Abstract:

Generative AI tools are increasingly present in undergraduate’s everyday study practices. However, their educational value in STEM depends on students’ ability to craft prompts that are accurate, constrained, and verifiable. Weak prompts tend to elicit vague or incorrect outputs; structured prompts help communicate audience, scope, and validation steps, skills that are aligned with scientific practice.

This paper reports a short, in-class intervention to develop AI prompting skills in a first-year Micro and Nanotechnology Engineering course, immediately after a lecture on band structure (valence band, conduction band, band gap) and material classes (conductors, intrinsic/extrinsic semiconductors n/p, insulators).

We implemented a within-subject pre/post design to avoid contamination of the diagnostic phase. In the pre-test, students (individual work) wrote a single prompt to explain the differences between conductors, semiconductors, and insulators using band-structure concepts, then executed it in a chatbot and saved the output. After a brief micro-intervention introducing a structured prompting pattern (ROLE/CONTEXT → OUTPUT → LIMITS → CHECK), students were asked to perform a new prompt on another class related subject with additional details and explicitly requested to consider the best practices presented to achieve structured outputs (for example, a “layer table” summarising material–band relations). Students also validated two factual claims against the lecture PDF and completed a three-item self-assessment. Submissions were collected via Moodle (online text with optional attachments) in two distinct groups.

Our objectives were to:
- characterise baseline prompting competence;
- teach and elicit application of a structured prompting pattern;
- evaluate change in prompt quality and format compliance;
- examine students’ confidence and fact-checking; and
- provide a replicable 60–90-minute protocol integrated with the institutional LMS.

Scoring focused on clarity, specificity, alignment with taught concepts, format compliance, and (when applicable) the correctness of the concept-device linkage. We contribute a classroom-ready pattern that connects band-gap reasoning with concise, auditable AI outputs and a lightweight analytic schema combining pre/post scores with PDF-based fact checks.

Keywords:

AI literacy, prompting, micro and nanotechnology education, band structure, first-year STEM, Moodle, classroom intervention.