Action Election: Fundamentals and Applications

By Fouad Sabry

What Is Action Election

The most fundamental challenge faced by intelligent systems is deciding what actions to take next, and action selection is one method to characterize this challenge. "The action selection problem" is often connected with intelligent agents and animats in the fields of artificial intelligence and computational cognitive science. Intelligent agents and animats are artificial systems that demonstrate complicated behavior in an agent environment. In the field of ethology, which studies animal behavior, the term is also occasionally employed.

How You Will Benefit

(I) Insights, and validations about the following topics:

Chapter 1: Action selection

Chapter 2: Putamen

Chapter 3: Striatum

Chapter 4: Basal ganglia

Chapter 5: Dopaminergic pathways

Chapter 6: Nigrostriatal pathway

Chapter 7: Ventral tegmental area

Chapter 8: Frontostriatal circuit

Chapter 9: Neuromodulation

Chapter 10: Neural binding

(II) Answering the public top questions about action election.

(III) Real world examples for the usage of action election in many fields.

(IV) 17 appendices to explain, briefly, 266 emerging technologies in each industry to have 360-degree full understanding of action election' technologies.

Who This Book Is For

Professionals, undergraduate and graduate students, enthusiasts, hobbyists, and those who want to go beyond basic knowledge or information for any kind of action election.

• Language: English
• Publisher: One Billion Knowledgeable
• Release date: Jul 3, 2023

Book preview

Chapter 7: Action selection

The most fundamental challenge faced by intelligent systems is deciding what actions to do next, and action selection is one method to characterize this challenge. The action selection issue is often connected with intelligent agents and animats in the fields of artificial intelligence and computational cognitive science. Intelligent agents and animats are artificial systems that demonstrate complicated behavior in an agent environment. In ethology and the study of animal behavior, the word is also sometimes employed.

Understanding action selection may be difficult since it might be difficult to determine the degree of abstraction that is employed when selecting a act. At the most fundamental level of abstraction, an atomic act may be anything from the contraction of a single muscle cell to the instigation of a conflict. In most cases, the collection of different actions that may be chosen by an action-selection system is predetermined and unchangeable.

The majority of researchers who are currently working in this subject have high expectations for their agents:

In most cases, the acting agent is required to choose its action in contexts that are both dynamic and unexpected.

Since the agents often operate in real time, it is essential that they get to conclusions as quickly as possible.

In most cases, the agents are developed to carry out a variety of distinct responsibilities. It's possible that allocating resources to these activities may result in a conflict (for example, can the agent put out a fire and provide a cup of coffee at the same time?).

There may be others present in the setting in which the agents work, who might make things more challenging for the agent (either intentionally or by attempting to assist.)

The agents themselves are often meant to imitate animals or people, yet the behavior of animals and humans is notoriously difficult to predict.

Because of these factors, the selection of actions is not easy and generates a significant amount of interest in study.

The intricacy of the situation is the primary challenge in decision-making. Because every calculation requires both time and space (in memory), it is impossible for agents to take into account all of the possibilities that are open to them at each and every moment in time. As a consequence, they must have some kind of prejudice, and their search must be limited in some manner. When it comes to AI, the topic of action selection is: what is the most effective strategy to restrict this search? The issue that has to be answered by biologists and ethologists is how different kinds of animals limit their search. Does every kind of animal behave in the same way? Why do they choose to make use of the ones that they do?

In the context of action selection, one of the most important questions to ask is whether or not it is really a problem for an agent, or whether or not it is merely a description of an emergent attribute of the behavior of an intelligent agent. However, if we think about how we are going to construct an intelligent creature, it seems obvious that there must be some kind of process for choosing what actions to do. This mechanism might be extensively dispersed (as it would be in the case of spread animals such as social insect colonies or slime mold), or it could be a module designed for a specific function.

The action selection mechanism (ASM) not only decides how the agent's actions will have an effect on the outside world, but it also defines where the agent will focus its attention perceptually and how it will store new information. Because updating memory suggests that some form of machine learning is possible, these sorts of egocentric actions may in turn result in changing the agent's fundamental behavioral capacities. This is especially true given that updating memory implies that it is possible for the agent to learn on its own. In an ideal world, action selection itself should also be capable of learning and adapting, however there are several challenges of combinatorial complexity and computational tractability that may necessitate constraining the search space for learning.

In the field of artificial intelligence, an agent support module (ASM) is also sometimes referred to as an agent architecture or thought of as a significant component of one.

Artificial action selection processes may, in general, be broken down into many distinct groups. These categories include symbol-based systems, which are often referred to as classical planning; distributed solutions; and reactive or dynamic planning. A few of these methods cannot be placed in any of these categories with complete accuracy. Others are more concerned with the provision of scientific models than with the actual control of AI; these latter are elaborated upon more in the next section.

It was considered early on in the development of artificial intelligence that the best approach for an agent to determine what to do next would be to calculate a probable optimum plan, and then execute that plan. This was done under the assumption that this would be the most efficient method. This gave rise to the physical symbol system theory, which states that in order to have intelligence, a physical agent that is capable of manipulating symbols is both required and sufficient. This method is still used by many software agents as a means of action selection. In most cases, it is necessary to describe all sensor readings, the world, all of one's behaviors, and all of one's objectives using some kind of predicate logic. Critics of this method argue that it is unsuitable for real-time planning and that, despite the proofs, it is still unlikely to produce optimal plans due to the fact that reducing descriptions