2026 ASEE Annual Conference & Exposition

The accelerating integration of generative artificial intelligence (Gen-AI) into academic and professional workflows is transforming expectations of how engineers learn, design, and communicate. For first-year students, this transformation occurs now when foundational engineering identity and study habits are formed. Understanding how freshmen perceive, adopt, and critically engage with AI is therefore essential for designing curricula that promote responsible and effective use. This research paper extends a previous study on students’ understanding and expectations of AI, especially on freshmen perspective by conducting a comparative analysis of with new data from a cohort of new freshman in 2025. The goal is to identify evolving patterns in students’ familiarity, confidence, ethical awareness, and practical engagement with AI tools while situating these findings within broader frameworks of technology acceptance and AI literacy development.

Motivation and Background -- Prior work demonstrated that introducing AI-focused lectures and reflective assignments in an “Introduction to Engineering and Computing” course increased awareness and confidence among first-year students but also revealed uncertainty about accuracy, reliability, and ethical boundaries. Since then, widespread media discourse, institutional policy discussions, and the proliferation of free AI tools have dramatically expanded exposure. These changes prompted new research questions: (1) How have freshmen’s perceptions and practices regarding AI evolved in just one year? (2) What relationships exist among familiarity, reliability, trust, and frequency of use? (3) How can early curricular experiences foster critical and ethical engagement with AI as a learning partner rather than a substitute for learning? The present study thus aims to trace shifts in student understanding while refining pedagogical strategies for embedding AI literacy in the engineering curriculum.

Methods - The study employs a mixed-methods design combining quantitative analysis of Likert-scale survey data with qualitative thematic coding of open-ended responses. Parallel pre- and post-surveys were administered (e.g., sample size n = 52 post-survey; n = 53 pre-survey) surrounding a guest lecture on use of AI and a reflection essay assignment on a specific topic, completed both independently and with ChatGPT assistance. Quantitative data were analyzed using paired-sample t-tests to measure changes in familiarity, confidence, and perceived helpfulness, while Pearson correlations examined associations between reliability, frequency of use, and trust. Qualitative responses were coded inductively and triangulated across both survey phases to identify emergent themes. Content validity was established through faculty review, and limitations of self-reported data were explicitly acknowledged. Although confined to a single institution and a course with three sections, the study’s design enables longitudinal comparison with the previous cohort and contributes to a work-in-progress cross-cohort dataset on first-year AI engagement.

Results - Initial findings reveal measurable growth in both familiarity and strategic use of AI tools. Baseline familiarity ratings increased significantly from the prior year, indicating that nearly all students in this course had prior experience with AI before formal instruction. Yet skepticism about reliability persisted, with average ratings of AI accuracy and dependability remaining moderate (around 3.9 of 5). Post-survey results show significant gains (p-value < 0.05) in self-efficacy for using AI productively and ethically, alongside greater appreciation of prompt design as a literacy skill. Students described iterative refinement such as rephrasing, adding context, and specifying constraints as key to improving ChatGPT’s responses, demonstrating emerging metacognitive awareness of human-AI interaction.

Cross-cohort comparison underscores that students now arrive with heightened expectations of AI integration but also more nuanced critical stances. This suggests a maturing digital culture in which early exposure breeds literacy rather than dependence. Correlational results show that perceived reliability and helpfulness strongly predict usage frequency, consistent with technology acceptance models, thereby supporting the conceptual validity of the survey instrument for longitudinal use.

Implications - Pedagogically, the study confirms that early structured engagement, through guided reflection, explicit ethics instruction, and comparative human-versus-AI exercises, can transform passive awareness into active literacy. Embedding “AI communication competence” in first-year courses can help students learn how to interrogate, verify, and integrate AI outputs responsibly. Curriculum design should therefore treat AI literacy as a foundational engineering skill comparable to programming or technical writing. The results also highlight the importance of balancing enthusiasm for efficiency with reinforcement of human creativity and ethical accountability.