Background: The ongoing adoption and use of digital interventions offer promising opportunities to meet the growing demand for mental health support. The effectiveness, implementation, and usage of these interventions depend on how well they are designed and evaluated. However, given the emerging nature of design research in this area, there is still no clear consensus on the specific principles and guidelines for developing digital mental health interventions (DMHIs). There seems to be a lack of clarity regarding the best practices for designing and evaluating these tools. Objective: We aimed to investigate and report on the design principles and evaluation approaches used in digital interventions specific to mental health care. Additionally, we sought to outline how these principles and approaches are applied in research. Methods: This scoping review was conducted in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines for scoping reviews. The literature search was performed in 2 electronic databases, SCOPUS and Web of Science, across 3 iterations from January 2024 to January 2025. A total of 2 independent reviewers screened and selected papers based on predefined inclusion and exclusion criteria, followed by data extraction from the selected studies. The data were then synthesized by categorizing the papers according to the primary research aim of each study. The inclusion criteria covered studies involving populations with mental health challenges or users of DMHIs, any digital tools for mental health care, and principles or strategies related to the design, evaluation, or implementation of DMHIs. Results: Our search identified 401 papers, of which 17 met the inclusion criteria for this review. Among these, 11 focused on evaluation studies, while 6 covered both design and evaluation studies (mixed). An iterative user-centered development process, expert inclusion, usability testing, specification of design elements, and user tracking and feedback were identified as common design principles used in studies focused on DMHIs. Evaluation approaches were shaped by the evaluation goal, which influenced the chosen methodologies. We also summarize the recommendations for implementation highlighted in some studies. Based on our findings, we propose 8 guidelines emphasizing stakeholder involvement in the development process and the need for clear justifications for design decisions, among other considerations. Conclusions: Design principles used in DMHI development include user-centered development, expert inclusion, and usability testing, while evaluation approaches often rely on randomized controlled trials to assess efficacy. Qualitative and mixed-method approaches are commonly adopted by studies to capture user experience and bridge both process and outcome measures. We recommend that future research explicitly report its design justification and adopt a multiperspective approach in the research and design of DMHIs.

Inverse Reinforcement Learning (IRL) seeks to infer reward functions from expert demonstrations. When demonstrations come from multiple experts with different intentions, the problem becomes Multi-Intention IRL (MI-IRL). MI-IRL methods generally follow a rewardfirst paradigm: they ask which reward, if followed, could have generated each trajectory. This leads to trajectory similarity being based on reward likelihood rather than on the relationships between behaviors. In deep generative MI-IRL methods, this paradigm further couples behavior clustering and reward learning, making such methods dependent on prior knowledge of the true number of modes K^∗. This limits adaptability to unseen behaviors, and restricts the analysis to learned rewards rather than relationships across behaviors. In contrast, we approach MI-IRL from a different standpoint. We propose Contrastive Multi-Intention IRL (CoMI-IRL), a transformer-based unsupervised behavior-first framework: it first learns a representation of trajectory dynamics that captures intrinsic behavioral similarity, then clusters trajectories in that space, and finally learns a reward for each cluster. Our experiments show that CoMI-IRL achieves higher clustering and reward learning performance when compared to Ess-InfoGAIL without needing K^∗ or labels, while allowing for visual interpretation of behavior relationships and adaptation to unseen behavior without full retraining.

Real-world decision-making often requires optimizing multiple competing objectives simultaneously. In reinforcement learning (RL), this is typically addressed by combining reward signals into a single scalar objective via a scalarization function, which can be fragile: small changes in the weights can induce drastically different policies. Multi-objective reinforcement learning (MORL) instead produces sets of policies that explicitly represent trade-offs between objectives. However, these policies are typically presented to the decision maker only through their value vectors, which can obscure substantial behavioral variation: policies that induce distinct trajectories may appear indistinguishable when evaluated solely by expected returns. We propose an exploratory diagnostic workflow that automatically highlights behavioral variation along the Pareto front that objective values alone do not reveal, providing both quantitative and visual tools to support policy inspection. We validate our approach on simple grid examples and scale it to continuous control benchmarks, demonstrating that it remains effective as problem complexity increases.

Human values capture what people and societies perceive as desirable, transcend specific situations and serve as guiding principles for action. People’s value systems motivate their positions on issues concerning the economy, society and politics among others, influencing the arguments they make. Identifying the values behind arguments can therefore help us find common ground in discourse and uncover the core reasons behind disagreements. Transformer-based large language models (LLMs) have exhibited remarkable performance across language generation and analysis. However, leveraging LLMs in sociotechnical systems that assist with discourse and argumentation necessitates systematically evaluating their ability to analyse and identify the values behind arguments, an under-explored research direction. Using a multi-level human value taxonomy inspired by the Schwartz Theory of Basic Human Values, we present a systematic and critical evaluation of GPT-3.5-turbo in human value identification from a dataset of multi-cultural arguments, across the zero-shot, few-shot and chain-of-thought prompting strategies, carrying forward from prior research on this task which leveraged a fine-tuned BERT model. We observe that prompting strategies exhibit performance levels close to, but still behind fine-tuning for value classification. We also detail some challenges associated with value classification with LLMs, offering potential directions for future research.

One of the key issues in human-robot collaboration is the development of computational models that allow robots to predict and adapt to human behavior. Much progress has been achieved in developing such models, as well as control techniques that address the autonomy problems of motion planning and decision-making in robotics. However, the integration of computational models of human behavior with such control techniques still poses a major challenge, resulting in a bottleneck for efficient collaborative human-robot teams. In this context, we present a novel architecture for human-robot collaboration: Adaptive Robot Motion for Collaboration with Humans using Adversarial Inverse Reinforcement learning (ARMCHAIR). Our solution leverages adversarial inverse reinforcement learning and model predictive control to compute optimal trajectories and decisions for a mobile multi-robot system that collaborates with a human in an exploration task. During the mission, ARMCHAIR operates without human intervention, autonomously identifying the necessity to support and acting accordingly. Our approach also explicitly addresses the network connectivity requirement of the human-robot team. Extensive simulation-based evaluations demonstrate that ARMCHAIR allows a group of robots to safely support a simulated human in an exploration scenario, preventing collisions and network disconnections, and improving the overall performance of the task.

Smart home technologies embed values such as sustainability, comfort, privacy, and security, which can sometimes conflict with one another, considering the complexities of domestic environments. This paper investigates the potential implications of these value conflicts and the corresponding design challenges. Through an enactment session and co-speculations with professional actors, we explored what it means to navigate multiple values simultaneously, live with products that impose their own values, and manage value conflicts both with and among smart products. The findings challenge the seamless and harmonious vision of smart homes conceived by technologists, proposing shifts in the common narrative: from value alignment to value transparency, from service provision to mutual care, and from autonomy to responsiveness. We discuss that acknowledging value conflicts, rather than eliminating them, is an opportunity to gain a deeper understanding of users and home environments and guide the design of smart home technologies.

This handbook presents the concept of ‘meaningful human control’ (MHC) over AI systems from the perspectives of (i) philosophy and ethics, (ii) law and governance, and (iii) design and engineering. The introductory chapter addresses the motivations and recent developments in MHC, introducing each perspective and related chapters. These three disciplinary perspectives scrutinize how MHC intertwines with philosophical debates on moral responsibility, societal concerns regarding control over technological advancements in legal frameworks, and the engineering complexities of designing and developing AI systems while ensuring human control and responsibility. Additionally, cross-cutting aspects on MHC over AI systems are also introduced and discussed through (iv) interdisciplinary and systemic perspectives. By offering a contextualized introduction to the perspectives considered in this handbook, this chapter aims to present the handbook’s various approaches and points of interest for a diverse audience, highlighting potential entry points into this multidisciplinary volume.

Although popular and effective, large language models (LLM) are characterised by a performance vs. transparency trade-off that hinders their applicability to sensitive scenarios. This is the main reason behind many approaches focusing on local post-hoc explanations recently proposed by the XAI community. However, to the best of our knowledge, a thorough comparison among available explainability techniques is currently missing, mainly for the lack of a general metric to measure their benefits. We compare state-of-the-art local post-hoc explanation mechanisms for models trained over moral value classification tasks based on a measure of correlation. By relying on a novel framework for comparing global impact scores, our experiments show how most local post-hoc explainers are loosely correlated, and highlight huge discrepancies in their results—their “quarrel” about explanations. Finally, we compare the impact scores distribution obtained from each local post-hoc explainer with human-made dictionaries, and point out that there is no correlation between explanation outputs and the concepts humans consider as salient.