Starting with AI: First Year Engineering Students Perceptions of Generative Artificial Intelligence
Open accessAbstract
This complete paper discusses how Generative AI (Gen AI) is poised as a disruptive tool for industry and a challenge for education.Academic institutions are issuing policies concerning student use of GenAI, yet student perspectives on use, value, and risk remain unprobed.If GenAI is to effectively join the ranks of routine engineering tools like Finite Element Analysis, Computer Aided Design, and Computational Fluid Dynamics, its adoption will require a more active pedagogical role than mere permission or prohibition through policy.Using Rogers' diffusion of innovations, we track first-year engineering students as they move along the five perceived attributes that shape adoption: relative advantage, compatibility, complexity, trialability, and observability.Early results from a survey administered in the third week of a Introduction to engineering course in a large, southeastern R1 university show that students claim relative advantage (speed, convenience) and basic operational know-how with respect to GenAI tools but report ambiguous compatibility with academic ethical norms and uneven institutional support that makes competent use feel uncertain.In most prior courses, sanctioned opportunities to try GenAI and to see its use in practice have likely been scarce, leaving gaps in trialability and observability and, in turn, dampening diffusion.We examine how these perceptions evolve across a first-year general engineering course in which GenAI use is structured, scaffolded, and required.By situating classroom activities as institutional supports that explicitly increase trialability and observability, we explore how students' reported understanding develops with respect competence and ethical practice.The contribution of this work is to provide a baseline of student views at the point of where institutional policy is under development, and to provide evidence on how targeted scaffolding and organizational support shift the perceived attributes about GenAI's diffusion in early engineering education.
AI snapshot
Written from the paper’s full text.
TL;DR
In a two-section first-year general engineering course where GenAI use was structured, scaffolded, and sometimes required, students’ reported GenAI use and perceived support increased, and their reflections moved from vague aspirations to more concrete, bounded future-use intentions. The authors interpret this as partial diffusion via increased trialability/observability, while ethical compatibility remained contested.
Key findings
- Reported GenAI use shifted upward: never-use fell from 17.0% (23/135) pre to 0.8% (9/115) post; daily use rose from 14.1% (19/135) to 20.9% (24/115); >4x/week rose from 35.6% (48/135) to 51.3% (59/115).
- Five Likert items improved significantly (Welch’s t-test, p<0.001): environmental impacts 3.20 ± 1.18 to 3.80 ± 1.02; familiarity with specific tools 3.69 ± 0.90 to 4.21 ± 0.71; list advantages 3.94 ± 0.75 to 4.32 ± 0.59; list disadvantages 4.08 ± 0.77 to 4.47 ± 0.60; perceived support 3.57 ± 0.75 to 3.94 ± 0.76.
- Qualitative coding showed RA.QualityInsight rising from 76% (76/100) midsemester to 97% (72/74) summative; OBS.CapabilityLimits rising from 50% (50/100) to 76% (56/74); ACTION.VagueGeneric falling from 47% (47/100) to 11% (8/74); ACTION.ConcreteTargeted at 49% (36/74) in summative; and TRIAL.FutureIntent rising from 10% (10/100) to 51% (38/74).
- Ethical compatibility remained contested: SHIFT.ExplicitChange was 27% (20/74), SHIFT.ExplicitUnChanged 18% (13/74), COMP.PriorNormsRestriction 2.7% (2/74), and among students expressing future intent, 30/38 also referenced at least one limitation.
Methods
- Mixed methods in two sections of a first-semester general engineering course at a large southeastern R1; total end-of-semester enrollment was 135; pre survey n=135, post survey n=115; analyses were cohort-level and de-identified.
- Quantitative data: categorical GenAI use frequency and Likert-style items on GenAI literacy and perceived support; frequency reported as counts/percentages with chi-square tests, Likert items as means/SD on a 1–5 scale, with Welch’s t-test for pre/post gains.
- Qualitative data: midsemester formative and end-of-semester summative reflections; deductive coding with an a priori Rogers’ Diffusion of Innovations codebook, plus an LLM as an independent alternate coder; interrater checks used micro-precision, micro-recall, micro F1, and mean row-wise Jaccard similarity, with notable disagreement in STANCE/DEPTH/ROLE families.
Limitations
- Stated: findings are highly context-specific to two sections, one instructor, and one institution, so they are not broadly generalizable.
- Stated: evidence comes from surveys and reflection artifacts, not independent observation of GenAI behavior outside assignments; required/framed activities may reflect assignment design; pre/post results are cohort-level; and the a priori DOI lens with limited emergent coding may under-emphasize themes outside the framework.
- Evident from the text: the post “never” count “0.8% (9/115)” is arithmetically inconsistent with n=115, and the institution is described as southeastern R1 in the abstract/context but as mid-Atlantic in the introduction.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
0 of 7 cited sources checked · 0 supported · 0 flagged
- Findingown data
Five Likert items show statistically significant gains from pre- to post-assessment (Welch's t-test, p<0.001).
“Five Likert items show statistically significant gains from pre-to -post assessment (Welch's t-test, p<0.001) (see Table 2).”
Quote in the paperSubstantial claim - Findingown data
Even when students acknowledge the learning value of the tool, several excerpts describe GenAI as feeling like "cheating" or as "morally wrong".
“Even when students acknowledge the learning value of the tool, several excepts describe GenAI as feeling like "cheating" or as "morally wrong".”
Quote in the paperSubstantial claim - Findingown data
In the pre survey (n=135), 17.0% of respondents (23/135) reported never having used GenAI in a typical week, compared to 0.8% (9/115) in the post survey (n=115).
“In the pre survey (n=135), 17.0% of respondents (23/135) reported never having used GenAI in a typical week, compared to 0.8% (9/115) in the post survey (n=115).”
Quote in the paperSubstantial claim - Findingown data
Reported GenAI use frequency among first-year engineering students shifted upward from the beginning of the semester to the end of the semester.
“Reported GenAI use frequency shifted upward from the beginning of the semester to the end of the semester (see Table 1).”
Quote in the paperSubstantial claim - Findingown data
OBS/CapabilityLimits increased from 50% (50/100) in the midsemester to 76% (56/74) in the end-of-semester reflections.
“OBS/CapabilityLimits increased from 50% (50/100) in the midsemester to 76% (56/74) in the end of semester reflections.”
Quote in the paperNot a substantial claim
- Findingown data
Daily GenAI use increased from 14.1% (19/135) at the start of the semester to 20.9% (24/115) at the end of the semester.
“At the high-use end, daily use increased from 14.1% (19/135) at the start of the semester to 20.9% (24/115) at the end of the semester.”
Quote in the paperSubstantial claim - Findingown data
In summative reflections, most students expressing future intent to use GenAI also referenced at least one limitation (30/38).
“In summative reflections, most students expressing future intent to use GenAI also referenced at least one limitation (30/38), suggesting that future use often co-occurred with recognition of tool limitations.”
Quote in the paperNot a substantial claim - Findingown data
The increase in concrete intent statements was accompanied by a decrease in language about vague aspirational use (ACTION.VagueGeneric 11%, 8/74).
“The increase in concrete intent statements correlated with a decrease in language about vague aspirational use (ACTION.VagueGeneric 11%, 8/74).”
Quote in the paperSubstantial claim - Findingown data
Across both midsemester and end-of-semester reflections, students commonly described GenAI as beneficial for improving the quality of their work or thinking (RA.QualityInsight).
“Across both midsemester (n=100 coded excerpts) and end of semester reflections (n = 74 coded excerpts), students commonly described GenAI as beneficial for improving the quality of their work or thinking (RA.QualityInsight).”
Quote in the paperSubstantial claim - Findingown data
Ethical compatibility of GenAI remained contested across the cohort.
“Ethical compatibility did remain contested across the cohort.”
Quote in the paperSubstantial claim - Findingown data
Prior prohibition of GenAI (mostly in K-12 contexts) was present but quite rare in the data sets.
“Finally prior prohibition (mostly in in K-12 contexts) was present but quite rare in the data sets.”
Quote in the paperNot a substantial claim - Findingown data
Students often noted limitations in what GenAI could reliably do in the context of the assignments (OBS.CapabilityLimits).
“Students often noted the limitations in what GenAI could reliably do in the context of the assignments (OBS.CapabilityLimits).”
Quote in the paperNot a substantial claim - Interpretationreasoning
The observed pattern aligns with DOI's claim of uncertainty reduction through trialability and observability, encouraged through organizational support.
“This pattern aligns with DOI's claim of uncertainty reduction through trialability and observability, encouraged through organizational support.”
Quote in the paperNot a substantial claim - Interpretationreasoning
Prohibition-only approaches to GenAI policy risk the exact conditions under which DOI predicts slow or nonexistent adoption.
“Against the context of rapidly increasing use of GenAI in the workplace, prohibition-only approaches to GenAI policy risk the exact conditions under which DOI predicts slow or nonexistent adoption.”
Quote in the paperSubstantial claim - Interpretationown data
Students' change in perceptions over the semester is characterized as a shift from general, abstract claims of awareness of GenAI's capabilities and limitations to more bounded and specific future use intentions.
“Here we characterize students' change in perceptions as a shift from general, abstract claims of awareness of the capabilities and limitations of GenAI to more bounded and specific future use intentions.”
Quote in the paperSubstantial claim - Interpretationreasoning
In instances where policy precludes legitimate trialability and observability, students are more likely to position GenAI usage as cheating rather than as a bounded learning tool.
“As such, in these instances, students are more likely to position GenAI usages as cheating rather than as a bounded learning tool.”
Quote in the paperSubstantial claim - Interpretationreasoning
Across the semester, the clearest shifts in student sensemaking were in trialability/observability and in compatibility rather than in simple positivity towards GenAI.
“Across the semester, the clearest shifts in student sensemaking were in trialability/observability and in compatibility rather than in simple positivity towards GenAI.”
Quote in the paperSubstantial claim - Premisecitation
Unlike many other disruptive technologies, GenAI adoption is complicated by questions of academic integrity, authorship, transparency, fairness, and sustainability.
“Unlike many other disruptive technologies, GenAI adoption is complicated by questions of academic integrity [6], authorship [7], transparency [8], fairness [9], and sustainability [10].”
Quote in the paperNot a substantial claim0 of 3 sources checked- cites [6]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- cites [7]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- cites [8]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Interpretationown data
The study presents evidence that explicit policy clarity coupled with structured, learning-centric and modeled use can increase students' reported support and authentic understanding of the capabilities and limitations of GenAI.
“In contrast, our study presents evidence that explicit policy clarity coupled with structured, learning-centric and modeled use can increase student's reported support and authentic understanding of the capabilities and limitations of the technology.”
Quote in the paperSubstantial claim - Interpretationown data
Social and ethical stigma around GenAI persist even after successful trial observations of the tool.
“Others continue to note the environmental impacts as ethically untenable, indicating that social and ethical stigma persist even after success trial observations of the tool.”
Quote in the paperSubstantial claim - Interpretationreasoning
Direct observation and ethical norming can move students toward concrete, ethically bounded intentions for future use.
“Direct observation and ethical norming can move students toward concrete, ethically bounded intentions for future use.”
Quote in the paperSubstantial claim - Interpretationreasoning
The mandated interventions in the course appear to have partially relieved some of the tension by providing students with legitimate exemplars of learning-supportive use cases.
“The mandated interventions in the course appear to have partially relieved some of the tension by providing students with legitimate exemplars of learning-supportive use cases.”
Quote in the paperNot a substantial claim - Interpretationown data
Students' pairing of future use intentions with explicit acknowledgement of tool limitations suggests that their adoption of GenAI was a mature, bounded posture rather than a simple observation of efficiency or usefulness.
“Importantly, students' end of semester reflections often paired future use intentions against explicit acknowledgement of tool limitations, suggesting that the adoption of GenAI tools was a mature, bounded posture as opposed to a simple observation of efficiency or usefulness.”
Quote in the paperSubstantial claim
- May combine two claims. 13
- Classified as probably not a substantial claim. 4
May combine two claims.
Findingown dataRA.QualityInsights was present in 76% of midsemester excerpts (76/100) and 97% of summative excerpts (72/74).
“RA.QualityInsights was present in 76% of the midsemester excerpts (76/100) and 97% of summative excerpts (72/74).”
Quote in the paperNot a substantial claimMay combine two claims.
Findingown dataChanges in student positivity towards GenAI were not observed, but more detail and intent about how students planned to use GenAI moving forward were observed.
“Changes in student attitudes in positivity towards GenAI were not observed, but more details and intent on how students planned on using GenAI moving forward were.”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataSummative reflections more often included concrete targeted next steps (ACTION.ConcreteTargeted 49%, 36/74) and explicit intent to continue using GenAI in the future (TRAIL.FutureIntent 51%, 38/74).
“Summative reflections more often included concrete targeted next steps (ACTION.ConcreteTargeted 49%, 36/74) and students explicitly stated intent to continue using GenAI in the future (TRAIL.FutureIntent 51%, 38/74).”
Quote in the paperSubstantial claimMay combine two claims.
Findingown dataIn midsemester reflections, students frequently concluded with broad aspirational intentions (ACTION.VagueGeneric 47%, 47/100) and limited explicit discussion of future tool use (TRIAL.FutureIntent 10%, 10/100).
“In midsemester reflections, students frequently concluded the reflections with statements displaying broad aspirational intentions (ACTION.VagueGeneric 47%, 47/100) and limited explicit discussion of future use of the tools (TRIAL.FutureIntent 10%, 10/100).”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S3 (load-bearing).
Findingown dataA small fraction of students referenced prior educational contexts where GenAI use was discouraged or prohibited (COMP.PriorNormsRestriction 2.7%, 2/74).
“A small fraction of students also referenced prior educational contexts where GenAI use was discouraged or prohibited (COMP.PriorNormsRestriction 2.7%, 2/74).”
Quote in the paperNot a substantial claimMay combine two claims.
Findingown dataThe largest reported impacts were in knowledge of environmental impacts of GenAI usage and the ability to name multiple GenAI tools.
“The largest reported impacts were in knowledge of environmental impacts of GenAI usage and the ability to name multiple GenAI tools (ChatGPT, Claude, Copilot, etc.).”
Quote in the paperNot a substantial claimMay combine two claims.
Findingown dataA notable subset of end-of-semester summative excerpts included direct statements that students' views of GenAI had shifted over the semester (SHIFT.ExplicitChange 27%, 20/74), while others reported explicitly that their views had not changed (SHIFT.ExplicitUnChanged 18%, 13/74).
“A notable subset of end of semester summative excerpts included direct statements that students' views of GenAI had shifted over the course of the semester (SHIFT.ExplicitChange 27%, 20/74), though others reported explicitly that their views had not changed (SHIFT.ExplicitUnChanged 18%, 13/74).”
Quote in the paperNot a substantial claimMay combine two claims.
Findingown dataStudents reported increased ability to list the advantages and disadvantages of GenAI tools and an increased perception of support to use GenAI tools.
“Students also reported increased ability to list the advantages and disadvantages of GenAI tools and noted an increased perception of support to use GenAI tools.”
Quote in the paperSubstantial claimClassified as probably not a substantial claim.
InterpretationreasoningThe study's findings should be interpreted as highly context-specific rather than as broadly generalizable.
“As such, our findings should be interpreted as highly context-specific rather than as broadly generalizable.”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S2 (checkable), S4 (based).
InterpretationreasoningIf GenAI is to mature into a normalized engineering tool like FEA, CAD and CFD, higher education will need to play a more active pedagogical role than simple permission or prohibition through policy.
“If GenAI is to mature into a normalized engineering tool like finite element analysis (FEA), computer aided design (CAD), and computation fluid dynamics (CFD), higher education will need to play a more active pedagogical role than simple permission or prohibition through policy.”
Quote in the paperNot a substantial claimMay combine two claims.
Interpretationown dataStudents had more direct experience and a first-hand view of GenAI's limitations and arrived at more explicit ethical framing of GenAI use.
“Students had more direct experience and a first-hand view of limitations and arrived at more explicit ethical framing of GenAI use.”
Quote in the paperSubstantial claimProbably not a substantial claim: may not pass S2 (checkable). May combine two claims.
PremisecitationGenerative AI is a disruptive technology with unprecedented capabilities to both enhance and undermine student learning, especially in higher education.
“Generative artificial intelligence (GenAI) is a disruptive technology with unprecedented capabilities to both enhance [1] and undermine student learning [2], especially in higher education [3].”
Quote in the paperNot a substantial claim0 of 3 sources checked- cites [1]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- cites [2]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- cites [3]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
May combine two claims.
InterpretationreasoningStudents may be able to name the pros and cons of GenAI usage abstractly but may not be able to arrive at a legitimate model of how to use GenAI without short-cutting learning or compromising ethical norms.
“Here we see that students may be able to name the pros or cons of tools usage abstractly but may not be able to arrive at a legitimate model of how to use GenAI without short cutting learning or compromising ethical norms.”
Quote in the paperSubstantial claimMay combine two claims.
InterpretationreasoningThe course interventions increased observability not only of productive and ethical uses of GenAI but also of its failure modes in practical use, enabling students to align their ethical viewpoints with the technology.
“This suggests that the inventions increased observability not only of productive and ethical uses of the technology, but also of failure modes of the technology in practical use, enabling students to align their ethical viewpoints with the technology.”
Quote in the paperSubstantial claimMay combine two claims.
PremisecitationMany institutions are in the process of developing policies on GenAI usage, but student perspectives on its use, value, and risk have been less systematically examined.
“While many institutions are in the process of developing policies on GenAI usage [4], student perspectives on its use, value, and risk have been less systematically examined.”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [4]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
May combine two claims.
Interpretationown dataWhen GenAI use is structured, scaffolded and unambiguous, students' perceptions shifted in three ways: intent moved from vague generalized aspirations to concrete plans for future use, students more frequently articulated specific limitations and capabilities, and perceived support and reported use increased.
“When GenAI use is structured and scaffolded and unambiguous, students' perceptions shifted in three ways: 1) intent moved from vague, generalized aspirations to concrete plans for future use, 2) student more frequently articulated specific limitations and capabilities, and 3) perceived support and reported use increased.”
Quote in the paperSubstantial claim
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Post-survey percentage for never using GenAI is incorrect
MinorNine of 115 respondents is approximately 7.8%, not 0.8%.
Reported 0.8 · recomputed 7.826087
“compared to 0.8% (9/115) in the post survey (n=115)”
Cites
The publisher hasn’t deposited this paper’s reference list.
Research works printed but not linked: 1
- M. Sag, "Fairness and Fair Use in Generative AI," Fordham Law Review, vol. 92, pp. 1887-1921, 2024.
- Authors
- Chaback, B., & Gray, D.
- Affiliations
- Virginia Tech (US)
- Venue
- 2026 ASEE Annual Conference & Exposition Proceedings
- Publisher
- ASEE Conferences
- Published
- 2026-08-21
- OpenAlex
- W7203862023
- Semantic Scholar
- fc1ea67566bae5aff40a426b2306ca6a2d5eec70
- Cited by
- 0
- References
- 0
- FWCI
- 0.00
- Type
- conference paper
- Open access
- bronze
- doi.org
- Locations
- ASEE Annual Conference & Exposition (publishedVersion)
- Funders
- American Society for Engineering Education
- Topics
- Physical Sciences › Computer Science › Information Systems › Artificial Intelligence in Education
- Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
- Physical Sciences › Engineering › Architecture › Engineering Education and Pedagogy
- Keywords
- generative AI, first-year engineering students, technology adoption, diffusion of innovations, student perceptions, relative advantage, compatibility, complexity, trialability, observability, scaffolding, institutional support, academic ethics, engineering education, AI policy in higher education
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-14.
Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-14.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
