Generative AI (GenAI) is rapidly transforming design education, but universities primarily govern GenAI through macro-level policies that may fail to match micro-level course practices, leaving students and instructors without practical guidance. We reviewed 23 top-ranked university GenAI policy documents and 48 empirical studies of GenAI-integrated design courses, using inductive analysis to build paired thematic frameworks and then comparing policies and practices. Our findings reveal strong alignment around academic integrity, transparency, and the emphasis on critical thinking and human judgment. However, gaps appear with policies prioritizing institutional risk governance and academic writing support, whereas practices emphasize pragmatic benefits and iterative multimodal production. Additionally, policies stress independent learning and lifelong skill development, while practices highlight creativity and efficiency alongside persistent inquiry bottlenecks, overreliance, and fixation. We provide a macro-micro analytical lens linking policy to practice, integrative frameworks grounded in student and instructor perspectives, and actionable implications for university governance and course design.

The number and layout of options within pie menus are crucial for determining the direction of radial swiping in mid-air interactions, influencing the physical demands imposed on users. This research examined six distinct option quantities, referred to as menu breadth, and investigated two layouts, designated as menu orientations. These orientations encompass landscape and portrait formats, with the former positioning a top option centred parallel to the horizontal axis, while the latter situates two top options on either side of the vertical axis. The study employed comprehensive metrics, including muscle activity, wrist movement, user performance, and subjective assessments. Our results indicated that the breadth should be controlled to no more than six to ensure high efficiency and align with users’ interactive experience, while breadth 2 should be avoided to eliminate the potential risk of wrist musculoskeletal disorders. Besides, the portrait orientation for breadth 3 and the landscape orientation for breadth 6 were recommended for their lower physiological loads and higher efficiency. Also, the landscape orientation for breadth 4 should be favoured as it outperformed portrait orientation in both wrist movement dimensions. These findings could provide scientific design principles for using pie menus in mid-air interaction within future HCI applications.

Freehand interaction is a promising method for virtual locomotion. While previous studies have proposed various freehand gestures, little is known about the optimal choice among the most frequently used gestures at different locomotion speeds. This study investigates the differences in physiological metrics, performance, and subjective assessments of three distinct gestures (IndexSlide, PalmSlide, and FistSlide) during waypoint navigation tasks at three speeds. Results indicated that PalmSlide was an optimal choice for prolonged use and effective locomotion at slower and moderate speeds during waypoint navigation tasks, as it offered significant benefits in reducing muscle strain, shortening task completion time, and enhancing the user experience. Additionally, IndexSlide showed reduced task completion time, fewer collisions, and an overall more enjoyable experience at the faster speed, suggesting that it might be the preferable choice for virtual navigation scenarios that require rapid manipulation during high-speed locomotion, such as competitive gaming settings.

Evidence shows that visual properties of on-screen objects determine their function and operation, thus influencing gesture elicitation. Therefore, this two-stage experimental study investigated how user-defined mid-air gestures elicited by on-screen visual properties affect human biomechanics, behaviour, and perception. Results showed that, psychologically, for the command Next, swiping with the visual flow provided benefits over swiping perpendicular to it; for Remove, swiping perpendicular to the visual flow offered advantages compared to swiping with it. Behaviourally, for Next, swiping left generated shorter response times than swiping up with a stimulus possessing a horizontal visual flow. Physiologically, for Next, swiping up induced higher forearm muscular load than swiping left under the vertical visual flow condition; for View, tapping twice resulted in less finger muscular load under the multiple entities condition than the one entity condition. These findings could provide a scientific basis for guiding appropriate mid-air gesture design considering on-screen visual properties.

Research on visual menus, specifically panel and pie menus, has gained widespread attention in recent times. However, limited research focuses on the mid-air interaction when comparing these two menu types. This study aimed to conduct a comprehensive comparison between panel and pie menus in the context of cursor-based mid-air interaction within a photo booth setting. The evaluation involved analyzing muscle activities, performance metrics, and user experience to quantify the differences between the two menus across three menu operations. Our results indicated that the pie menu offered notable advantages by reducing muscular loads, achieving shorter task completion times, and delivering an overall more enjoyable user experience, making it a preferable choice for mid-air interaction in photo booth settings. Conversely, the panel menu exhibited superior scores of learnability and demonstrated enhanced accuracy in the switching operation, suggesting that it might be a better choice for complex and precision-demanding tasks.

Research on gesture input modalities, including touchscreen and mid-air interactions, is getting more widespread. A few studies have compared these two input modalities according to subjective methods. However, there is no empirical research on their differences in physiological measures. The purpose of this study was to quantify the differences between touchscreen and mid-air interactions when performing swipe gestures in four orthogonal directions on a tablet combining the indices of electromyography (EMG), electrogoniometry, user performance, and subjective assessment. Our results indicated that mid-air interaction obtained significantly lower muscular loads in the upper limb, smaller wrist joint excursions, shorter task completion time, and better subjective ratings than touchscreen interaction when users with elbow support performed swipe gestures on a tablet. We also found that performing swipe gestures in the vertical direction tended to possess higher muscular loads than in the horizontal direction during touchscreen interaction. Besides, we revealed that swipe right had the largest radial/ulnar deviation, and swipe down had the largest flexion/extension excursion. Furthermore, swipe up brought the worst subjective ratings among the four gesture types. These findings could provide a scientific basis for guiding the appropriate selection and use of the two input modalities in the future HCI field.

Traditional gesture elicitation studies generally adopted the frequency ratio to select popular gestures among users. However, the chosen gestures were not always optimal and might pose a potential risk of musculoskeletal disorder under long-term use. The purpose of our research was to apply a novel assessment system combining the indices of electromyography (EMG), electrogoniometry, and subjective preference to the end-user elicitation experiment. In this study, we conducted a two-stage experiment to compare and analyze the results of the physiological and psychological measures for 33 candidate gestures of 16 given commands. Our results indicated that this assessment system could effectively determine the candidate gestures with lower physiological loads for all given commands except Maximize, and most of the selected gestures were in line with users’ mental models. Finally, we developed an optimal user-defined gesture set for the commands accommodated to human-computer interaction (HCI).