Paper ID #47761Assessing ChatGPT 4o for AI-Assisted Problem Solving in Electric CircuitsTeachingDr. Bin Chen, Purdue University Fort WayneDavid S Cochran, Purdue University Fort WayneJeffrey Andrew Nowak Ph.D., Purdue University Fort WayneGuoping Wang, Purdue University Fort Wayne Guoping Wang, Ph.D. is an Associate Professor in the Department of Electrical and Computer Engineering at Purdue University Fort Wayne. He earned his Ph.D. from the University of Oklahoma in 2003, following a Master’s from Nanjing University and a Bachelor’s from Tsinghua University. Dr. Wang’s research interests include the Internet of Things, edge
Capstone CourseKeywords: Capstone Projects, Electrical Engineering Education, Generative AI in Education,ChatGPT, Entrepreneurship in Engineering, Marketing and Design Requirements, ABET.1. IntroductionIn recent years, many engineering programs have integrated entrepreneurship education into thecapstone experience, blending technical engineering skills with entrepreneurial processes,namely ideation, customer discovery, client validation, and commercial viability [3] Theseprocesses enable students to translate their technical knowledge into economically relevantengineering practice. The objective is to produce graduates who are not only technicallyproficient but also capable of navigating the business landscape, ethically aware, and responsiveto
secured multiple grants for innovative projects. A senior member of IEEE, he actively contributes to the field through publications and conference presentations. ©American Society for Engineering Education, 2025 Case Studies of ChatGPT for Embedded Systems TeachingAbstractThe rise of AI technology, particularly Generative AI, has significantly transformed the landscapeof higher education. Generative AI, such as ChatGPT, has been extensively studied in fields likeComputer Science to assess its effectiveness in enhancing learning. However, its impact on morespecialized areas, such as bare-metal embedded systems, remains underexplored. Bare-metalembedded systems, which include hardware (e.g
education. ©American Society for Engineering Education, 2025 AI-Assisted Learning of VHDL Yumin Zhang, Brad Deken Department of Engineering and Technology Southeast Missouri State University Cape Girardeau, MO 63701AbstractVHDL is widely used in digital systems design courses in electrical engineering programs, yetmany students struggle with its steep learning curve. Integrating AI tools like ChatGPT presentsan innovative and effective solution to these challenges. By leveraging AI in the learningprocess, students can benefit from real-time, interactive feedback on coding
. However, there are a few restrictions: • The minimum version of ChatGPT is specified. • Catching ChatGPT making oversimplifications is insufficient. • No purposely incorrect questions are allowed.After finding a misconception, students have to analyze the error and provide a detailedexplanation of what makes ChatGPT wrong. Additionally, they have to explain the concept thatwould correct its misunderstanding.For documentation, students must provide a link to the conversation, so that it is impossible tofake. However, there is still room for academic dishonesty, as there is no way to know if the ideawas suggested by someone else.3 Results and DiscussionWhile the assessment strategy has been well received and enjoyed by students
the latest OpenAI and Anthropic models: ChatGPT 4o [9] and Claude 3.5Sonnet [10]. These two models were selected based on previous benchmark results for reasoningand mathematics [11, 12]. In some of our experiments, the AI LLMs are used ‘as is’ or ‘off theshelf’ - no training and no instructions. In other experiments, the AI LLMs are trained, i.e.instructions are fed into the AI program along with one or more chapters of the course textbook.In evaluating the ability of LLM chatbots to act like a very good TA, we sought to investigatehow the AI TA performs for different amounts of instruction / training prior to asking questions.In our tests, all chatbots are configured with temperature set to 0.3 and a maximum token size of2048.3.3.1
withsports. These findings suggest the need for alternative analogies that better resonate with diversestudent backgrounds.For the solar charging station analogy, 72% of students matched all terms correctly, although someconfusion persisted. For example, 10% mistook ‘DC Source’ for the interface controller, and 15%confused ‘computer controller’ with the image of a cell phone. These findings suggest areas forrefining analogies, particularly in distinguishing components with similar terminology.A survey conducted at the end of the semester confirmed that students preferred real-worldanalogies over AI tools like ChatGPT, highlighting their value in establishing a strong conceptualfoundation and boosting confidence. Table 1 presents key survey results
used search en-gines—primarily Google—or increasingly, LLM-based services like ChatGPT. While Googleprovides reliable but fragmented information, ChatGPT can produce more coherent, user-friendlysummaries but may occasionally “hallucinate” or fabricate content. Users relying on ChatGPTmust manually verify any suggested sources.Despite these limitations, ChatGPT still offers a valuable starting point for discovering potentiallyrelevant books, especially for non-expert users. We leveraged ChatGPT-4o to generate candidateground truth data, then applied rigorous human verification to mitigate misinformation. Our pro-cedure was as follows: • We issued arbitrary queries (main query plus one or more topics) to ChatGPT, asking it to identify
of course mapping and alignment is neither challenging nor time-consumingwith the assistance of ChatGPT. By providing the course coverage, outcomes, and content,ChatGPT was able to generate units, lessons, and related assignments efficiently. However, fine-tuning is necessary to align the generated lessons and units with the specific teaching materialsand objectives of the course.Future work could focus on refining the modular approach by incorporating more interactive andhands-on activities to address feedback regarding engagement. Additionally, expanding the useof this structured alignment method across other disciplines or multi-disciplinary courses couldvalidate its broader applicability. Continuous enhancement through feedback loops
, especially in individual project implementations. Group project work (maximum of two students per group) can help lower the required amount of interaction time. Moreover, hiring a Teaching Assistant (TA) can help lower the interaction time if hiring funds are available, which can be a challenge. Of course, finding a good TA for the job is also a challenge! Continuous creativity: To keep projects challenging and minimize cheating and copying past executed projects, creating new and varied project specifications can be challenging and time consuming. This issue is especially observed in courses that are offered frequently. Threat from AI: To ensure students are not using AI tools, such as ChatGPT, to find solutions for