Artificial Intelligence as the Epistemological Maturation of Computer‐Assisted Engineering Education

Recent work in engineering education describes the current moment as an inflection point. In a commemorative editorial marking the 35th anniversary of Computer Applications in Engineering Education, Iskander 1 traces the development of the field from early simulation software to AI-driven educational ecosystems. The notion of inflection captures rapid acceleration—technological, institutional, and cognitive—but acceleration alone does not explain the structural nature of the current transformation. This paper argues that AI does not replace previous educational paradigms but extends them by supporting higher-order reasoning processes, such as generating solution steps or identifying inconsistencies. To understand this transformation, it is not sufficient to describe new technologies; it is also necessary to examine how they change the way knowledge is produced and validated in engineering education. Since its founding in 1992, CAE has documented an evolving relationship between computational tools and engineering formation. Early contributions emphasized visualization, interactive modeling, and multimedia modules. The key innovation of that period was not simply digitization, but the transformation of how abstract phenomena became accessible through dynamic visualization. Simulation did not replace theory; it mediated it. The expansion of virtual and remote laboratories extended mediation from representation to experimentation, enabling parameter manipulation and comparison between simulated and physical systems. Learning analytics later introduced descriptive insight into learner behavior. At each stage, computational mediation reorganized—rather than displaced—engineering reasoning. Within this historical development, artificial intelligence should be understood not as a rupture, but as a continuation and deepening of computer-assisted engineering education. If simulation externalized dynamic models and analytics externalized behavioral patterns, AI extends computational mediation into interpretive and generative processes. The transformation is therefore primarily epistemological rather than technological: the key shift concerns how engineering knowledge is produced, validated, and applied. In this sense, AI marks the epistemological maturation of computer-assisted engineering education, extending rather than displacing the computational paradigm documented in CAE since 1992. Systemic analyses of AI integration 2 further indicate that adoption cannot be treated as a pedagogical add-on. AI reshapes curriculum structure, assessment design, faculty roles, and governance mechanisms. Two common interpretive errors must be avoided. Technological determinism assumes that AI will inevitably transform engineering education regardless of curricular and disciplinary choices. Defensive isolation, by contrast, attempts to preserve “pure” human cognition by excluding AI from meaningful analytical participation. Between these extremes lies the need for structured human–AI collaboration. Engineering reasoning has always relied on external artifacts—mathematical notation, diagrams, and computational tools. AI extends this lineage by incorporating pattern recognition, generative modeling, and adaptive feedback 3. Yet these capacities introduce qualitative shifts: AI systems can propose solutions, optimize parameters, and generate symbolic derivations. The central challenge, therefore, is not whether to adopt AI but how to govern its use effectively. How can engineering education preserve disciplined reasoning while incorporating systems capable of autonomous generation? How can assessment verify understanding when generative systems produce technically correct outputs? These questions frame the present inquiry. This paper advances three propositions. In doing so, it complements existing empirical and application-focused studies by offering a conceptual framework that clarifies the role of AI within the evolution of engineering education. First, the evolution of computer-assisted engineering education can be understood as a layered architecture of mediation, in which AI represents an adaptive layer continuous with prior developments. Second, AI integration requires reconceptualizing engineering education as disciplined co-intelligence—a structured collaboration between human reasoning and AI systems, in which students may use AI to generate solutions but remain responsible for verifying, interpreting, and justifying them. For example, an AI tool may generate a transfer function, but the student must validate its correctness and explain its physical meaning. Third, assessment practices must evolve accordingly. The objective is not to celebrate technological novelty, but to clarify conceptual continuity and define the conditions under which AI strengthens rather than dilutes engineering formation. If the early 1990s marked the beginning of computational mediation, the present moment marks its epistemological maturation. The task is not to choose between human and artificial intelligence, but to articulate the principles governing their structured collaboration. To understand the integration of artificial intelligence, it is necessary to examine how knowledge is developed and validated in computer-assisted engineering education. The adoption of computational tools was not arbitrary, but aligned with constructivist and model-centered approaches in engineering education and cognition. Engineering education has long been based on a constructivist approach in practice. Knowledge develops through engagement with problems, models, and experimental systems rather than passive reception. In engineering contexts, this orientation manifests in modeling, laboratory experimentation, and problem-based design. When computational simulations entered the curriculum, they did not introduce constructivism; they expanded its operational scope. Simulation environments enabled parameter manipulation, dynamic visualization, and iterative refinement of hypotheses. Abstract mathematical relationships, such as those describing dynamic systems, electromagnetic fields, or circuit behavior, became easier to explore through interactive representations. Reasoning unfolded not only symbolically but experimentally within structured digital environments. Cognitive apprenticeship theory 4 helps explain this process. Expertise develops through modeling, guided practice, articulation, and gradual reduction of support. Early computer-assisted systems operationalized aspects of this logic by externalizing expert reasoning pathways and scaffolding solution strategies. If simulation made domain models more accessible, intelligent systems now support aspects of diagnostic reasoning, such as identifying errors or suggesting solution paths. Research on model-based reasoning supports this perspective. Hestenes 5 argued that engineering cognition is fundamentally model-centered: engineers construct, test, and validate models of physical systems. Gilbert 6 emphasized the mediating role of models between theory and empirical observation. Computer simulations extended this tradition by enabling dynamic and manipulable representations of system behavior. Research on simulation-based learning further demonstrated that unguided exploration rarely produces robust conceptual understanding 7, 8. Structured scaffolding remains essential. This insight becomes even more significant in AI-mediated contexts. If simulation required instructional regulation, generative systems require governance. Artificial intelligence can produce symbolic derivations, propose optimizations, and generate explanatory narratives 9. The key issue is not whether AI supports reasoning, but how responsibility for interpretation and validation is shared between the system and the learner. When AI generates intermediate steps or alternative designs, the locus of understanding must remain disciplined and accountable. The historical trajectory documented within CAE reveals continuity: technological mediation deepens, yet responsibility for understanding remains with the learner. Simulation enhanced visualization; virtual laboratories enhanced experimentation; analytics enhanced insight into learning patterns. Artificial intelligence should be understood as part of this ongoing development. Recent reviews and studies (e.g., Holmes and Miao 3 and Liu et al. 10) describe AI applications in engineering education, including intelligent tutoring, automated grading, predictive analytics, and conversational agents. However, they primarily focus on technological implementation and pedagogical applications, while leaving less explicit how AI reshapes the structure of knowledge and reasoning. This paper addresses this gap by proposing an epistemological interpretation of AI as a new stage in the evolution of computer-assisted engineering education. A systemic perspective emphasizes that AI integration spans classroom practice, curriculum design, institutional governance, and ethical oversight. The maturation thesis advanced here rests on a historical pattern: each extension of cognitive artifacts—notation, diagrams, computational tools, and simulations—required new forms of regulation without displacing human responsibility. AI extends interpretive and generative processes, but extension does not imply substitution. It requires structured integration within established epistemic traditions. Before discussing implementation or assessment, a key principle must be stated clearly: artificial intelligence must reinforce model-based reasoning, sustain disciplined guidance, and preserve the learner's active construction of understanding. Only under these conditions can AI be seen as a continuation of existing educational practices, rather than a disruption within a coherent epistemological framework. If artificial intelligence is to be interpreted as epistemological maturation rather than disruption, the evolution of computer-assisted engineering education must be described in structural terms. The layered mediation model provides such a framework. Rather than listing technologies, it identifies progressively deeper forms of mediation between disciplinary knowledge and analytical practice. Each layer changes how knowledge is accessed, used, and validated in engineering learning. The layered mediation model describes how computational tools have progressively changed engineering learning, from visualization (simulation), to experimentation (virtual labs), to adaptive reasoning support (AI). For example, circuit simulation software visualizes system behavior, while AI systems can now suggest solution steps or detect algebraic errors. In practice, this means that instructors should design tasks in which students are required to explain and verify AI-generated steps, rather than simply accept them. Representational mediation characterized the early decades of the CAE journal. Computational simulations transformed abstract mathematical relationships into dynamic visual models. Differential equations describing circuit transients, electromagnetic fields, or stress distributions became interactively observable. The main contribution of this layer was to make representations more accessible: reasoning gained perceptual support. Yet the systems remained deterministic. The learner manipulated predefined models; the tool performed calculations but did not interpret results. Experimental mediation extended representation into structured experimentation. Virtual and remote laboratories 11 enabled parameter variation, hypothesis testing, and comparison between simulated and physical systems. This stage strengthened the connection between models and their validation. Interpretation, however, remained fully human. Even with the rise of learning analytics, which provided meta-level insight into learner interaction patterns, decision-making remained entirely in the hands of the learner. Adaptive mediation introduces a qualitative shift. Artificial intelligence systems can diagnose partial reasoning, suggest symbolic transformations, generate derivations, or propose optimized parameter configurations. In Electric Circuit Theory, for example, an AI system may analyze an incomplete transfer-function derivation and identify algebraic inconsistencies. At this stage, computational systems no longer only execute calculations, but begin to support parts of the reasoning process. However, this does not mean that the system replaces human judgment. Without structured validation, adaptive mediation risks replacing careful analysis with unverified automated outputs. Strategic mediation extends beyond classroom interaction to the organization of educational institutions. Curriculum alignment, assessment design, accreditation requirements, and governance of AI use become central. Learning management infrastructures integrate AI-mediated interaction logs, validation checkpoints, and assessment workflows. At this level, mediation involves not only learning processes, but also rules and structures that guide their use. The distribution of analytical authority between human and artificial intelligence becomes a matter of institutional design rather than individual preference. The layered model clarifies two principles. First, artificial intelligence does not abolish prior layers; representational and experimental infrastructures remain foundational. Second, AI externalizes interpretive and generative processes, introducing new capabilities that require clear rules and guidance. The model thus preserves historical continuity while acknowledging structural novelty. It supports the claim that AI constitutes maturation within an evolving mediation architecture rather than abrupt discontinuity. The integration of artificial intelligence requires rethinking how human reasoning and AI systems work together, not just adopting new technologies. If AI constitutes an adaptive layer within the historical architecture of mediation, then the relationship between human and artificial intelligence must be clearly defined. Disciplined co-intelligence refers to a structured form of collaboration between human reasoning and AI systems, in which the system may support solution generation, but responsibility for interpretation, validation, and justification remains with the learner. Engineering reasoning has always relied on external tools, such as notation, diagrams, computational solvers, and simulations. Artificial intelligence continues this lineage with expanded capacity. Human engineers provide judgment, interpretation, and decision-making under real-world constraints. AI contributes rapid computation, large-scale pattern recognition, symbolic manipulation, and generative modeling. These roles are complementary, not identical. In Electric Circuit Theory, an AI system may compute transfer functions or simulate transient responses efficiently; interpretation of stability, physical meaning, and design trade-offs remains a human responsibility. This distinction becomes particularly relevant in analytical contexts such as Electric Circuit Theory, where AI can efficiently generate transfer functions or simulate responses, while interpretation remains a human task. For example, classroom activities can require students to compare AI-generated solutions with their own derivations and justify any differences. Complementarity alone does not ensure rigor. Generative systems may produce outputs that appear correct, even when their underlying assumptions are not clear. By governance, we mean the rules and practices that ensure AI-generated results are checked, understood, and validated by students. Such governance includes structured validation checkpoints, independent derivation requirements, explicit comparison between AI-generated and manually derived results, and documentation of reasoning processes. The objective is not restriction but disciplined engagement. AI should enhance analytical fluency rather than bypass it. Learning management infrastructures function within this regulatory layer by embedding validation sequences, logging AI-mediated interactions, and structuring assessment workflows. These practices should be built into the educational system, not left to individual choice. Engineering education cultivates habits of precision, validation, and responsibility. AI should be used in ways that strengthen, not weaken, students' analytical skills. AI integration must reinforce—rather than erode—these habits. Instant symbolic transformation or numerical optimization must not replace conceptual understanding or procedural transparency. The shift from assistance to partnership changes how future engineers are trained. Students must learn not only to solve problems but to supervise algorithmic reasoning, recognize limitations, and justify decisions in the presence of machine-generated alternatives. As AI becomes embedded in professional practice, engineering identity expands to include algorithmic oversight. Disciplined co-intelligence, therefore, connects earlier forms of computer-assisted learning with current AI-based approaches, linking historical mediation layers to institutional transformation. It frames AI not as external disruption but as internal expansion of engineering cognition—an expansion that requires explicit governance to remain aligned with disciplinary values. If artificial intelligence represents epistemological maturation, it also introduces vulnerabilities. Every historical expansion of cognitive artifacts—notation, computational solvers, the distribution of cognitive AI extends this by in intermediate reasoning processes. The of epistemic risks is not to but to define the conditions under which disciplined co-intelligence remains describes the to machine-generated outputs as even when is In engineering validation is a significant For example, a generative system may produce a solution for a or generate stress and that appear technically Yet without independent of and the reasoning remains The is not results, but procedural Engineering reasoning has relied on of analytical testing, and empirical AI outputs should therefore be treated as to explicit validation. In educational practice, this incorporating explicit validation steps into such as students to and A concerns the gradual of symbolic Engineering education on In Electric Circuit Theory, students governing symbolic transformations, and interpret and conditions When AI generates intermediate symbolic steps the may be correct, while procedural is of intermediate reasoning can analytical from the concerns through the concerns to the need for clearly of independent reasoning within workflows. A is epistemic of understanding When AI systems generate and may primarily with outputs rather than principles. cannot be to or Engineering responsibility requires the to diagnose when outputs physical or underlying Disciplined co-intelligence therefore between reasoning, and independent conceptual AI systems through internal In on outputs with professional of this that AI-mediated reasoning must include explicit of assumptions and structural Students must example, understanding the physical of in circuit than accept symbolic becomes a pedagogical of design. AI integration should validation checkpoints, reasoning define of independent and institutional for thus becomes rather than The presence of procedural and epistemic does not the maturation the it the need for structured forms of human–AI collaboration. of governance capable of within expanded cognitive ecosystems. The central assessment in engineering education is not but how can and knowledge be when artificial intelligence in intermediate reasoning If AI becomes embedded in modeling, and design processes, assessment must evolve accordingly. educational systems artificial technological or of disciplinary rigor. provides a interpretive Generative systems support and procedural may in analytical However, under and within disciplinary remain on structured validation and judgment. In framework becomes operational rather than must three and professional requires independent of derivations and governing principles. In Electric Circuit Theory, this includes transformations, and system without AI assessment of analytical concerns interaction with AI systems. Students must derivations, detect algebraic verify and justify or of outputs. the cognitive task from to validation. on the of rather than the correctness of results. professional structured human–AI collaboration in design contexts. Students may AI to optimize or simulate responses, yet must articulate reasoning, justify and validation is analytical reasoning with AI A clarifies the In a Electric Circuit students analyze a circuit under AI system may generate a transfer function, symbolic and transient In a assessment students must the governing compare their results with AI verify and behavior, and explain physical therefore, analytical validation conceptual interpretation, and of human–AI Such requires support. Learning management environments can validation checkpoints, AI-mediated interactions, and require explicit of reasoning thus from to governance. This requires that and of reasoning, rather than only correct models correctness with understanding. In contexts, correctness alone is of learning. must verify and collaboration with AI systems. In doing so, becomes aligned with the layered mediation architecture between and institutional governance. If artificial intelligence represents a new stage in the evolution of computer-assisted engineering education, where computational tools beyond simulation and begin to support intermediate reasoning processes, its beyond classroom The adaptive layer cannot remain it reshapes curriculum faculty roles, accreditation and institutional governance. The is no longer whether AI will be but how its presence will be structured to preserve disciplinary while enabling Engineering education has they became foundational. Artificial intelligence must a but with careful AI does not imply with generative tools. It requires understanding model limitations, validation and the ethical of design. In this sense, AI extends engineering responsibility. as engineers must understand the of numerical solvers, they must understand the assumptions and of generative systems. must therefore integrate algorithmic reasoning, and validation practices within disciplinary rather than them as skills. a this perspective that engineering should not the adoption of AI tools, but structured human–AI collaboration. Curriculum design should preserve analytical tasks that students must while also incorporating activities in which and justify machine-generated outputs. should therefore learning three independent conceptual analytical practice, and design tasks that require validation under constraints. For curriculum development, this means embedding explicit for reasoning validation and documentation of AI that students learn not only to solutions but to their and engineering learning and professional In a must structured AI of and explicit between and that AI with professional that integrate AI without governance of The layered mediation model a conceptual for institutional AI integration within and infrastructures rather than as Artificial intelligence faculty roles from of solutions to of cognitive ecosystems. to include structuring AI-mediated embedding validation checkpoints, and epistemic development must therefore beyond tool governance design. Students not only but reasoning When faculty model disciplined of AI systems, they professional of As a the of and engineering education, CAE a point. historical documentation of computational mediation now of governance principles. the of AI in engineering education, it emphasizes innovation and The present work contributes a perspective by on governance, validation, and epistemological a structured framework for AI without disciplinary rigor. The from early simulation environments to AI-driven systems is not technological it is transformation of cognitive By engineering may within cognitive in which derivations remain AI modeling, validation are and institutional collaboration between human and artificial The of such on curriculum design, assessment faculty development, and governance structures must disciplined co-intelligence as a For may include on validation and the responsible use of generative systems. AI does not the distinction between human and it clarifies their roles within structured collaboration. three computer-assisted engineering education has through of pedagogical Simulation enhanced Virtual laboratories extended experimentation learning Artificial intelligence now introduces adaptive and generative capacity. this it a deeper of how engineering knowledge is Artificial intelligence externalizes interpretive within an evolving mediation The future of computer-assisted engineering education will not be by the presence of artificial intelligence, but by the to which analytical responsibility is The central is governance. Without structured AI risks procedural and epistemic explicit regulation, it can validation practices, and engineers for professional environments in which design is The trajectory from computational mediation to intelligent partnership does not engineering reasoning, but it. The responsibility now rests with the engineering education to ensure that this partnership understanding rather than dilutes it. new or in this are from which are in the