TY - GEN
T1 - Evaluating AI Chatbots in RL and RLC Circuit Problems: Benefits, Pitfalls, and Student Perceptions
AU - Horne, Christopher
AU - Robledo Rella, Victor Francisco
PY - 2026
Y1 - 2026
N2 - This study investigates the use of generative AI chatbots in supporting undergraduate students’ understanding of RL and RLC circuits at Institution 1 in the USA and Institution 2 in Mexico, during the Spring and Summer of 2025. Three course sections were studied (Ntotal = 79): G1 solved an RL circuit with AC, G2 solved an RLC circuit with filters and AC, and G3 solved an RL circuit with constant input voltage. Students were assigned a multi-part task involving (1) solving the problem by hand, (2) prompting an AI chatbot of their choice, and (3) reflecting on the comparison between their work and the chatbot’s output. Submissions were evaluated using a structured rubric to assess conceptual depth, hallucination frequency, and prompting skills. An Interaction Parameter (IP) is defined to classify the student interaction with the chatbot, ranging from IP = 1, which corresponds to an unstructured prompt with superficial interaction, to IP = 4, which corresponds to a well-structured prompt where the student can identify mistakes made by the chatbot. The results from the G1 and G2 groups indicate that while AI chatbots were advantageous for procedural guidance, students often encountered partially correct or misleading outputs. High-quality prompts led to more coherent responses, while low-quality prompts produced vague or erroneous advice. Students reported improved confidence, though some showed signs of over-reliance on AI suggestions. For the G3 group, which dealt with a simpler problem, almost all the answers provided by the chatbots were correct. A student perception questionnaire administered to the G3 students indicates that 95% of the students consider that the use of the chatbots helped them better understand the concepts, while 82% of them admit to regularly use AI chatbots to do their assignments. These results contribute to the growing body of work showing that, although structured student–AI interactions enhance procedural accuracy, explicit guidance is still required for higher-order conceptual understanding.
AB - This study investigates the use of generative AI chatbots in supporting undergraduate students’ understanding of RL and RLC circuits at Institution 1 in the USA and Institution 2 in Mexico, during the Spring and Summer of 2025. Three course sections were studied (Ntotal = 79): G1 solved an RL circuit with AC, G2 solved an RLC circuit with filters and AC, and G3 solved an RL circuit with constant input voltage. Students were assigned a multi-part task involving (1) solving the problem by hand, (2) prompting an AI chatbot of their choice, and (3) reflecting on the comparison between their work and the chatbot’s output. Submissions were evaluated using a structured rubric to assess conceptual depth, hallucination frequency, and prompting skills. An Interaction Parameter (IP) is defined to classify the student interaction with the chatbot, ranging from IP = 1, which corresponds to an unstructured prompt with superficial interaction, to IP = 4, which corresponds to a well-structured prompt where the student can identify mistakes made by the chatbot. The results from the G1 and G2 groups indicate that while AI chatbots were advantageous for procedural guidance, students often encountered partially correct or misleading outputs. High-quality prompts led to more coherent responses, while low-quality prompts produced vague or erroneous advice. Students reported improved confidence, though some showed signs of over-reliance on AI suggestions. For the G3 group, which dealt with a simpler problem, almost all the answers provided by the chatbots were correct. A student perception questionnaire administered to the G3 students indicates that 95% of the students consider that the use of the chatbots helped them better understand the concepts, while 82% of them admit to regularly use AI chatbots to do their assignments. These results contribute to the growing body of work showing that, although structured student–AI interactions enhance procedural accuracy, explicit guidance is still required for higher-order conceptual understanding.
UR - https://2026.ieee-educon.org/
M3 - Other contribution
T3 - IEEE EDUCON 2026
ER -