Levels of Learning in Natural and Artificial Agents

Abstract

Throughout the history of psychology, there has been a recognition that there are multiple time-scales of decision making in human cognition. This includes early dual-process theories, but was recently popularized by Daniel Kahneman in his book, Thinking Fast and Slow (2011), where he proposed System 1 and System 2. We propose that it is useful to make a related distinction in learning for humans and intelligent autonomous agents based on how learning fits into the overall cognitive architecture. We initially define Level 1 (L1) and Level 2 (L2), where L1 are fixed, innate, automatic (architectural) learning mechanisms, and L2 are (knowledge-based) learning strategies that are controlled by the agent to create experiences such that L1 mechanisms can learn useful knowledge. The purpose of this project was to explore and refine these distinctions across both artificial and natural intelligent autonomous systems.

Document Details

• Document Type: Technical Report
• Publication Date: Jul 15, 2021
• Accession Number: AD1145883

Entities

People

• John E. Laird

Organizations

• Board of Regents of the University of Michigan

Tags

Communities of Interest

• Autonomy

DTIC Thesaurus Topics

• Air Force Research Laboratories
• Algorithms
• Animal Behavior
• Animals
• Artificial Intelligence
• Automatic
• Autonomous Agents
• Classification
• Cognition
• Computer Languages
• Deep Learning
• Educational Psychology
• Environment
• Instructions
• Instructors
• Intelligent Agents
• Literature
• Literature Surveys
• Machine Learning
• Michigan
• Neural Networks
• Psychology
• Reasoning
• Reinforcement Learning
• Scientific Research
• Taxonomy

Readers

• Agent-Based Social Robotics and Mobile-Assisted Learning in Virtual Environments.
• Neural Network Machine Learning.
• Team-Based Human-Centered Cognitive Task Decision Making and Information Performance.

Technology Areas

• Autonomy
• Autonomy - Human-Robot Interaction