsection{Personality Traits} \begin{itemize} \item \textbf{Openness}: The degree to which an individual is willing and eager to have new experiences. This also has implications for creativity and problem solving. An individual with low openness will behave more conservatively and will be less welcoming to change. - \item \textbf{Conscientiousness}: The degree of conscientiousness of an individual determines their level of organization, discipline and thoughtfulness. Conscientious individuals are also more strategic and forward thinking. + \item \textbf{Conscientiousness}: The degree of conscientiousness of an individual determines their level of organisation, discipline and thoughtfulness. Conscientious individuals are also more strategic and forward thinking. \item \textbf{Extraversion}: Extraversion is the extent to which an individual is willing and eager to socialise and interact with other individuals. An extroverted individual is more likely to occupy positions of leadership and influence. @@ -39,7 +39,7 @@ \subsection{Personality Traits} \subsection{Behavioural Traits} \label{subsec: Behaviour} -In contrast to personality traits, behavioural characteristics of an agent change over the course of the simulation. The behaviours are initialized based on the agent's personality traits and can range from 0-100 over the lifetime of the agent. These behaviours are then updated daily by environmental factors such as food availability, floor changes and interactions with other agents. +In contrast to personality traits, behavioural characteristics of an agent change over the course of the simulation. The behaviours are initialised based on the agent's personality traits and can range from 0-100 over the lifetime of the agent. These behaviours are then updated daily by environmental factors such as food availability, floor changes and interactions with other agents. The behaviour characteristics are described below: \begin{itemize} @@ -63,7 +63,7 @@ \subsection{Behavioural Traits} \end{itemize} \item \textbf{Responsiveness}: - This behaviour influences the likelihood of an agent accepting the treaties of other agents. It is initialized as the average of openness, extraversion, and agreeableness. + This behaviour influences the likelihood of an agent accepting the treaties of other agents. It is initialised as the average of openness, extraversion, and agreeableness. In the current implementation, the responsiveness trait does not vary during the simulation. In an implementation where dishonesty is factored in, the agent would keep record of the trustworthiness of each agent that it interacts with. If other agents are largely untrustworthy, overall responsiveness would go down proportionally. @@ -253,7 +253,7 @@ \subsubsection{Significance of Treaties} \label{fig: Cumulative Deaths with Treaties} \end{figure} -This demonstrates that the agents perform better collectively when treaties are activated. This is expected as treaties enable agents to cooperate and organize the distribution of scarce food. +This demonstrates that the agents perform better collectively when treaties are activated. This is expected as treaties enable agents to cooperate and organise the distribution of scarce food. \newpage \subsubsection{Impact of Personalities} @@ -401,7 +401,7 @@ \subsubsection{Kindness} A high kindness score sees double the deaths of a low kindness score. A high kindness score does not achieve stability however a low kindness score allows for the tower to reach it in less than 200 days. The most likely explanation for this is that agents with high kindness are potentially too selfless. It is likely that they consume in small quantities when food is available and thus do not have sufficient health when food isn't available. Thus, making them more vulnerable. \subsubsection{Inter-team Comparison} -\Cref{tab: Inter-team data} shows the total number of deaths for simulations of various combinations of agent teams. All simulations are run with a simulation period of 500 days, shuffle period of 7 days and 10 team 7 agents with 10 agents of one other team type. The table is further broken down into food available per agent (5, 10 and 15). +\Cref{tab: Inter-team data} shows the total number of deaths for simulations of various combinations of agent teams. All simulations are run with a simulation period of 500 days, shuffle period of 7 days and 10 Team 7 agents with 10 agents of one other team type. The table is further broken down into food available per agent (5, 10 and 15). \begin{table} \begin{center} @@ -441,7 +441,7 @@ \subsubsection{Inter-team Comparison} \hline \end{tabular} \end{center} - \caption{Comparison between agent teams and team 7.} + \caption{Comparison between agent teams and Team 7.} \label{tab: Inter-team data} \end{table} @@ -453,7 +453,7 @@ \subsubsection{Inter-team Comparison} \label{fig: agent deaths} \end{figure} -All team agents, bar team 6, were able to coexist with team 7 and achieve zero deaths for a food availability of 15 per agent, indicating that Team 6 agent was the most incompatible with our agent. Team 4 achieved zero deaths for food availability of 10 demonstrating a greater compatibility with team 7 and team 5's results are not too dissimilar. +All team agents, bar Team 6, were able to coexist with Team 7 and achieve zero deaths for a food availability of 15 per agent, indicating that Team 6 agent was the most incompatible with our agent. Team 4 achieved zero deaths for food availability of 10 demonstrating a greater compatibility with Team 7 and Team 5's results are not too dissimilar. \subsubsection{Conclusion} -The agent designed by team 7 is unique in terms of the use of personalities to determine the general behaviour of the agent. These personality traits assigned allow it to mimic human behaviour as much as possible. The results presented above show that team 7's agents can have vastly different behaviours with each other agent, showing our compatibility with some agents and incompatibility with others. In short, the team 7 agents have been intricately designed to meaningfully engage in a huge variety of scenarios and attempt to accomplish its to major goals, to survive individually and collectively. \ No newline at end of file +The agent designed by Team 7 is unique in terms of the use of personalities to determine the general behaviour of the agent. These personality traits assigned allow it to mimic human behaviour as much as possible. The results presented above show that Team 7's agents can have vastly different behaviours with each other agent, showing our compatibility with some agents and incompatibility with others. In short, the Team 7 agents have been intricately designed to meaningfully engage in a huge variety of scenarios and attempt to accomplish its to major goals, to survive individually and collectively. \ No newline at end of file