AI and Robotics Policy Module

The AI & Robotics Policy Modules are designed to help undergraduate students reason about AI technologies as socio-technical systems. The modules ask students to consider how a system’s capabilities and limitations, human decisions, safety concerns, and broader societal impacts are connected.

Each module uses the Narrative Policy Framework (NPF) to organize students’ reasoning around four questions:

• Who is or could be affected?

• Who or what may contribute to the problem?

• Who could help prevent or address the problem?

• What solutions could improve the situation?

Students answer these questions more than once as they move through the module. New technical and computational information is introduced along the way, allowing students to reconsider and refine their initial reasoning.

We developed modules for three AI-powered technologies: humanoid robots, self-driving cars, and generative AI.

How the Modules Are Structured

Each module follows five connected components. Below, we explain the purpose of each component and use the humanoid robot module as an example.

 1. Introducing a Focus Event 

Each module begins with real-world narratives about the potential benefits and risks of the technology. These narratives introduce students to a broader issue surrounding how the technology is being developed, deployed, or used in society.

Before receiving detailed technical information, students use the NPF questions to form an initial interpretation of the issue: who may be affected, who or what may contribute to problems, who could take action, and what solutions they would propose.

Humanoid robot example. Students begin by considering the use of humanoid robots around people. They encounter a positive perspective through the AI for Good Global Summit, which describes ways humanoid robots may contribute to several United Nations Sustainable Development Goals. They also examine two incidents involving the commercially available Unitree H1 humanoid robot: a factory-test malfunction and an incident during a public festival in which the robot unexpectedly moved toward spectators.

2. A Case Study Narrative

Students next examine a specific real-world incident involving a different system within the same technology category. They analyze the incident before receiving detailed technical information about that system.

Using a different system is intentional. It allows students to develop their reasoning around one case and later consider whether what they learned can be applied to another system and context within the same technology area.

Humanoid robot example. After considering narratives involving the Unitree H1, students examine an incident involving Digit, a humanoid robot developed by Agility Robotics. During a demonstration in which Digit was transporting boxes, the robot unexpectedly collapsed.

Students watch the incident and consider who could have been affected, what may have contributed to the failure, who could help prevent similar incidents, and what actions might improve safety. 

4. Computational Activities

Students next explore a simplified computational scenario related to the case study. The purpose of these activities is not to teach programming. Instead, they allow students to see how a software decision or technical limitation can contribute to different system behaviors and safety outcomes.

Students first receive an English-language description of the algorithm and then interact with a small Python program. They can run the simulation, modify selected values or settings, and observe how those changes affect the behavior of the system.

Humanoid robot example. Students explore a simplified scenario involving battery depletion and robot shutdown. They compare two implementations:

Unsafe shutdown: The robot continues working until its battery reaches 0% and collapses where it is operating. Click on the Run button below to run the program and observe the unsafe shutdown.

Safe shutdown: When the battery reaches a predefined threshold, the robot puts down its load, moves to a designated safe area, kneels, and shuts down.

Students can modify selected parts of the code to explore how different settings affect the outcome. For example, they can change the BATTERY_DRAIN value to control how quickly the battery decreases or switch SAFE_SHUTDOWN from False to True on line 8 of the code to compare unsafe and safe shutdown behavior. These small changes make it possible to see how software and design decisions can lead to different safety outcomes.

Activity: Switch SAFE_SHUTDOWN from False to True (or vice versa) on line 8, then click Run again to observe how the robot’s behavior changes.

5. Revisiting the Focus Event

At the end of the module, students return to the real-world narratives they encountered at the beginning. After examining a different system in greater technical depth, students revisit the original NPF questions from the Introduction and reflect on whether the additional technical and computational context changes their reasoning.

This final component also allows us to examine whether students can transfer what they learned from a specific case to another system and a broader issue within the same technology category.

Humanoid robot example. After analyzing the Digit incident, reviewing Digit’s technical profile, and exploring different shutdown behaviors, students return to the original narratives about humanoid robots interacting with people, including the Unitree H1 incidents.

They again consider who may be affected, who or what may contribute to potential problems, who could help address them, and what solutions might support safer integration of humanoid robots.

In this way, students move from an initial understanding of humanoid robots, to a detailed examination of one robot and its technical behavior, and then apply that additional understanding back to the broader questions introduced at the beginning of the module.