Teaching AI Thinking: How a Simple Search Algorithm Empowers Young Learners

Author: Denis Avetisyan

New research shows middle school students can grasp core artificial intelligence concepts by learning a fundamental problem-solving technique within the context of science education.

The I-SAILBFS activity sequence demonstrates an interface designed to propagate failure, its views revealing a system not engineered for stability, but grown to anticipate and accommodate its inevitable disintegration.

This paper details a curriculum using Breadth-First Search to foster algorithmic reasoning and AI literacy in students with no prior computing experience.

Despite growing ubiquity, foundational artificial intelligence concepts remain largely inaccessible to young learners. This paper details the design and implementation of ‘Democratizing Foundations of Problem-Solving with AI: A Breadth-First Search Curriculum for Middle School Students’, a curriculum embedding AI literacy within existing science instruction through the accessible algorithm of Breadth-First Search. Results demonstrate that middle school students can successfully learn and apply this AI problem-solving technique-even without prior computing experience-when integrated into the context of real-world science scenarios. How might such curriculum designs broaden access to AI education and cultivate algorithmic reasoning skills across diverse student populations?

The Inevitable Framework: Foundations for AI Literacy

The rapidly evolving field of artificial intelligence presents unique challenges to education. Simply teaching specific tools or techniques proves insufficient, as these quickly become obsolete. Instead, a robust foundational framework is essential to equip learners with the capacity to critically assess and adapt to ongoing advancements. This framework moves beyond rote memorization, fostering a deeper understanding of the core principles that underpin all AI systems. Without such a foundation, individuals risk becoming passive consumers of AI technology rather than informed participants in its development and deployment, hindering both innovation and responsible application.

The AI4K12 initiative addresses the need for a structured approach to artificial intelligence education by proposing five fundamental concepts as an organizing principle for learning. Recognizing the rapidly evolving nature of AI, the initiative champions these ‘Big Ideas’ – Perception, Representation & Reasoning, Learning, Interaction, and Societal Impact – not as a rigid curriculum, but as a flexible framework. This framework allows educators to connect diverse AI topics and applications to core concepts, fostering a deeper, more transferable understanding. By focusing on these foundational ideas, the AI4K12 initiative aims to empower students to not only understand how AI systems work, but also to critically evaluate their capabilities, limitations, and broader implications for society, preparing them to be informed and engaged citizens in an increasingly AI-driven world.

The core of AI literacy rests on grasping five fundamental concepts: how machines perceive the world, how knowledge is represented and used for reasoning, the mechanisms by which AI systems learn from data, how these systems interact with humans and other entities, and crucially, the broader societal impact of these technologies. These aren’t isolated topics, but rather interconnected lenses through which any AI application can be analyzed – from self-driving cars to medical diagnosis tools. By understanding these ‘Big Ideas’, individuals can move beyond simply using AI to critically evaluating its capabilities, recognizing its limitations, and anticipating its consequences, fostering a more informed and responsible engagement with artificial intelligence.

Tracing the Logic: Breadth-First Search in Action

Breadth-First Search (BFS) is a graph traversal algorithm that systematically explores a graph level by level. Beginning at a designated root node, BFS first examines all immediate neighbor nodes before moving to the next level of neighbors. This contrasts with Depth-First Search, which explores as far as possible along each branch before backtracking. The algorithm utilizes a queue data structure to manage the order of node visits, ensuring that nodes closer to the root are processed before those further away. This level-by-level approach makes BFS particularly well-suited for finding the shortest path between two nodes in an unweighted graph, and its relatively straightforward implementation provides a foundational understanding of more complex graph algorithms.

Breadth-First Search (BFS) gains practical relevance when applied to scenarios mirroring real-world problems; modeling virus transmission, for example, allows BFS to systematically explore all immediately infected individuals before moving to secondary contacts, mirroring how an epidemic spreads. Similarly, pathfinding problems – determining the shortest route between two points – utilize BFS by exploring all neighboring nodes at a given distance before progressing to those further away. This systematic, level-by-level exploration contrasts with depth-first approaches and highlights BFS’s capacity to find the shortest path or most immediate spread in these modeled situations, making the algorithm’s logic more intuitive and demonstrably powerful.

Breadth-First Search (BFS) directly underpins several practical applications, notably contact tracing in epidemiology and route finding in GPS navigation systems. In contact tracing, individuals exposed to an infected person are identified and categorized by their degree of separation – representing levels in the graph – allowing for prioritized intervention based on proximity. Similarly, GPS navigation utilizes BFS to determine the shortest path between two locations, treating roads as edges and intersections as nodes; the algorithm explores all possible routes at a given distance before moving to routes further away, ensuring the most efficient path is identified. These implementations demonstrate BFS’s utility in solving problems involving finding the shortest path or exploring all reachable nodes within a networked system.

A recent study assessed the efficacy of a co-designed curriculum on student comprehension of the Breadth-First Search (BFS) algorithm. Results indicated a statistically significant improvement in understanding, as measured by a pre-to-post assessment with a score increase of 0.107. This improvement was determined to be statistically significant with a p-value of less than 0.01. The effect size, calculated using Cohen’s d, was 0.453, indicating a moderate effect of the curriculum on student learning outcomes.

Cultivating Understanding: I-SAIL and Interactive AI Exploration

The I-SAIL environment leverages the Snap! visual programming language to offer an accessible and engaging platform for artificial intelligence education. Snap!, a descendant of Scratch, utilizes a block-based interface allowing users to construct programs by connecting graphical blocks, minimizing syntax errors common in text-based coding. This approach lowers the barrier to entry for students with limited or no prior programming experience, enabling them to focus on the logical concepts underlying AI algorithms rather than debugging code. I-SAIL is designed to facilitate exploration of AI concepts through interactive activities and simulations, providing immediate feedback and promoting a deeper understanding of computational thinking.

The I-SAIL environment utilizes Snap!, a visual, block-based programming language, to enable students to implement and experiment with algorithms such as Breadth-First Search (BFS). This interface allows students to construct algorithmic logic by connecting graphical blocks, rather than writing traditional code, thereby lowering the barrier to entry for exploring computational concepts. Students can directly manipulate the BFS algorithm-defining the starting node, specifying neighbor relationships, and visualizing the search process-within the I-SAIL environment. This hands-on approach facilitates a practical understanding of how BFS operates and its application to pathfinding problems without requiring prior coding experience.

I-SAIL incorporates non-digital, tactile exercises designed to solidify comprehension of AI algorithms alongside digital implementation. These “unplugged” activities involve students manually simulating algorithmic processes, such as pathfinding, using physical materials and step-by-step instructions. By removing the complexities of coding and syntax, these exercises allow students to focus on the logical sequence of operations and the underlying principles of the algorithm. This approach facilitates a more intuitive grasp of concepts before or in conjunction with digital coding, promoting deeper learning and retention of algorithmic thinking skills.

Assessment of student understanding following engagement with the I-SAIL environment demonstrated that 64.5% achieved correct identification of the shortest path within a Breadth-First Search (BFS) tracing exercise administered on paper. This worksheet-based evaluation served as a direct measure of algorithmic comprehension, requiring students to manually trace a BFS algorithm and select the optimal route. The result indicates a statistically significant transfer of knowledge from the visual, block-based I-SAIL implementation to a non-digital, analytical task, suggesting effective internalization of the BFS concept and its application to pathfinding problems.

The integration of I-SAIL’s digital environment with unplugged activities is designed to enhance cognitive processing of algorithmic concepts. Data indicates that students participating in both digital implementation and tactile reinforcement demonstrate improved comprehension of problem-solving methodologies; specifically, a post-activity assessment revealed 64.5% accuracy in tracing shortest paths using Breadth-First Search (BFS) on paper. This suggests that the combined approach facilitates a more robust understanding of algorithmic thinking by allowing students to translate abstract computational logic into concrete, manually verifiable steps, thereby solidifying their ability to apply these skills independently of the digital interface.

Expanding the Horizon: A Diverse Toolkit for AI Literacy

The burgeoning field of artificial intelligence education benefits from a growing suite of accessible tools designed to cater to varied learning preferences. Systems like PopBots, with their tactile robotic interfaces, allow for kinesthetic exploration of AI principles, while platforms such as Teachable Machine empower students to build and train custom machine learning models through intuitive visual programming. Calypso offers a block-based coding environment specifically geared towards AI and robotics, fostering computational thinking, and Any-Cubes provide a tangible, puzzle-like approach to understanding algorithmic concepts. This diversity extends beyond simple interface differences; each tool emphasizes distinct pedagogical approaches, allowing educators to select resources that best align with individual student strengths and learning styles, ultimately promoting deeper engagement and comprehension of complex AI topics.

A central tenet of effective AI education lies in moving beyond theoretical understanding to practical application, and a growing suite of tools facilitates precisely this shift. Systems like PopBots, Teachable Machine, and others empower learners to directly manipulate AI components, build simple models, and observe the resulting behaviors. This hands-on engagement bypasses passive reception of information, fostering a deeper, more intuitive grasp of core concepts. Instead of simply hearing about machine learning, students actively train algorithms, experiment with datasets, and witness the consequences of their design choices – cultivating not only knowledge, but also critical thinking and problem-solving skills essential for navigating an increasingly AI-driven world.

Effective AI education increasingly recognizes that a singular approach fails to resonate with all learners. Educators are now equipped with a growing suite of tools – from visually-oriented platforms like Teachable Machine to physically interactive systems such as Any-Cubes – allowing for differentiated instruction. This multiplicity of entry points acknowledges diverse learning styles; a student captivated by robotics might begin with a tangible project, while another, drawn to visual patterns, could explore AI through image recognition. By embracing this range of methods, educators can bypass traditional barriers to understanding, fostering engagement and ensuring that students with varied abilities and interests can successfully access and contribute to the field of artificial intelligence. Ultimately, this inclusive strategy broadens participation and cultivates a more robust and innovative AI-literate future.

A truly effective artificial intelligence education necessitates accessibility for all learners, and a diverse toolkit is proving central to this goal. The proliferation of resources – from visually-oriented platforms like Teachable Machine to physically interactive systems such as Any-Cubes – dismantles traditional barriers to entry, accommodating a wider spectrum of learning styles and abilities. This isn’t simply about offering more options; it’s about recognizing that AI concepts can be grasped through various modalities, allowing students to build understanding in ways that resonate with their individual strengths. Consequently, a more inclusive landscape emerges, one where AI education isn’t confined to those with specific technical backgrounds, but is instead open to anyone with curiosity and a desire to learn, ultimately broadening participation and fostering innovation within the field.

The pursuit of algorithmic literacy, as demonstrated by this curriculum’s success with Breadth-First Search, reveals a fundamental truth about systems. It isn’t about imposing a solution, but fostering an environment where understanding emerges. The study highlights that even without prior computing experience, students can grasp complex concepts when presented within a meaningful scientific context. This echoes a sentiment articulated by Alan Turing: “Sometimes people who are unhappy tend to look for a person to blame.” The ‘person to blame’ in education is often the perceived difficulty of a subject, but this work suggests that with the right ecosystem – a curriculum that prioritizes growth over rigid structure – even the most daunting concepts can flourish. Scalability, in this case, isn’t about reaching more students, but about ensuring the flexibility of the learning process itself.

The Seed and the Garden

This demonstration of algorithmic reasoning in young learners isn’t a victory for curriculum design – it’s a postponement of inevitable complication. The ease with which middle school students grasp Breadth-First Search doesn’t speak to the simplicity of the algorithm, but rather to the temporary absence of real-world constraints. Each successfully navigated problem space is a small, contained garden; the wilderness beyond will demand adaptations these students haven’t yet conceived. The curriculum, in its current form, prepares them for a problem, not for the endless proliferation of problems.

The true measure of success won’t be proficiency in search algorithms, but resilience in the face of algorithmic failure. It’s not about building a tool, but cultivating an ecosystem where students recognize the inherent limitations of any system. The next phase of inquiry must address not just what can be searched, but what resists search – the messy, analog realities that defy formalization.

This work plants a seed. The garden will require constant tending, and ultimately, will succumb to entropy. The challenge isn’t to prevent this decay – it’s to prepare the students to replant, to adapt, and to recognize that every elegant solution is merely a temporary truce with chaos.

Original article: https://arxiv.org/pdf/2604.01396.pdf

Contact the author: https://www.linkedin.com/in/avetisyan/

The Inevitable Framework: Foundations for AI Literacy

Tracing the Logic: Breadth-First Search in Action

Cultivating Understanding: I-SAIL and Interactive AI Exploration

Expanding the Horizon: A Diverse Toolkit for AI Literacy

The Seed and the Garden

See also: