This study evaluates the usage of OpenAI’s ChatGPT Large Language Model (LLM) as a tool for constructing multiple choice questions for assessing student academic performance through quizzes and exams. We randomly deploy questions constructed with and without use of the LLM tool and gauge the ability of the students to correctly answer, as well as their ability to correctly perceive the difference between human-authored and LLM-authored questions. In determining whether the questions written with the aid of ChatGPT were consistent with the instructor’s questions and source text, we computed representative vectors of both the human and ChatGPT questions using SBERT and compared cosine similarity to the course textbook. A non-significant Mann- Whitney U test (z = 1.018, p = .309) suggests that students were unable to perceive whether questions were written with or without the aid of ChatGPT. However, student scores on LLM-authored questions were almost 9% lower (z = 2.702, p < .01). This result may indicate that either the AI questions were more difficult or that the students were more familiar with the instructor’s style of questions. Overall, the study suggests that while there is potential for using LLM tools to aid in the construction of assessments, care must be taken to ensure that the questions are fair, well-composed, and relevant to the course material.

Recent decades have seen a resurgence of interest in electroencephalography (EEG), as neuroscience develops new models of cognition and refines old ones, associating them with detectable indicators of brain activity. This article presents a more direct measure of programmer expertise, derived from noninvasive observation of the brain’s electrical activity. This article provides a foundational approach for investigating the role of expertise in programming language comprehension, showing that this electrical activity in the brain can indicate (1) prior programming experience by class level (current state of progression through an undergraduate computer science program), and (2) self-reported experience levels.

Increasingly power-hungry processors have reinforced the need for aggressive power management. Dynamic voltage scaling has become a common design consideration allowing for energy efficient CPUs by matching CPU performance with the computational demand of running processes. In this paper, we propose Interaction-Aware Dynamic Voltage Scaling (IADVS), a novel fine-grained approach to managing CPU power during interactive workloads, which account for the bulk of the processing demand on modern mobile or desktop systems. IADVS is built upon a transparent, fine-grained interaction capture system. Able to track CPU usage for each user interface event, the proposed system sets the CPU performance level to the one that best matches the predicted CPU demand. Compared to the state-of-the-art approach of user-interaction-based CPU energy management, we show that IADVS improves prediction accuracy by 37%, reduces processing delays by 17%, and reduces energy consumed of the CPU by as much as 4%. The proposed design is evaluated with both a detailed trace-based simulation as well as implementation on a real system, verifying the simulation findings.

Wireless Network Interface Cards (WNICs) are part of every portable device, where efficient energy management plays a significant role in extending the device's battery life. The goal of efficient energy management is to match the performance of the WNIC to the network activity shaped by a running application. In the case of interactive applications on mobile systems, network I/O is largely driven by user interactions. Current solutions either require application modifications or lack a sufficient context of execution that is crucial in making accurate and timely predictions. This paper proposes a range of user-interaction-aware mechanisms that utilize a novel approach of monitoring a user's interaction with applications through the capture and classification of mouse events. This approach yields considerable improvements in energy savings and delay reductions of the WNIC, while significantly improving the accuracy, timeliness, and computational overhead of predictions when compared to existing state-of-the-art solutions.