Analyzing the Goal-Finding Process of Human Learning With the Reflection Subtask

Analyzing the Goal-Finding Process of Human Learning With the Reflection Subtask

Tomohiro Yamaguchi (National Institute of Technology, Nara College, Japan), Yuki Tamai (National Institute of Technology, Nara College, Japan) and Keiki Takadama (The University of Electro-Communications, Japan)
Copyright: © 2018 |Pages: 18
DOI: 10.4018/978-1-5225-2993-4.ch019

Abstract

This chapter reports the authors' experimental results on analyzing the human goal-finding process in continuous learning. The objective of this research is to make clear the mechanism of continuous learning. To fill in the missing piece of reinforcement learning framework for the learning robot, the authors focus on two human mental learning processes, awareness as pre-learning process and reflection as post-learning process. To observe mental learning processes of a human, the authors propose a new method for visualizing them by the reflection subtask for human to be aware of the goal-finding process in continuous learning with invisible mazes. The two-layered task is introduced. The first layer is the main task of continuous learning designing the environmental mastery task to accomplish the goal for any environment. The second layer is the reflection subtask to make clear the goal-finding process in continuous learning. The reflection cost is evaluated to analyze it.
Chapter Preview
Top

Introduction

Recent years in Robotics, there are several researches on biologically inspired approaches (Habib, 2011 & Shi et al., 2015). One of the possible challenges is to make robots more intelligent assistants for a human. The mission of this research is to realize the human-mimetic intelligent system which finds out learning goals from the learning environment. Shi (2015) refers that “Biologically inspired robotics has enabled today’s robots to operate in a variety of unstructured and dynamically changing environments” and “learning often plays an important role in achieving adaption”. To design the learning robots having human-mimetic learning skill, there are several targets on human’s intelligent and mental learning processes such as individual learning or creative thinking (Habib, 2008).

This chapter will contribute the basis of human skill-inspired approach toward designing the human-mimetic intelligent system. Research issues of this chapter are as follows:

  • 1.

    What is the mechanism of the human’s goal finding process in continuous learning?

  • 2.

    How to design the environmental mastery task to accomplish the goal?

  • 3.

    How to design the goal finding process in continuous learning?

Our approaches for them are as follows:

  • 1.

    Designing the reflection subtask for human to be aware of the goal finding process in continuous learning;

  • 2.

    To design the environmental mastery task, we focus on the complexity of the learning environment which has an affect on the performance of the learning task;

  • 3.

    We discuss this issue through the analysis of the experiment of human subjects.

Researches on learning process are divided into two fields. As one of them Russell and Norvig (2009) discuss a learning algorithm for robot in robotics and Artificial Intelligence, as the other Marton and Booth (1997) discuss learning of a human in psychology. For the learning robot, reinforcement learning is the major framework since it automatically learns after a learning goal is set in the learning environment. The main feature of reinforcement learning is that the learning goal is given by the human designer. On the other hand, researches on human learning ability have been performed in various research fields such as psychology, education, business, and so on. One of the main features of human learning ability is that it covers a vast territory of learning ability including discovery of learning goals, awareness, reflection, self-regulated learning (Schunk, & Zimmerman, 2007), or continuous learning.

The objective of this research is to bring the learning ability of the learning agent close to that of a human. The authors focus on both reinforcement learning framework for the learning agent and continuous learning model of a human. However, there are two kinds of questions. First question is how to bridge an enormous gap between them. Second question is how to observe mental learning processes of a human. Previous methods of human learning researches mostly depend on observable behaviors or activities. On the other hand, a learning process of a human has a major difficulty in observing since it is a mental process. Then a human learning process is yet-to-be-defined. So it is necessary to add a new twist to observe the human learning process.

To solve these problems, the authors propose a new method for visualizing mental learning processes with invisible mazes. Yamaguchi, Takemori, and Takadama (2013) focus on continuous learning and aim for modeling the unified continuous learning process model based on reinforcement learning framework for both a human and the learning agent. Our new approaches are following:

Key Terms in this Chapter

Reflection Cost: The time a learner worked through the reflection subtask per the number of found solutions. It suggests that the continuous learner can perform the reflection subtask in a certain amount of time and can perform a very good job of it.

Goal-Finding Process in Continuous Learning: A process of meta-learning to discover learning goals while learning the main learning task.

Environmental Mastery Task: To accomplish the goal for any environment as one of the continuous learning tasks.

Reflection: Those intellectual and affective activities in which individuals engage to explore their experiences in order to lead to new understandings and appreciations.

Reinforcement Learning: A learning algorithm for a robot or a software agent to take actions in an environment so as to maximize the sum of rewards through trial and error.

Reflection Subtask: To describe the awareness from a sign of the invisible walls to visualize his/her reflection process. It is defined to attach a marker on a state in the reflection map for expressing and summarizing a learner’s mental learning processes.

Continuous Learning: Continuous learning at the individual level is regularly changing behavior based on a deepening and broadening of one’s skills, knowledge, and worldview.

Maze-Sweeping Task: A maze task to find paths from a fixed-start state S to the invisible goal state G which visit all states only once in the maze.

Learning Process: A process that consists of several mental processes. It results in changed behavior.

Awareness: Increasingly related to finding appropriate learning objects, peers and experts, or the “right” learning path.

Invisible Maze: A learning environment with an invisible goal state and invisible walls as the boundary of maze or inner obstacles.

Complete Chapter List

Search this Book:
Reset