You're Not Helping: Using Imperfect Robots to Foster Critical Thinking

• In Event: Transfer for Learning Outside the Classroom: Informal Contexts to Examine STEM Learning Skills and Strategies

Abstract

A key component in the development of STEM skills is critical thinking. Of the many out there, several definitions of critical thinking require the learner to “go deeper” and not accept the information given at face value (Norris, 1985). In particular, as artificial intelligence becomes increasingly ubiquitous, critical thinking in interactions with digital agents (e.g., social robots) will become crucial. People will need to determine the accuracy of the suggestions provided by agents and determine whether they should accept those suggestions or override them with their own knowledge or intuitions, as we often do with Global Positioning Systems today.

To date, the role of social robots in learning environments has been varied—they have been used as a tutor, a peer, and as a learner for a human student to teach. Independent of the role, a meta-review of the research indicates favorable gains for learning (Belpaeme et al., 2018). In addition to content or skill area learning, interacting with social robots may present an opportunity to explore how students develop critical thinking skills, particularly with respect to digital agents.

As an initial test, this study explores how and whether social robots can support the development of critical thinking, specifically whether students will accept or question new information. After a pre-test, participants are asked to play a game with a robot teammate. In a 2x2 design, we vary the robot’s knowledge of the rules of the game (knows rules vs. does not) and we vary the robot’s knowledge of the strategy of the game (can determine if a move is “good” or not). The robot verbally offers a suggestion for a move each turn. After several games, we give the participant a test on the rules and strategy, and ask them to rate the competence of the robot at the game. We then ask participants to play a new game with a different robot. Before playing the game, participants are asked if they have any questions about the robot and its capabilities. The second robot has low knowledge of both the rules and the strategy and offers bad advice for every move. After gameplay, we will study patterns of non-compliance with the robot’s suggestions. We hope the results will provide insight into the design of capabilities for robot teammates to foster critical thinking in learners. It may be the case that an imperfect robot is better for a learner than one that always has the correct information.

Author

• Robert Semmens, Naval Postgraduate School