The Art of Creating Believable Agents
This paper describes how 'agents' are developed to be intelligent to our society. Anything in the new world I lan to build that might have 'behavior' will be able to be informed by the research in this paper.
THE ART OF DESIGNING SOCIALLY INTELLIGENT
AGENTS: SCIENCE, FICTION, AND THE HUMAN IN
THE LOOP
Kerstin Dautenhahn
An agent is an object that has a goal. It is able to complete a task as commanded by another person.
Autonomous agents are ones that are robotic, computational or biological.
Believability
Humans give bias to life, we see any object that exerts kinetic forces to produce movement we tend to think it is intentional. We are most specifically bias towards biological agents as we are them ourselves.
Artificial intelligence agents on computers once were not believable agents but have since become better and better. Although they give the illusion of life and we want them to be real but they are not in reality. Robots are similar but they can be physically made so allow the gap to be bridged even further. Robotics is were AI can converge with engineering to create a believable agent.
Believability of agents isn't necessarily in the intelligence of the behavior, but it is more in the subjectivity of the observer. Some people will have the personality, psychology or empathetic traits to connect with certain agents while other will not. Many animated movies have very believable characters but in their physical forms they are often at all feasible or even slightly in our reality. Agents that don't resemble human or any other biological creature can be very believable to us if identifiable behavior is shown.
Believable agents need to cater to humans cognitive and social needs.
The author compares certain virtual pets, the most popular of which is Tamagotchi by Bandai. They have all the parameters of real organisms except for the appearance, yet they are still believable.
Biological Believability
Although there are many mythical creatures that look real e.g dragons, physically they would not work on earth. But the Snouter was much more convincing for biologists and regular people alike despite not being rendered in exquisite details. This is due to the fact that scientifically it is feasible, there internal structure and thought behind the evolution process shows how a well thought out but slightly ridiculous looking creature can seem much more real. For the snouters the environment is very important because we as humans understand animals have adapted to the environments so a believable character will be a logical fit for it's environment.
Summary points
-Embodied agents. Agents are influenced by their environments as to how they are constructed.
-Autonomous agents. Interacting with our world, biological agents are influenced by history so context is very important. Also, aliveness, embodiment and autonomy are always together when considering biological agents.
-Artificial agents that are believable are perceived by humans as characters and can present them selves to us with behavior rather than being a piece of software only.
-Socially intelligent agents can interact with human social nuances.
"We and the societies we are living in are changing as are our conceptions of being "social". Thus the process and products of SIA design might be seen as a mirror that reflects our own sociality"
THE ART OF DESIGNING SOCIALLY INTELLIGENT
AGENTS: SCIENCE, FICTION, AND THE HUMAN IN
THE LOOP
Kerstin Dautenhahn
An agent is an object that has a goal. It is able to complete a task as commanded by another person.
Autonomous agents are ones that are robotic, computational or biological.
Believability
Humans give bias to life, we see any object that exerts kinetic forces to produce movement we tend to think it is intentional. We are most specifically bias towards biological agents as we are them ourselves.
Artificial intelligence agents on computers once were not believable agents but have since become better and better. Although they give the illusion of life and we want them to be real but they are not in reality. Robots are similar but they can be physically made so allow the gap to be bridged even further. Robotics is were AI can converge with engineering to create a believable agent.
Believability of agents isn't necessarily in the intelligence of the behavior, but it is more in the subjectivity of the observer. Some people will have the personality, psychology or empathetic traits to connect with certain agents while other will not. Many animated movies have very believable characters but in their physical forms they are often at all feasible or even slightly in our reality. Agents that don't resemble human or any other biological creature can be very believable to us if identifiable behavior is shown.
Believable agents need to cater to humans cognitive and social needs.
The author compares certain virtual pets, the most popular of which is Tamagotchi by Bandai. They have all the parameters of real organisms except for the appearance, yet they are still believable.
Biological Believability
Although there are many mythical creatures that look real e.g dragons, physically they would not work on earth. But the Snouter was much more convincing for biologists and regular people alike despite not being rendered in exquisite details. This is due to the fact that scientifically it is feasible, there internal structure and thought behind the evolution process shows how a well thought out but slightly ridiculous looking creature can seem much more real. For the snouters the environment is very important because we as humans understand animals have adapted to the environments so a believable character will be a logical fit for it's environment.
Summary points
-Embodied agents. Agents are influenced by their environments as to how they are constructed.
-Autonomous agents. Interacting with our world, biological agents are influenced by history so context is very important. Also, aliveness, embodiment and autonomy are always together when considering biological agents.
-Artificial agents that are believable are perceived by humans as characters and can present them selves to us with behavior rather than being a piece of software only.
-Socially intelligent agents can interact with human social nuances.
"We and the societies we are living in are changing as are our conceptions of being "social". Thus the process and products of SIA design might be seen as a mirror that reflects our own sociality"
Comments
Post a Comment