It's like "X": How Engineering Faculty Metaphors Construct (and Constrain) GAI Understanding in Engineering Education
Plain Language Summary
In this paper, we explore the everyday comparisons — metaphors like "assistant" or "search engine" — that engineering instructors reach for when asked to describe generative AI (GAI), since those comparisons are not just color commentary but shape how instructors, and in turn their students, come to understand what these systems are and how they work. Drawing on interviews with 57 engineering instructors, we organize the language they use into five dimensions: what GAI fundamentally is, whether it truly "understands" anything, how it operates, how people relate to it, and what it can and cannot do. The point is not to catalog every metaphor instructors used, but to show that instructors hold no single shared picture of GAI, which matters for anyone designing training or policy around it.
Contribution
Building on prior figurative-language research that has focused mostly on students, this paper contributes an empirically-derived, five-dimensional codebook — refined across three coding passes from an initial 68-item literature-based list to a consolidated 48-item Version 3 codebook — that captures engineering instructors' figurative conceptualizations of GAI along ontology, epistemology, operation, relationship, and power/capability continua, rather than treating individual metaphors as discrete, non-overlapping types.
Research Questions
- What figurative language — metaphors and analogies — do higher education engineering instructors use when asked to describe GAI?
- Along what conceptual dimensions (e.g., what GAI "is," how it "knows," how it operates, how instructors relate to it, and what it can do) does this figurative language vary across instructors?
- What do these patterns imply for how faculty developers might support instructors' understanding of, and communication about, GAI in engineering education?
Methods
We draw from 57 semi-structured interviews, conducted in Spring 2025 with engineering instructors across 17 disciplines and seven universities, part of a larger multi-institutional study of instructors' mental models of GAI; the analysis here focuses specifically on instructors' responses to a single protocol question asking them to offer an analogy or metaphor for GAI. Deductive coding started from a literature-derived compendium of 68 GAI metaphors organized via Maas' five-dimensional taxonomy, with inductive coding that let new categories emerge from the data. We iterated across three coding passes (Version 1, 2, and 3 codebooks) with member-checking after each. The final codebook reorganizes 48 metaphors/analogies into five continua, each with three positions, applied across 417 coded response segments.
Key Findings
Instructors most often conceptualized GAI ontologically as a "Social Being/Agent" (54.2%) rather than a "Technical Object/Artifact" (44.1%), with human-like framings co-occurring heavily with "Collaborative/Partnership," "Human-Level/Comparable," and "Genuine/True Understanding" codes. Epistemologically, instructors were nearly evenly split between viewing GAI as genuinely understanding (29.8%), functionally/contextually understanding (31.6%), or merely mimicking understanding (38.6%), though notably none used the word "consciousness." Relationally, 87.7% of instructors described some human-GAI relationship, and of those, 64% framed it as "Collaborative/Partnership" versus 32% as "Instrument/Tool" and only 4% as "Adversarial/Threat"; on the Power/Capability continuum, half described GAI as roughly "Human-Level/Comparable," with "Limited/Deficient" (31.5%) more common than "Superhuman/Superior" (18.5%).
Implications
Because instructors do not share a unified mental model of GAI — even within the same department or under the same institutional policy — we argue faculty development efforts should treat this taxonomy as a conversation-starter rather than push toward standardized language, helping instructors distinguish functional from ontological anthropomorphism and recognize where their own metaphors obscure GAI's actual technical operation. The low occurrence of "Adversarial/Threat" and "Environmental Force/Systemic Phenomenon" framings, despite active concerns about bias and infrastructure costs in the broader literature, suggests these considerations warrant more explicit attention in training than this single-question analysis surfaces.