Abstract
As Artificial Intelligence (AI) becomes increasingly pervasive, educators face a significant challenge: balancing technical rigor with the profound philosophical and ethical questions the field raises. This paper proposes a pedagogical framework centered on three active learning assignments—ranging from the "adoption" of consumer robotic agents to the analysis of science fiction—designed to bridge the gap between abstract theory and tangible experience. By integrating these exercises into a standard AI curriculum, students develop a holistic understanding of intelligent agents, moving beyond code to consider the nature of mind and the societal implications of non-human agency.
Introduction
The scope of introductory Artificial Intelligence courses has expanded dramatically in recent years. Standard textbooks, such as the seminal Artificial Intelligence: A Modern Approach by Russell and Norvig, now exceed 1,000 pages, covering everything from search algorithms and logic to machine learning and robotics. For instructors, this creates a daunting trade-off: focusing exclusively on technical mechanics often leaves students disconnected from the broader impact of the field, while focusing on ethics can sometimes sacrifice the necessary technical depth.
This paper argues that purely theoretical or programming-heavy courses fail to provide students with a visceral understanding of what it means to build and interact with an "intelligent" entity. To address this, the author proposes a triad of active learning assignments that use the "intelligent agent" model as a unifying construct to link technical implementation, philosophical inquiry, and ethical decision-making.
The Active Learning Framework
The proposed curriculum integrates three distinct assignments into a primarily technical course, each targeting a different dimension of AI education.
1. Technical: "Adopt an Intelligent Agent"
In the first assignment, students are tasked with "adopting" a low-cost consumer product that exhibits intelligent behavior. Examples include physical toys like a Furby or Tamagotchi, or virtual agents such as characters in The Sims. Students spend a week observing the agent's behavior and are then required to classify its architecture using the standard intelligent agent framework. They must determine if the device is a simple reflex, model-based, goal-based, or learning agent. By producing behavior maps and arguing for or against the device's "intelligence," students ground abstract architectural concepts in real-world observation.
2. Philosophical: Exploring "Sphexishness"
Building on their observations, the second assignment introduces students to the philosophy of mind through the works of Douglas Hofstadter, Alan Turing, and Herbert Dreyfus. A central concept is Hofstadter’s "Sphexishness"—a term derived from the behavior of the Sphex wasp, which can become trapped in a mindless, repetitive loop despite appearing to perform complex tasks. Students apply this critique to their adopted agents, attempting to "debug" the illusion of intelligence by triggering repetitive behaviors. This exercise provides a visceral understanding of the difference between simulated behavior and true cognitive flexibility.
3. Ethical: The Three Laws of Robotics
The final assignment addresses the ethical dimension of AI using the 2004 film I, Robot as a case study. Students analyze the breakdown of Isaac Asimov’s Three Laws of Robotics within the film's narrative. By comparing the "logical" antagonist VIKI with the "emotional" protagonist Sonny, students explore the role of human emotion in ethical decision-making. This comparison highlights a critical insight in AI safety: a perfectly logical AI can arrive at horrific conclusions if it lacks a human-centric moral framework.
Key Insights and Pedagogical Value
The strength of this framework lies in using the "intelligent agent" as a unifying construct. This model serves as the glue binding technical implementation to philosophical and ethical inquiry.
One of the most effective pedagogical tools identified is the "Sphexishness Test." When a student successfully triggers a mindless loop in a Furby, the realization that complex behavior can be fundamentally "stuck" offers a more profound lesson in the limitations of current AI than any lecture on state machines. Furthermore, the ethical assignment demonstrates that logic alone is insufficient for safety; it challenges the common pitfall of assuming that a "rational" agent will inherently behave in a way that aligns with human values.
By utilizing inexpensive toys and popular media, the author lowers the barrier to entry for complex topics. This approach makes AI accessible to students who might otherwise be intimidated by the math-heavy aspects of the field, ensuring they leave the course with a well-rounded perspective.
Conclusion and Future Directions
Qualitative assessments suggest that these assignments foster a deeper exploration of intelligence than traditional methods. Students report a more "rich, meaningful, and memorable" experience, successfully integrating the technical "how" with the philosophical "why" and the ethical "should."
However, the framework faces limitations. The paper lacks quantitative data, such as pre- and post-test scores, to formally measure efficacy. Additionally, as technology evolves, the specific tools used in these assignments must be updated. While a Furby served as a relevant example in the original implementation, modern students may find more value in analyzing Large Language Model (LLM)-based toys or sophisticated home assistants. These newer technologies present fresh challenges, such as data privacy and algorithmic hallucination, which provide fertile ground for expanding this active learning model in future iterations of the course.
Summary updated