A robot / an artifact which make us feel its intentions

1. 在利益冲突和合作需要并存的情况下，人类如何读懂他人的想法。我们发明了一种模拟这种情况的两人游戏（非零和游戏），并要求受试者玩这个游戏。在这里，玩家被要求递归地读懂别人的想法，也就是读懂对手如何读懂自己的想法，并根据对手的反应来控制自己的行动。主体的行为被分析，从而形成一个心智模型。本文还开发了这种游戏的学习算法，并将其与人类的行为进行了比较。研究了注视和面部表情在人机交流中的作用，并探讨了将这些信息应用于人机交流的方法。我们开发了一种可以检测人类目光并控制自己目光的机器人，并评估了目光控制在诱导人类对机器人采取有意姿态，最终读懂机器人思想方面的效果。在人与神器的互动中，人类可能将神器视为具有心灵的东西。根据丹尼特的三种立场，即，物理的，设计的和有意的，我们调查了人类在各种情况下对人工制品采取的立场。首先，我们评估了确定人类采取何种立场的方法，然后调查了人类（受试者）面对未知工件并不得不猜测其功能时的行为。一些提示诱导人采取有意的立场。

1. How humans read other's mind under the situation where the conflict of interest and the need for cooperation both exist. We invented two person game (a non-zero-sum game) simulating such a situation, and asked subjects to play the game. Here the players are made to read other's mind recursively, i.e., to read how the opponent read his own mind, and to control his action taking into account the opponent's reaction. The behavior of the subject is analyzed to make a model of mind. The learning algorithm for such a game is also developed and compared to the behavior of humans.2. The role of gaze and facial expression in communication is investigated, and the method to employ this information in man-machine communication is pursued. We developed a robot which can detect human gaze and can control its own gaze, and evaluated the effect of gaze control in inducing a human to take an intentional stance to the robot, and ultimately to read robot's mind.3. In human-artifact interaction, humans may regard artifact as having mind. Following the Dennett's three stances, i.e., physical, design, and intentional, we investigated what stance a human takes toward artifact under various conditions. First, we evaluated methods to determine what stance a human takes, and then investigated the behavior of a human (subject) who is confronted with an unknown artifact and have to guess its functions. Some hints to induce a human to take an intentional stance is obtained.