It's Like "X": How Engineering Faculty Metaphors Construct (and Constrain) AI Understanding in Engineering Education

Synthesis: Gerhardt, Shiekh, Katz, and Chaback (2026) analyze the figurative language — metaphors and analogies — that engineering instructors use to describe generative AI, drawn from 57 semi-structured interviews across 17 disciplines and seven universities. They map instructors' language onto a five-dimensional taxonomy (Ontology, Epistemology, Operation, Relationship, Power/Capability), finding that instructors overwhelmingly frame GAI either as a human-like "social being/agent" (e.g., "a PhD student," "an assistant") or as a "technical object/artifact" (e.g., a search engine or tool), with almost no framing of GAI as an environmental or systemic force. Because these metaphors both construct and constrain understanding, and because instructors within the same department often hold fundamentally different mental models, the authors argue that developing a shared language is essential for Educational Development and productive departmental discussions about GAI in engineering education.

Metaphor as understanding

The paper begins from the premise that figurative language is not decorative but fundamental to how we think and act: "the way we think, what we experience, and what we do every day is very much a matter of metaphor." Instructors inevitably describe GAI through metaphors — a "mirror" that reflects biases, a "crutch" or "drug" enabling Cognitive Offloading and intellectual degradation, an "assistant," or an "advanced student." Common metaphors for LLMs in wider discourse include "stochastic parrots," "simulators," "crowds/zeitgeist," "gods," "e-bikes," and "drop-in remote workers." In engineering specifically, figurative language is instrumental for conceptual understanding. Yet a lack of shared language between stakeholders can compromise Educational Development programs and departmental discussions.

Five dimensions of instructor language

Analyzing instructors' responses to the prompt "if you had to think of an analogy or metaphor to describe GAI, what would you say?", the authors developed a five-continuum codebook, each with contrasting positions:

• Ontology — what "is" GAI? Instructors most often framed GAI as a Social Being/Agent (54.2%) with human-like agentic and relational qualities (e.g., "a person with deep knowledge who thinks far faster than I do"); second was a Technical Object/Artifact (44.1%) likened to a search engine or tool; least common was an Environmental Force/Systemic Phenomenon (1.7%) such as a tsunami, iceberg, or paradigm shift.
• Epistemology — does GAI genuinely understand? Positions ranged from "Genuine/True Understanding" to "Mimics/Simulates Understanding," often co-occurring with the human-like vs. mechanical framings above.
• Operation — how does it work? "Learning/Developmental" versus "Mechanical/Pattern-Matching."
• Relationship — how do we relate to it? "Collaborative/Partnership" versus "Instrument/Tool-Use."
• Power/Capability — what can it do? "Human-Level/Comparable" versus "Limited/Deficient."

Nearly 75% of "Social Being/Agent" framings co-occurred with Collaborative/Partnership, Human-Level/Comparable, Genuine/True Understanding, and Learning/Developmental positions — forming a coherent human-like-assistant mental model. Conversely, the "Technical Object/Artifact" framing co-occurred with mechanical, tool-use, and limited positions.

Constructing and constraining understanding

The key finding is that engineering instructors do not hold a unified mental model of GAI: even within the same department, semester, and institutional policy environment, instructors construct fundamentally different accounts of what GAI "is" and "means." None is inherently "correct," but each is shaped by personal experience, use cases, and professional background. Critically, some metaphors are inaccurate and can constrain understanding: describing GAI as a "search engine" inaccurately positions it as retrieving pre-existing information, when in fact it algorithmically generates plausible token continuations. Instructors tended to focus on what GAI does rather than what it knows, says, or is.

What this means for practice

• Faculty developers. Open GAI training with metaphor elicitation — ask instructors what analogy they would use, then map the response onto the five dimensions (Ontology, Epistemology, Operation, Relationship, Power/Capability) to surface the mental model driving their teaching decisions.
• Faculty developers. Correct inaccurate artifact metaphors before they harden: describing GAI as a search engine mispositions it as retrieving stored information when it actually generates plausible token continuations, and such framings constrain what instructors expect of students.
• Instructors. Use the taxonomy as a shared conversation starter within a department, since colleagues in the same department, semester, and policy environment constructed fundamentally different accounts of what GAI is — bridging vocabularies rather than enforcing standardized language is the productive move.
• Administrators. Reach beyond early adopters and staff already "speaking the same language" as coordinators when rolling out GAI initiatives; incommensurable instructor mental models are a direct threat to policy coherence and Educational Development programming.

Limitations

• The analysis uses instructors' responses to a single interview question ("if you had to think of an analogy or metaphor to describe GAI to someone, what would you say?"); the rest of each hour-long interview was not analyzed.
• The reported results cover 57 engineering instructors from 17 disciplines at seven universities, collected in Spring 2025, drawn from a larger project of nearly 170 instructors at 18 universities.
• Instructors who could not produce an analogy were not described because of space constraints, so the taxonomy does not represent the struggling cases, and comparisons across disciplines were left to future work.
• This is self-reported figurative language in an interview; the study does not measure whether an instructor's metaphor predicts their actual classroom AI practice.

Citation

Gerhardt, M., Shiekh, K., Katz, A., & Chaback, B. E. (2026). It's Like "X": How Engineering Faculty Metaphors Construct (and Constrain) AI Understanding in Engineering Education. ASEE Annual Conference & Exposition, Paper ID #50720.