Learning by Dismantling

How Reverse Engineering Could Transform Liberal Education

As a child of engineers and a parent of an engineer, I’ve spent a lifetime stepping on small parts – rivets, plates, screws, circuit boards, wires – and putting them in a box for whoever it was who took something apart to find it (eventually).

In an era where AI can write essays, solve equations, and generate code, education faces an existential challenge. How should universities respond? The answer may lie in one of engineering's oldest practices: reverse engineering. While traditionally used to understand how physical objects are built, to understand their components and principles, instilling a reverse engineering mindset may be the answer for higher ed in the AI age. We need to teach students to systematically deconstruct AI-generated knowledge to understand sources, assumptions, and limitations. This approach transforms students from passive consumers of information into active analysts of how knowledge is constructed and transmitted.

Reverse engineering (RE) is the systematic process of deconstructing something to understand not just how it works, but why it works that way. It's about asking three fundamental questions (variations on questions I’ve asked here):

• What are the components?

• How do they interact?

• Why was it designed this way?

The true power of RE lies in its dual nature as both an analytical tool and a learning methodology. When you RE something, you're not just taking it apart, you're rebuilding your understanding from the ground up. This rebuilding process often reveals insights that aren't apparent when simply studying the finished product.

This definition and approach transcends engineering. Chefs deconstruct a complex dish to understand its ingredients and techniques. Business analysts dissect a successful marketing campaign to understand its key components. The goal is the same: to understand the underlying principles well enough to either recreate or improve upon the original.

Every child has the desire to take things apart, to tear things down. In my households, this has always been welcomed and encouraged. In my poetry classes it has always been welcomed and encouraged. How does this sonnet work? How was it engineered? Let’s take it apart and see.

In the AI era, higher education must foster a mindset of "learning by dismantling," encouraging innovating by understanding what has already been done, to do it better. If AI knows as much as it does (and can reason better every day), higher ed needs to engineer its place in the new AI reality.

Paradoxically, RE may be the best way to approach AI if you aren’t an AI engineer. The vast majority of users will never know how LLMs work. No matter. Begin with the question: how does it do what it does and know what it knows? You will find this approach always useful and (sadly) not always asked in higher ed.

While RE is not a familiar term outside of engineering, the concept translates to other disciplines. In biology, dissection is the most obvious parallel (even without Dr. Frankenstein). Researchers use RE techniques to analyze genomic data, identify regulatory relationships between genes, and decipher the intricate mechanisms governing cellular processes. All of biology is a form of RE, where living systems are analyzed like complex machines to understand design and function.

Even humanities fields like literature or history involve RE principles to deconstruct narratives, analyze historical events, and piece together motivations and influences.

The attraction of DeepSeek’s “how did I get there” responses shows the public wants this. Users are delighted that DeepSeek “thinks out loud” and deconstructs the question, seeming to turn it over in its mind before answering it. This is not RE in the formal sense but I want to point out the conceptual appeal.

That is, DeepSeek starts with dissecting and dismantling. Then it combines patterns, knowledge, and logical inference from training data to forward-engineer a response. It doesn’t RE its answer but it REs your question. (The analogy works best in cases where logical decomposition or problem-solving mirrors backward reasoning.) It’s a mechanistic process aimed at optimizing relevance in its responses.

People love it and that’s a good thing for my argument, which is that the time has come to bring RE methods to all education, as we rethink education for the AI era.

As AI systems become increasingly sophisticated, the ability to take apart and critically examine their outputs becomes not just valuable but essential. The skills developed through RE — systematic analysis, critical questioning, and methodical reconstruction — are precisely what students need to thrive in a world where information is abundant but understanding is scarce.

Reverse Engineering in Higher Ed

First, formal definitions. Reverse engineering (RE) is the systematic process of analyzing a product, device, or system to understand its design, functionality, and underlying principles. Sometimes the goal of taking something apart is simply curiosity. But typically the goal of RE is replicating or improving something. RE is used in software (decompiling code to identify vulnerabilities), hardware design (dissecting to study circuitry), and mechanical engineering (reconstructing components). Cybersecurity, for example, is all about RE for malware analysis and legacy system modernization. While the corporate espionage fear of RE is copying, RE is a legitimate practice for innovation, quality control, and product development.

Reverse engineering is primarily taught in engineering schools, in computer science, electrical engineering, or cybersecurity courses. Engineering departments generally integrate RE into product design or systems analysis curricula, using tools like disassemblers or 3D scanners.¹

In high schools, exposure is more limited but growing through specialized STEM tracks, career and technical education (CTE) programs, or project-based courses in robotics or coding²

Nobody talks about reverse engineering in the humanities, though they should. As arts & humanities dean at Sonoma State some years ago I wanted to pilot a disassembly project, where students would bring old cell phones and computers to disassemble and learn what’s inside of the technology they use every day. It is still a good idea.

Can RE improve general education? Yes. AI could work with RE principles to ensure learning that students can’t simply ask AI to analyze data or write a smooth final paper to demonstrate learning.

Systematic analysis, methodical problem-solving, and iterative improvement are ideal for gen ed in the age of AI. The RE mindset transforms traditional coursework from passive knowledge acquisition into active discovery, learning and how and why things are the way they are. The RE approach of breaking down complex systems into understandable components provides students with transferable analytical skills essential for navigating an AI-enhanced world. This methodology bridges the perceived gap between technical and liberal arts education: systematic thinking traditionally associated with engineering can enhance learning across the entire curriculum.

For example, in a traditional biology gen ed class students are asked to demonstrate their ability to research and synthesize information. Now that AI can perform these tasks, students would "reverse engineer" AI knowledge and responses about, say, cell division. Students could analyze how the AI constructs explanations by asking what sources and concepts it draws upon. Students would try different prompts about cell division and document how varying the prompts leads to different responses.

Students would compare multiple AI responses with scientific literature or textbooks to identify where the AI's knowledge comes from and how it aligns with or diverges those sources. This process of tracing knowledge lineage develops deeper understanding than simply writing a paper, as it shows about how scientific knowledge is built and transmitted. The ultimate goal would be for students to "reconstruct" accurate knowledge about cell division (or any other topic) by understanding how information flows from primary research through various knowledge intermediaries (including AI) to create current understanding of biological processes.

In other words, students would be doing RE: taking something apart to understand how it was built, then using that understanding to create something new.

History works the same way. The reality is that AI can write good papers summarizing the causes and key events of, say, the American Revolution. With an RE approach, students would ask AI systems to explain the American Revolution, then analyze how the AI constructs its historical narrative. Students would follow up with questions like "What primary sources inform your understanding of the Boston Tea Party?" or "How do you weigh different historical interpretations of colonial taxation?"

Again, students would compare the AI's responses with primary documents, academic histories, textbooks, and even TV. Students might discover that the AI's interpretation reflects dominant narratives while overlooking others. The culminating assessment would ask students to "reconstruct" their own understanding of the American Revolution by documenting how different pieces of historical evidence and interpretation fit together. This RE approach transforms history education from memorization and summary into an active investigation of how we know what we know about the past. It uses AI as a tool for understanding both historical events and the process of historical knowledge construction.

Would this work with an intro to economics course? Again, students would ask AI how they model and explain economic principles. A question might be how AI constructs its understanding of market dynamics, with follow-ups like "How do you determine the equilibrium price in the smartphone market?" or "What data sources inform your understanding of elasticity of demand for gasoline?"

Students would then compare AI's economic modeling with real-world market data, academic theories, and economic forecasts. Students might discover that while AI can perfectly describe theoretical supply and demand curves, real-world markets often deviate from these idealizations due to factors like imperfect information or regulatory constraints. A final exam project might require students to "reconstruct" an economic model by documenting how different variables interact. They might build their own predictive model of a simple market, explaining how they incorporated various factors like price elasticity, substitute goods, or external shocks.

Given the rapid advancement of AI and its increasing use in economic forecasting and analysis, developing a RE approach to economics education might be truly valuable.

Critics might argue that incorporating RE principles into gen ed creates unnecessary complexity in courses that should focus on foundational knowledge. They could argue that making students analyze how AI constructs knowledge adds an extra layer of abstraction that might confuse rather than clarify. They might worry that this approach requires significant faculty retraining and curriculum redesign at a time when higher ed already faces numerous challenges. But these objections overlook how AI is transforming the educational landscape. When AI can instantly generate polished essays and solve complex problems, traditional assessments no longer effectively measure student learning. RE approaches provide a necessary framework for ensuring actual learning in an AI-enhanced world. The end is how knowledge is constructed, verified, and transmitted. While implementation will require planning and faculty development, the alternative is continuing with educational approaches that AI has already rendered obsolete.

I’ve spent countless hours at engineering workbenches where literal RE is being performed, watching new things being designed with the manufacturing process in mind, to know that the rigorous thinking that goes into RE is applicable to other fields. I’ve seen how RE requires understanding technology and manufacturing workflow design, asking what equipment is available, how stringent the accuracy requirements are, about material properties (will something melt?) and surface complexity. If you’ve taken something apart you have a better sense of how to put something together.

Where to begin? Business Schools

Looking around higher ed, I suspect business school leaders might be most open to bringing RE principles into business schools. RE is not yet a mainstream component of most b-school curricula, though elements appear in niche courses or programs focused on competitive strategy, innovation, or product management. For example, courses like Competitive Strategy or Business Analytics might use reverse-engineering principles to dissect market leaders’ strategies or workflows. MBA programs with a tech focus may integrate reverse engineering into innovation labs or partnerships with engineering departments. However, explicit teaching of reverse engineering as a formal methodology remains rare, limited to electives.

If business schools were to systematically adopt reverse engineering, particularly with AI-driven tools, courses or modules might include:

AI-Powered Competitive Analysis: While many courses use case studies and SWOT analyses are common, new reverse-engineering frameworks could integrate AI tools to automate the deconstruction of competitors’ public data (financial reports, marketing campaigns, or customer reviews) to identify hidden patterns. AI could reverse-engineer a rival’s pricing algorithm from historical sales data or simulate their supply chain logic using public procurement records. Students would gain real-time, data-driven insights into competitors’ strategies, enabling faster, evidence-based decision-making.

Product and Service Deconstruction with AI Simulations. While engineering students regularly take machinery apart to see how they’re made, physical product teardowns or service-model analyses might not translate well to business programs. But AI could simulate virtual “teardowns” of digital versions of physical products or digital products (SaaS platforms) using code analysis tools or customer journey mapping. For physical products, AI-powered image recognition (e.g., analyzing product photos) or 3D modeling could estimate material costs or assembly processes. Students would be able to benchmark innovations efficiently, identifying cost-saving or differentiation opportunities even without physical access to products.

Ethical IP Navigation with Synthetic Data. Right now, legal and ethical training around IP is theoretical, lacking hands-on practice. But AI-generated synthetic datasets or “mock” proprietary systems could let students safely practice reverse engineering in a controlled environment. For example, dissecting a synthetic AI model’s training data to study bias without legal risks. Students could develop ethical judgment and technical skills while avoiding real-world IP violations.

Operational Workflow Optimization via Process Mining. Currently, process improvement courses often rely on generic frameworks like Six Sigma. AI process-mining tools (Celonis) could reverse-engineer workflows from event logs or user data, revealing bottlenecks in hypothetical (or anonymized) business operations. Students could then redesign workflows for efficiency. This would combine technical reverse engineering and operational strategy, preparing students for roles in supply chain or tech-driven industries.

Entrepreneurship and MVP Development. Current startup courses focus on lean canvases but rarely teach “hacking” successful models. AI could analyze startups’ growth metrics, user feedback, or app features to reverse-engineer their traction drivers. Students could use results to refine their own MVPs. Such a course would accelerate entrepreneurial experimentation by learning from (but not copying) existing successes.

A course like AI-Driven Business Innovation might task students with using ChatGPT to reverse-engineer a viral social media campaign’s engagement triggers or employ no-code AI tools to replicate a competitor’s recommendation engine. Collaborations with computer science departments could enable cross-disciplinary projects, such as reverse-engineering a product’s technical design and business model simultaneously.

Integrating RE into all higher ed will be the future. In entrepreneurship especially, RE using AI would produce graduates adept at decoding complex systems, ethically leveraging competitors’ insights, and driving innovation in an AI-augmented business landscape.

As higher education grapples with how to maintain academic integrity and ensure genuine learning in an era where AI can generate sophisticated papers and responses, an RE approach is a framework that could transform how we assess and develop student understanding.

1. Some examples: Purdue University's "Applications of ECE Through Reverse Engineering" (ECE 39595); UC’s San Diego’s "Introduction to Reverse Engineering with Solidworks"; Grand Valley State University’s "Reverse Engineering and Malware Analysis" (CIS 456); the Naval Postgraduate School’s "Software Reverse Engineering and Malware Analysis" (CS4648); and University of North Georgia’s "Advanced Reverse Engineering" (CSCI 7250).

2. Advanced high schools with engineering academies might include reverse engineering in prototyping labs. Extracurricular clubs (robotics, cybersecurity teams, maker spaces) might explore it informally. Structured coursework remains rare at the K-12 level.