Abstract This research commentary explores how requirements engineering (RE) can contribute to achieving digital humanism in AI-enabled healthcare, ensuring that technological innovation is balanced with human values. Drawing on insights from a panel discussion at the REWBAH’25 (RE for Well-Being, Aging, and Health) workshop, we identify four key themes: the necessity of interdisciplinary collaboration, the nuances of cultural and individual differences, human-AI collaborative decision-making, and the challenge of balancing technological advancement with sustainability and emotional considerations. We argue that RE is well suited for addressing these issues because it brings together diverse stakeholder perspectives to define what systems should achieve and how they should be used. In this paper, we build on digital humanism principles, on RE-relevant frameworks for AI, and on the above themes to frame the role of RE in supporting human-centred healthcare. We highlight open challenges, and outline five promising research directions and related research questions to guide future research and practice in RE, aiming to ensure that AI-enabled healthcare systems better reflect the values of digital humanism.

Neurodegenerative diseases often result in pathological gait patterns, reducing mobility, stability, and overall functional capabilities. Given their impact on older adults' quality of life, early and accurate diagnosis is crucial for timely intervention. Traditional gait assessment technologies present some limitations related to low portability levels and user comfort. In this context, Socially Assistive Robots (SARs) offer an alternative by enabling non-intrusive gait monitoring while also supporting professional caregivers with objective measurements of users' motor performance. This study investigates the feasibility of using a mobile robotic platform to extract and analyze digital biomarkers related to gait activity. A novel pipeline was developed to automatically detect gait parameters from laser sensor data, segment the gait cycle, and compare these measurements against inertial measurement unit (IMU) data, which is the widely used approach. Results demonstrate a strong correlation (CI > 0.7) between laser-derived and IMU-based temporal gait parameters. However, discrepancies in step length measurements suggest that laser-based tracking provides more precise spatial information than IMU estimations. Additionally, this study explores the influence of the robotic platform on gait performance. Findings indicate that users walk faster when the robot is absent, despite its position behind them and out of sight. This suggests an unconscious adaptation to the robot’s presence, aligning with previous studies on human-robot interaction.

Socially Assistive Robots (SARs) represent a valid support to professional caregivers in providing care to person with need. To improve the quality of the interaction, SARs should be able to automatically assess user engagement. In this work, we addressed this problem by investigating user engagement dynamics during a robot-to-human handover task, considering 3 main components of engagement: affective, cognitive, and behavioral. For this study, we automatically extracted 10 visual features from the camera recordings of 31 participants. Each individual engaged in eight consecutive sessions with a robot manipulator designed with social cues. Our statistical analysis indicates that prolonged interaction with the robot could influence user engagement. Namely, we observed a decrease in positive emotions (affective), a more regulated quantity of motion (behavioral), and a reduced attention on the robot tasks (cognitive). Overall, the results of this study suggests that engagement dynamics can be described by the selected behavioral features, and that the a more predictable robot’s behavior could negatively influence user engagement.

Movement is one of the basic tools that humans use to convey emotional states. Body language and movement are also relevant in the perception that we have of other people. Many studies have been done on movement and emotion sharing, involving humans on one side and robots or automated agents on the other. From these studies, there is evidence of the importance of robots to improve their social capabilities and develop more effective social interaction. This work aims at embedding some social movements in a robot manipulator, that will elicit an emotional response from users. Laban Movement Analysis is used to do so, and social movements are developed on a robotic arm, then shown to participants in an online study and a questionnaire. The results show that it is possible to elicit emotions through movement only and that the perception is not affected by personal experiences.

This work aims to investigate if a "robot's personality" can affect the social perception of the robot in the navigation task. To this end, we implemented a dedicated human-aware navigation system that adapts the configuration of the navigation parameters (i.e. proxemics and velocity) based on two different human-like personalities, extrovert (EXT) and introvert (INT), and we compared them with a no social behavior (NS). We evaluated the system in a dynamic scenario in which each participant needed to pass by a robot moving in the opposite direction, showing a different personality each time. The Eysenck Personality Inventory and a modified version of the Godspeed questionnaire were administered to assess the user’s and the perceived robot’s personalities, respectively. The results show that 19 out of 20 subjects involved in the study perceived a difference among the personalities exhibited by the robot, both in terms of proxemics and velocity. Furthermore, the results highlight a general preference of a complementary robot’s personality, helping to suggest some guidelines for future works in the human-aware navigation field.

The European population is getting older; many elderlies would like to live independently in their homes as long as possible. In this context, a robotic telepresence service could support the frail persons in their homes, empowering their social relationships. In this work, 10 frail elderly were asked to live with a telepresence robot (i.e. Double Robot), through which the formal caregiver could remotely visit and chat with them. A total of 169 days of field-test trial was evaluated before (TO) and after (TF) the tests with a multidimensional framework, including acceptance, usability, and expectation domains. The system was used for 2871 mins, and the results underline good usability- and acceptance- related domains (average score equals to 70.83 at TF) and expectation (average score equal to 67.01 at TF). Results remark that expectation could influence the potential and the real use of the robot. Additionally, a positive trend in the answers was identified between T0 and TF. Indeed, the evaluation of a system should envisage a complex, multidisciplinary and holistic approach, that may influence the success or failure of the robot's purpose, if not properly analysed during the evaluation and design phase.

The aim of this work is to provide an automatic analysis to assess the user attitude when interacts with a companion robot. In detail, our work focuses on defining which combination of social cues the robot should recognize so that to stimulate the ongoing conversation and how. The analysis is performed on video recordings of 9 elderly users. From each video, low-level descriptors of the behavior of the user are extracted by using open-source automatic tools to extract information on the voice, the body posture, and the face landmarks. The assessment of 3 types of attitude (neutral, positive and negative) is performed through 3 machine learning classification algorithms: k-nearest neighbors, random decision forest and support vector regression. Since intra- and intersubject variability could affect the results of the assessment, this work shows the robustness of the classification models in both scenarios. Further analysis is performed on the type of representation used to describe the attitude. A raw and an auto-encoded representation is applied to the descriptors. The results of the attitude assessment show high values of accuracy (>0.85) both for unimodal and multimodal data. The outcome of this work can be integrated into a robotic platform to automatically assess the quality of interaction and to modify its behavior accordingly.

Aging society is characterized by a high prevalence of sarcopenia, which is considered one of the most common health problems of the elderly population. Sarcopenia is due to the age-related loss of muscle mass and muscle strength. Recent literature findings highlight that the Tinetti Balance Assessment (TBA) scale is used to assess the sarcopenia in elderly people. In this context, this article proposes a model for sarcopenia assessment that is able to provide a quantitative assessment of TBA-gait motor parameters by means of a cloud robotics approach. The proposed system is composed of cloud resources, an assistive robot namely ASTRO and two inertial wearable sensors. Particularly, data from two inertial sensors (i.e., accelerometers and gyroscopes), placed on the patient's feet, and data from ASTRO laser sensor (position in the environment) were analyzed and combined to propose a set of motor features correspondent to the TBA gait domains. The system was preliminarily tested at the hospital of “Fondazione Casa Sollievo della Sofferenza” in Italy. The preliminary results suggest that the extracted set of features is able to describe the motor performance. In the future, these parameters could be used to support the clinicians in the assessment of sarcopenia, to monitoring the motor parameters over time and to propose personalized care-plan.

Internet of Robotic Things paradigm offers a concrete support to daily life. The pervasiveness of smart things, together with advances in cloud robotics, can help the smart systems to perceive and collect more information about the users and the environment. Often citizens have experienced “one-size-fits-all” approach, since the delivered service was not personalized, therefore resulting far from user's expectations. Hence, future smart agents, like robots, should produce personalized behaviours based on user emotions and moods in order to be more integrated into ordinary activities. In this work, we investigated the performances with unsupervised and supervised approaches to recognize three different moods elicited during a social interaction by means of a wearable system capable of measuring the Electrocardiogram, the ElectroDermal Activity and the Electroencephalographic signals. Particularly, the classification problem was analysed using three unsupervised (K-Mean, Self-Organizing Map and Hierarchical Clustering) and three supervised methods (Support Vector Machine, Decision Tree and k-nearest neighbour). The supervised algorithms reached an accuracy of 0.86 in the best case. The outcomes show that even in an unsupervised context the system is able to recognize the mood, reaching an accuracy equal to 0.76 in the best case.

Human activity recognition is an important and active field of research having a wide range of applications in numerous fields including ambient‐assisted living (AL). Although most of the researches are focused on the single user, the ability to recognise two‐person interactions is perhaps more important for its social implications. This study presents a two‐person activity recognition system that uses skeleton data extracted from a depth camera. The human actions are encoded using a set of a few basic postures obtained with an unsupervised clustering approach. Multiclass support vector machines are used to build models on the training set, whereas the X ‐means algorithm is employed to dynamically find the optimal number of clusters for each sample during the classification phase. The system is evaluated on the Institute of Systems and Robotics (ISR) ‐ University of Lincoln (UoL) and Stony Brook University (SBU) datasets, reaching overall accuracies of 0.87 and 0.88, respectively. Although the results show that the performances of the system are comparable with the state of the art, recognition improvements are obtained with the activities related to health‐care environments, showing promise for applications in the AL realm.

Daily activity recognition can help people to maintain a healthy lifestyle and robot to better interact with users. Robots could therefore use the information coming from the activities performed by users to give them some custom hints to improve lifestyle and daily routine. The pervasiveness of smart things together with advances in cloud robotics can help the robot to perceive and collect more information about the users and the environment. In particular thanks to the miniaturization and low cost of Inertial Measurement Units, in the last years, body-worn activity recognition has gained popularity. In this work, we investigated the performances with an unsupervised approach to recognize eight different gestures performed in daily living wearing a system composed of two inertial sensors placed on the hand and on the wrist. In this context our aim is to evaluate whether the system is able to recognize the gestures in more realistic applications, where is not possible to have a training set. The classification problem was analyzed using two unsupervised approaches (K-Mean and Gaussian Mixture Model), with an intra-subject and an inter-subject analysis, and two supervised approaches (Support Vector Machine and Random Forest), with a 10-fold cross validation analysis and with a Leave-One-Subject-Out analysis to compare the results. The outcomes show that even in an unsupervised context the system is able to recognize the gestures with an averaged accuracy of 0.917 in the K-Mean inter-subject approach and 0.796 in the Gaussian Mixture Model inter-subject one.