Understanding human gaze communication from a video is critical for decoding complex social interactions in dynamic, real-world environments. Existing gaze communication models focus on a single interaction, such as mutual gaze or shared attention, leaving the full spectrum of dyadic gaze states unaddressed. Unlike low-level gaze tracking that focuses on eye movement anatomy, this work addresses high-level gaze behaviors such as mutual gaze, referential gaze, and shared attention, which reflect the social-cognitive functions of gaze in multi-agent contexts. To this end, a spatiotemporal transformerbased framework is proposed, which involves human-object detection and tracking, gaze-following prediction, and a robust spatiotemporal transformer architecture for fine-grained classification and localization of these gaze behaviors. Moreover, the proposed model incorporates human gaze information, which provides explicit, fine-grained cues about each individual’s focus of attention, allowing more precise alignment of visual features with underlying social intent. Evaluated on a benchmark dataset, the proposed model substantially improves over strong graph-based and transformer-based baselines, particularly in accurately identifying rare yet socially meaningful gaze behaviors. This study contributes a scalable architecture for multi-class gaze analysis, supporting socially aware AI systems in healthcare through applications like autism screening and social engagement assessment, as well as in robotics and behavioral science.

Autistic children often face challenges in reciprocal communication leading to difficulties with social turn-taking (STT), affecting their social interactions. Prior technical interventions show limitations in personalization and sensory adaptability, while also exhibiting scalability constraints. However, recent advances in generative artificial intelligence (GenAI) offer the opportunity to develop adapted tools. Therefore, this Ph.D. work aims to design and evaluate a GenAI-driven, gamified tool to support STT in reciprocal communication for verbal autistic children. The Gamified Adaptive Interaction Platform (GAIP) integrates GenAI personalization with gamification. This research will deliver this key contribution: a novel methodology for adaptive STT interventions using multimodal behavioural sensing to support reciprocal communication. GAIP plans to advance autism interventions through scalable, adaptive AI, advancing HCI and autism inclusion.

Gamification has emerged as a powerful educational tool, offering innovative ways to engage students and enhance learning outcomes. This study investigates the impact of gamifi-cation on seventh-grade students learning mathematics in Arabic, addressing a gap in research on its effectiveness in Arabic K-12 mathematics education. Using a between-subjects design with two groups, this study examined academic performance and motivation via a digital learning system. The system included three activities: a video tutorial, a messaging app simulation, and a treasure hunt game, each incorporating different gamification elements. Academic performance was measured through pre-and post-tests and an end-of-unit quiz, while motivation was assessed using the Instructional Materials Motivation Survey (IMMS). Results showed significant improvement in academic performance in the digital learning group compared to traditional methods, as measured by post-pre test scores$(\mathrm{M}_{\text{diff}}= 17.4\%, \mathrm{p} <. 001)$. However, the end-of-unit quiz showed no significant difference between the two groups. Additionally, students re-ported higher motivation scores for the messaging app simulation and treasure hunt game versus the video tutorial across all components of the IMMS. The study also observed increased peer learning and engagement, although initial unfamiliarity with the digital approach led to challenges that were overcome in subsequent classes during the study. These findings contribute to understanding the potential of gamification in Arabic mathe-matics education, highlighting promising outcomes in academic performance, motivation, and engagement.

Objective: Existing research reporting an association between smartphone usage and sleep quality has often utilised subjective self-reported smartphone usage and sleep data.This paper aims to study the associations of objectively collected smartphone near-bedtime usage and problematic internet usage (PIU) with parameters of sleep quality.Methods: The dataset had 269 (55% Female, 55.13% Adults) participants.From the acquired usage data, the daily averages of sleep duration, sleep distraction, and smartphone usage two hours before sleep were extracted.Results: The multivariate linear regression showed that the increase in PIU (β = -0.231,p < 0.001) and smartphone usage two hours before sleep (β = -0.246,p < 0.001) led to decrease in sleep duration.Regarding sleep distraction, multivariate linear regression analysis revealed that two hours before sleep was a significant and positive predictor of sleep distraction (β = 0.197, p = 0.003), whereas PIU was not significant.Conclusion: Longer duration of smartphone usage before sleep and higher PIU were associated with reduced sleep duration and continuity.PIU predicted the possibility of getting distracted while usage before sleep predicted the distraction duration.Our results confirm and elaborate on concerns about technology overuse near bedtime and call for specialised interventions to help healthier technology design and usage styles.

The paper explores the synergy between wearable data and augmented Artificial Intelligence (AI) through findings from two interconnected studies. The first study (study 1) focuses on medical professionals’ perceptions of wearable data and AI, and the second study (Study 2) extends it focuses on how differences in the level of granularity in the data presented affect the professionals’ understanding, interpretation, and trust in AI recommendations. This system allows medical professionals to view AI-generated recommendations for sleep and activity improvement and explanations of the underlying rationale. While each study has distinct research questions, Study 2 builds upon Study 1's foundation. Both studies employed scenario-based interviews. Thematic analysis of Study 1 identified trust as a crucial factor in the acceptance of wearable data and AI, influencing Study 2's exploration of factors affecting trust, such as explainability, data granularity, representativeness, and user interaction. Study 2 highlighted varying perspectives on information sufficiency and data sharing from the AI system linked to professionals’ roles and tasks. The work offers insights into data granularity’s impact on engagement with AI recommendations.

The proliferation of Artificial Intelligence (AI) technologies, exemplified by Large Language Models (LLM), has ushered in a transformative era across various fields. As the AI revolution will impact societies in complex and uncertain ways, it is likely that persons tending towards belief of conspiracy theories also tend to form more negative and less positive attitudes towards AI. Such persons might believe that some evil force will use AI to destroy human mankind. Drawing on the Interplay of Modality, Person, Area, Country/Culture, and Transparency categories (IMPACT) framework, this study aims to investigate the interplay of locus of control (LOC), belief in conspiracy theories, and the perception of the importance and availability of eXplainable AI (XAI) on attitudes towards AI (measured via AI acceptance and fear). The study used an online survey with 281 participants from the UK and 281 from the Arab world. Statistical analysis revealed that in the UK but not in the Arab sample, female participants reported higher fear of AI and lower acceptance of AI compared to males. The regression results consistently confirmed the role of internal LOC, perceived XAI importance, and perceived availability of XAI in fostering AI acceptance, as well as the role of belief in conspiracy theories, external LOC, and perceiving availability of XAI as being low in increasing fear of AI. The perceived availability of XAI emerges as a crucial influencing factor; addressing it appropriately could enhance societal awareness and acceptance of AI while reducing fear. Personal factors and XAI influence attitudes towards AI in both Arab and UK cultures, enhancing result robustness and revealing nuanced differences.

This article explores the multifaceted relationship between social media use and individual wellbeing (SM-WB). It focuses on personality traits, locus of control, social media competency, and cultural backgrounds. An online survey was conducted with 281 Arabs (141 females) and 281 British (155 females). Analyses revealed significant differences: Arabs exhibited higher belief in social media's positive impact on wellbeing, consistent across various wellbeing dimensions measured through a customized PERMA scale. Regression analysis identified significant predictors of SM-WB—social media competency, conscientiousness, agreeableness, neuroticism, and internal locus of control—for both samples, while females reported higher SM-WB than males. Age was a significant predictor exclusively in the British sample, whereas extraversion predicted SM-WB only in the Arab sample. Qualitative findings regarding future social media design to enhance wellbeing revealed several similarities between samples; however, some themes differed in their specifics, underscoring culture’s nuanced impact on social media developmental requirements for improving wellbeing.

The role of intelligent agents in research on AI is becoming increasingly significant. The study of embodied intelligent systems, such as social and companion robots, is growing steadily in proportion to the previsions that they will become increasingly involved in the everyday lives of individuals. Non-embodied intelligent systems, such as chatbots or virtual voice assistants, are already part of many people’s everyday experience, being constantly at their disposal. This necessitates a re-evaluation of society as we know it, which must be conceived as a ‘hybrid society’. The objective of this paper is to demonstrate how, in such a scenario, a truly human-centred development of intelligent agents necessitates a design that does not capitalise or exploit the natural inclination towards human attachment and empathy, but rather incorporates this into a more responsible framework for human-agent interaction (henceforth HAI). In order to achieve this objective, the topic of human dependency is presented, analysed, and interpreted in the light of one of its main theorisation, formulated in the socio-legal domain. The merit of such a theoretical framework is that it is flexible enough to be applied to all those elements that, once introduced into society, can affect it and alter its equilibrium. This is undoubtedly the case with the artificial agents designated for HAI. Building on this, the paper demonstrates how such an approach could be employed to analyse the dynamics of HAI in a truly human-centred manner, thereby providing a foundation for the future design of artificial agents and the dynamics of interaction with end-users.

A co-design process involving autistic children can provide a substantial benefit and optimal utilization of technologies to an off-the-shelf design-based one. Having a voice and making a contribution plays a major role in the co-design process. Yet autistic children exhibit varying communication and social skill and some of them may be minimally verbal or non-verbal. For these reasons, harmonizing the techniques of the co-design process with autistic children with varying characteristics requires detailed and careful consideration. To understand the techniques of the co-design process that accommodates all categories of autistic children, a systematic review of a co-design process involving autistic children was conducted, using six large databases (Scopus, ACM Digital Library, ScienceDirect, IEEE Xplore, SpringerLink, and Google Scholar). The search result includes 2482 papers of which only 82 met the inclusion criteria. The result of the data extraction analysis collection is classified according to techniques for accommodating autistic children of varying characteristics, the challenges encountered, and methods of minimizing those challenges. The review identifies four prominent themes within co-design research for autism: advances in co-design objectives and outcomes, participant recruitment determinants, core co-design methods, and the management of co-design challenges. Highlighting the need for inclusivity and equitable support, the study proposes recommendations for better integration of diverse communication abilities and multiple diagnoses in the co-design process, underlining the importance of adaptive technologies and methods to accommodate the needs of all children.

Abstract Detecting the attention of children with autism spectrum disorder (ASD) is of paramount importance for desired learning outcome. Teachers often use subjective methods to assess the attention of children with ASD, and this approach is tedious and inefficient due to disparate attentional behavior in ASD. This study explores the attentional behavior of children with ASD and the control group: typically developing (TD) children, by leveraging machine learning and unobtrusive technologies such as webcams and eye-tracking devices to detect attention objectively. Person-specific and generalized machine models for face-based, gaze-based, and hybrid-based (face and gaze) are proposed in this paper. The performances of these three models were compared, and the gaze-based model outperformed the others. Also, the person-specific model achieves higher predictive power than the generalized model for the ASD group. These findings stress the direction of model design from traditional one-size-fits-all models to personalized models.

In children with autism spectrum disorders (ASD), attention assessment plays a crucial role in understanding their behavioral and cognitive functioning. Difficulties with attention are a common feature of children with autism and have a significant impact on their ability to learn and socialize. In this paper, we propose a non-invasive and objective method to assess attention in children with autism from real videos by utilizing the head poses and motion parameters. The proposed approach is an ensemble of a deep learning model that extracts head pose parameters, an optical flow approach that extracts motion parameters from consecutive frames, temporal head pose parameters extraction and an autoencoder for attention assessment. The experimental study was conducted on 39 children (ASD = 19, neurotypical children = 20) by giving different attention tasks and capturing their video using an attached webcam. Results are analyzed for participant and task differences, which demonstrate that our approach is successful in measuring a child's attention control and inattention. In particular, the assessment of the head poses and motion parameters will enable the development of real-time attention recognition systems that can be used for both learning and targeted intervention.

Money laundering is a process where money from illegal sources is transferred through fraud and concealment to make it appear as legal and legitimate. It is estimated that 2-5% of global GDP is laundered every year. The identification of suspects involved in the laundering network is significantly challenging. This work reports on the utilization of a modified method to identify further money laundering nodes starting with two prior identified suspects. Link analysis was used to transform the weight of the link in a money laundering network starting with two suspects into a distance measured in a new graph representation. The shortest-path algorithm was then utilized to identify the optimal path between two nodes, which cannot be used directly to necessarily identify the strongest association between the pair of nodes. The results show that this method is efficient when large data sets are investigated and when data structure involves minimal attributes to start with.

Machine learning has made rapid advances in safety-critical applications, such as traffic control, finance, and healthcare. With the criticality of decisions they support and the potential consequences of following their recommendations, it also became critical to provide users with explanations to interpret machine learning models in general, and black-box models in particular. However, despite the agreement on explainability as a necessity, there is little evidence on how recent advances in eXplainable Artificial Intelligence literature (XAI) can be applied in collaborative decision-making tasks, i.e., human decision-maker and an AI system working together, to contribute to the process of trust calibration effectively. This research conducts an empirical study to evaluate four XAI classes for their impact on trust calibration. We take clinical decision support systems as a case study and adopt a within-subject design followed by semi-structured interviews. We gave participants clinical scenarios and XAI interfaces as a basis for decision-making and rating tasks. Our study involved 41 medical practitioners who use clinical decision support systems frequently. We found that users perceive the contribution of explanations to trust calibration differently according to the XAI class and to whether XAI interface design fits their job constraints and scope. We revealed additional requirements on how explanations shall be instantiated and designed to help a better trust calibration. Finally, we build on our findings and present guidelines for designing XAI interfaces.

Human-AI collaborative decision-making tools are being increasingly applied in critical domains such as healthcare. However, these tools are often seen as closed and intransparent for human decision-makers. An essential requirement for their success is the ability to provide explanations about themselves that are understandable and meaningful to the users. While explanations generally have positive connotations, studies showed that the assumption behind users interacting and engaging with these explanations could introduce trust calibration errors such as facilitating irrational or less thoughtful agreement or disagreement with the AI recommendation. In this paper, we explore how to help trust calibration through explanation interaction design. Our research method included two main phases. We first conducted a think-aloud study with 16 participants aiming to reveal main trust calibration errors concerning explainability in AI-Human collaborative decision-making tools. Then, we conducted two co-design sessions with eight participants to identify design principles and techniques for explanations that help trust calibration. As a conclusion of our research, we provide five design principles: Design for engagement, challenging habitual actions, attention guidance, friction and support training and learning. Our findings are meant to pave the way towards a more integrated framework for designing explanations with trust calibration as a primary goal.

Promoter regions of long non-coding RNA (lncRNA) genes are crucial to understand their transcriptional regulatory pattern. LncRNA genes, being more cryptic than protein-coding genes in terms of their functionality and biogenesis divergence, are lacking in number of existing studies to elucidate the roles of their promoters compared to their counterparts. Based on the overlap between epigenetic marks and transcription start sites, human lncRNAs were categorized into two broad categories: enhancer-originated lncRNAs (e-lncRNAs) and promoter-originated lncRNAs (p-lncRNAs) and hence these two groups are subject to distinct transcriptional regulatory programs. To understand the difference in the transcriptional regulatory mechanisms that governs p- and e-lncRNAs, we studied the promoter sequences of these two groups of lncRNAs including distinct transcription factor (TF) proteins that favor p-over e-lncRNA (and vice versa). In addition, we developed a convolution neural network (CNN) based deep learning (DL) framework DeePEL (deep p-, e-lncRNA promoter recognizer), to classify the promoter of p- and e-lncRNAs. To the best of our knowledge, this is the first attempt to classify these two groups of lncRNA promoters, using sequence and TF information, based on DL framework. We report several sequence specific signatures in the promoter regions as well as several distinct TFs specific to groups of lncRNAs that will help in understanding the promoter-proximal transcriptional regulation of p-lncRNAs and e-lncRNAs.

The Internet is the main source of information nowadays. Consequently, end users need to be knowledgeable about how to use search engines in order to locate relevant information in a reasonable time with minimal effort. On the other hand, search engines must provide different and alternative ways to represent the search results to facilitate the user access to the information especially for the visually impaired (VI) users. Our research aim is to produce a new representational model for the search engine results targeting VI users. The result of this study will be a functional prototype that summarizes the search results as main ideas that are identified as concepts. Formal Concept Analysis (FCA) defines a concept as the maximum number of objects that are sharing the maximum number of features or attributes. Concepts are discovered by analyzing data patterns for the text of the study. The outcome of the first step of summarization concepts as keywords is used to minimize the number of listed websites and URLs that match the user selection of the multi-level tree of concepts. This scenario of summarization can give the user different directions for the shortest path to reach the target information with the minimum amount of time and effort required. The purpose of these directions can be either to proceed with reading the whole document in detail, or to continue the search for finding other related documents that match the user's inquiry. Experiments run on an iterative testing basis until VI users find proper results that satisfy their needs for the search context. User observations and interpretations based on the experiments are used for the user evaluation. This study will guide us for designing a new model for summarizing search results based on the FCA algorithm to the VI end users, and with a new representation interface based on the discovered concepts' weights.

The objective of the work reported here is to develop an understanding of cross-modal collaborative information-seeking (CCIS) between visually impaired (VI) and sighted users in order to learn how best to support it. In a previous article, we reported the CCIS process that occurred when 14 pairs of users, one sighted and one VI, performed web-based collaborative information-seeking (CIS) tasks in two settings: co-located and distributed. In that study, participants used their tools of choice: web browsers, search engines, notetakers and communication tools. We discussed the difficulties encountered, including those imposed on VI users due to the current limitations of screen readers. In this article, we report a study using the same participants undertaking similar search tasks, but this time using a commercially available CIS system, which we enhanced to improve its accessibility. In this study, in order to examine the impact of the interface on the process, we looked at the CCIS process from two perspectives: the actions of individuals collaborating with one another and the interactions of each individual with each interface. The results showed that both sighted and VI users benefited from the use of an integrated, purpose-built tool, both in terms of task performance and levels of satisfaction. The analysis of these interactions is then used to formulate guidelines for the design of accessible CCIS systems.