AHFE Open AccessArtificial Intelligence and Social Computing
Editors: Tareq Z. Ahram, Jay Kalra, Waldemar Karwowski
Topics: Artificial Intelligence & Computing
ISBN: 978-1-964867-79-3
DOI: 10.54941/ahfe1007222
Table of Contents
Values-Driven AI Framework for Preschool and Elderly Learning in Bulgaria
The accelerated integration of artificial intelligence (AI) into educational settings has intensified debates concerning pedagogical quality, ethical responsibility, and alignment with human values. In Bulgaria, preschool education and elderly learning environments represent socially sensitive domains in which AI technologies may offer benefits such as adaptive learning, improved accessibility, and reduced professional workload, while simultaneously posing risks related to inequity, value erosion, and diminished human oversight. Situated within the context of societal turbulence and shifting value structures examined by COST Action CA24150 Values in Turbulent Times (VISTA), this paper responds to the need for analytical tools capable of evaluating AI’s educational implications without requiring immediate empirical deployment. The study develops a conceptually grounded framework for assessing AI-supported learning environments across early childhood and older-adult education. Drawing on media pedagogy, human–computer interaction, and value-sensitive design, the framework articulates three analytical dimensions: (1) Pedagogical Value Alignment, addressing developmental appropriateness, learner autonomy, curiosity, and educator–learner interaction; (2) Environmental and Relational Value Dynamics, encompassing inclusivity, emotional climate, accessibility, and social cohesion; and (3) Ethical-Systemic Value Integration, focusing on transparency, data protection, fairness, human oversight, and sociocultural contextualization within Bulgarian educational structures. These dimensions synthesize insights from European regulatory frameworks, national policy documents, and prior research on AI literacy, education governance, and generational differences in technology adoption. Rather than measuring outcomes, the framework provides a structured foundation for future experimental, observational, and comparative studies. By embedding value analysis into early-stage AI evaluation, the model supports human-centered technological integration and contributes to VISTA’s broader inquiry into how socio-technological turbulence reshapes educational values.
Lyubomir Kolarov
Open Access
Article
Conference Proceedings
History and Historians in the Age of AI
Artificial Intelligence and its tools have sparked numerous debates within the Humanities. In this study, I aim to explore how AI tools contribute to History, while also introducing new responsibilities for historians. I also intend to analyse how AI, in conjunction with digitization, may help address the challenges of the zeitgeist, even as it pushes historians toward specific subfields. On the one hand, AI tools enable historians to work with diverse datasets and analyse large volumes of material—from digitized archives to textual data derived from oral history studies. Additionally, AI-based language tools allow historians to engage with sources in multiple languages, offering a level of accessibility previously unattainable. In an increasingly polarized and fragmented world, where scholars face challenges such as visa restrictions, AI and digitization ease many practical difficulties. On the other hand, these advancements place a new kind of burden on historians: the need to remain vigilant about the potential transformation of the discipline. Most notably, as AI tools introduce algorithmic interpretation and rely on “big data,” they risk amplifying existing biases, subjectivities, or overgeneralizations within historical sources. Therefore, I argue that scholars must adopt a new critical approach to ensure human oversight remains central to historical scholarship. In this study, I aim to discuss the advantages and drawbacks of AI in the discipline of history, ultimately suggesting that AI will not herald the downfall of the Humanities or its scholars. However, it will undoubtedly require the development of new methodologies and approaches to navigate its impact effectively.
Hazal Papuççular
Open Access
Article
Conference Proceedings
Deliver cross-process automation across Finance, HR, Procurement by orchestrating actions across diverse systems -powered by AI & governed workflows
The rapid diffusion of data‑driven automation and agentic AI systems is reshaping the foundations of work, decision‑making, and human–technology interaction. As organizations move toward Society 5.0— Japan’s vision for a human-centered “super smart” society in which cyber-physical intelligence augments human capability across economic and social systems—there is an urgent need for operational architectures that are not only technologically capable but also fundamentally human‑centric. This paper presents an applied model using Intelligent Operations framework that integrates agentic AI, enterprise data fabric, human‑in‑the‑loop governance, and secure multi‑system orchestration, and enterprise digital twins that simulate processes and operational states for context-aware decision support. The result is an adaptive socio‑technical system that enhances human decision‑making rather than replacing it, while simultaneously enabling automation at operational scale.The research builds on fieldwork across finance, supply chain, HR, and complex asset‑intensive environments, where organizational processes are distributed across heterogeneous platforms such as ERP, HCM, workflow systems, enterprise data lakes, RPA tools, and emerging AI orchestration layers. Traditional human‑computer interaction models are insufficient in these environments because workers face fragmented data landscapes, inconsistent process execution, and increasing cognitive load. The proposed Intelligent Operations framework addresses these pain points by introducing an orchestration layer that harmonizes data, interprets context (including real-time insights from digital twin models), and deploys agentic AI workers capable of completing multi‑step tasks across systems.A key contribution of this work is the definition of agentic AI in enterprise socio‑technical ecosystems—AI agents equipped not only with language models and planning capability but also with secure access to enterprise systems through structured patterns such as passthrough APIs, workflow orchestration, Model Context Protocol (MCP), and agent‑to‑agent (A2A) collaboration. Rather than relying on brittle rule‑based workflows, the agents dynamically interpret goals, assess context, and plan actionable sequences while maintaining traceability, decision lineage, and auditability. This supports a new form of “digital labor” that works alongside human employees to augment cognitive, administrative, and analytical tasks. However, the framework insists on human‑in‑the‑loop governance, recognizing that human oversight remains essential for ethical, safe, and responsible AI deployment. The DMO acts as a security and compliance boundary—enforcing identity controls, audit trails, approval checkpoints, policy enforcement, and anomaly detection throughout the agentic automation lifecycle. This hybrid model ensures that automation amplifies human capability without bypassing institutional safeguards or creating new forms of risk.The paper also discusses the human‑centric business implications: reduced cognitive load for knowledge workers, increased transparency of decision processes, improvements in cross‑functional collaboration, and the redefinition of roles as humans transition from transactional executors to supervisors, interpreters, and strategic actors. Proposed framework becomes the backbone for Society 5.0 organizational design—linking people, processes, data, and intelligent systems through a unified operational fabric.This research demonstrates that when designed with ergonomics, human values, and socio‑technical principles at the center, agentic AI become powerful enablers of human‑centric, resilient, and adaptive enterprises.
Elizabeth Koumpan, Laurentiu Gabriel Ghergu, Łukasz Strack, Grzeg Jurek
Open Access
Article
Conference Proceedings
AI With and Within User Research Across the Product Lifecycle
Artificial intelligence (AI) is rapidly transforming user research workflows, yet its integration often lacks grounding in human factors principles. While generative and agentic AI can accelerate synthesis, pattern detection, and analysis, overreliance on automated outputs risks automation bias, misplaced trust, and erosion of research judgment. This paper presents a human factors framework for integrating AI with and within user research across the product lifecycle. AI is positioned as an augmentation tool that reduces analytical friction and supports sensemaking, while researchers remain accountable for interpretation and decision-making. The framework maps AI-assisted opportunities across discovery, solution development, delivery, and post-launch support. Explicit boundaries for AI use are defined to prevent misuse. By grounding AI in human judgment, this work offers practitioners practical guidance for designing research workflows that preserve rigor, accountability, and human-centered decision-making in increasingly intelligent sociotechnical systems.
Maria Natalia Russi-Vigoya, Leonidas Guadalupe
Open Access
Article
Conference Proceedings
Governing the Transition to Action: An Agentic Architecture for Situation-Aware LLMs
This paper proposes a framework of ten architectural requirements for AI decision support in safety-critical domains. Derived from Endsley’s SA model and the distributed SA perspective, the requirements span real-time perception, structured comprehension, projection, temporal depth, transparency, operator state modelling, auditability, governed activation, inter-agent coherence, and self-monitoring. We evaluate three classes of LLM-based architecture against these requirements and demonstrate that only agentic workflows with knowledge graph grounding can satisfy them. A technical architecture for a simulated air traffic control environment demonstrates how each requirement maps to concrete infrastructure components, and identifies which requirements are readily met and which remain open research challenges.
Michael Hildebrandt
Open Access
Article
Conference Proceedings
Multi-source Food Names Mapping Using OpenAI vision, Manual Dictionary and Fuzzy Matching Techniques
Accurate harmonization of food names across heterogeneous and multilingual datasets remains a major challenge in food informatics, dietary assessment systems, and data-driven public health research. Modern AI-based food recognition models such as LogMeal, FoodSAM, and OpenAI Vision can identify multiple components within complex dishes, but they frequently produce inconsistent, culturally specific, and multilingual labels. These inconsistencies complicate downstream tasks including nutritional analysis and cross-dataset integration. In this study, we evaluated practical methods for mapping AI-generated food component names to Finnish menu-based ground truth in a real-world restaurant setting. We collected 320 meal images using an integrated camera–scale system; 167 images containing multi-component dishes were selected for detailed evaluation against Finnish lunch-line menu labels. We compared (i) a segment-aware, menu-constrained mapping approach that uses LogMeal segmentations and prompts OpenAI Vision to select the best-matching item from the daily menu for each segment, and (ii) a hybrid manually curated canonical dictionary and fuzzy string matching pipeline applied separately to labels from different AI sources. Mapping performance is measured using precision, recall, and F1-score. The segment-aware OpenAI Vision approach achieved the best overall results (Precision = 0.90, Recall = 0.70, F1 = 0.79), while the hybrid dictionary+fuzzy method also improved consistency over direct label matching. These results indicate that menu-aware segment-level reasoning and lightweight lexical normalization are effective for food-name harmonization and can support scalable dietary monitoring and menu analytics.
Shyam Bhetuwal, Lauri Koivunen, Rehan Khalil, Sanna Koskimäki, Hanna Lähde, Veera Houttu, Kirsi Laitinen, Tuomas Mäkilä
Open Access
Article
Conference Proceedings
Product Design with Human-Machine Collaboration and AI Integration into Design Process
In order to enhance product design efficiency, this paper integrates AI technology into the design process by combining AI software with traditional workflows. Three design methodologies—text-to-image generation, image-to-image generation, and sketch-to-process generation—were explored, along with human-machine collaborative design approaches. Practical implementation demonstrated that this collaborative workflow significantly improves the effectiveness of creative solution generation, providing valuable insights for design professionals.
Huajie Wang, Rong Chen, Jinwu Xiang, Zhang Jin
Open Access
Article
Conference Proceedings
Trust and Calibration in AI-mediated Decision Support under Conditions of Risk
AI-mediated decision support systems are increasingly deployed in domains characterized by risk, uncertainty, and time pressure. In such environments, appropriate reliance on AI recommendations requires not only initial trust formation but also dynamic recalibration when system performance fluctuates or conflicts with other information sources. Although determinants of perceived trust (e.g., explainability, authority cues, and ethical framing) have been widely studied, less attention has been given to how reliance behavior adjusts following observed system error. This paper presents a focused qualitative synthesis of empirical studies examining trust and reliance in AI-based decision support under conditions of risk or informational divergence. Across the included studies, trust was frequently operationalized as an attitudinal construct or predictor of adoption. In contrast, fewer investigations directly measured behavioral reliance following performance degradation or assessed calibration accuracy, defined as the alignment between perceived system capability and actual performance over time. Findings suggest that reductions in reported trust do not consistently translate into commensurate changes in reliance behavior. This divergence highlights the need to distinguish attitudinal trust from behavioral calibration when evaluating AI systems in safety-relevant contexts. We argue that calibration-aligned design (rather than trust maximization alone) should guide the development and assessment of high-stakes AI decision support.
Angela Fike, Tian Wang, Masooda Bashir
Open Access
Article
Conference Proceedings
Identification of Influential Nodes and Discourse Features within Synthetical Hierarchical Communities in Online Social Networks
Social fragmentation and information gap are leading to a growing number of communication barriers and social issues. To analyse root causes of the above phenomena from the view of interactive influence, we develop two models to identify influential nodes and discourse features in online social networks to reveal their influence on information dissemination in synthetical hierarchical communities. Firstly, a Node Influence Calculation Model is constructed based on network topology by integrating multi-dimensional indicators such as degree centrality, closeness centrality, and betweenness centrality, to evaluate the influence of nodes with the holistic information in online social network. Secondly, a Discourse Features Model is built based on semantic information by incorporating topics and sentimental features with fine-grained semantic analysis to decode the strategic adjustments and sentimental polarization mechanisms of influential nodes. Finally, with empirical research in real networks, the above two models effectively calculate the influence of nodes and reveal their roles in reinforcing communication effects, cross-community connections, and fostering community integration through themes and emotions. The findings can provide theoretical basis and guiding strategies to promote balanced information dissemination for online social network management, public opinion guidance and societal cohesion
Yufan Wu, Zhouhai Chen, Peihan Wen
Open Access
Article
Conference Proceedings
Review of Supervised Machine learning cost estimation techniques for building projects
Cost is considered a vital parameter in determining the success of a construction project. Project costs control and monitoring prevent budget overruns and safeguard expected profits, regardless of the project's size, scope, or complexity. Traditional methods for estimating project costs are facing growing challenges as demand for more accurate, adaptable strategies that respond to evolving market dynamics and technological progress increases. This study offers insight into supervised ML-based cost estimation techniques, highlighting the models employed, the geographical area of the studies, sample sizes, input and output variables, and property types. The findings indicate that there has been some progress in applying supervised ML for cost estimation. Asia accounts for the most studies (65.96%), followed by Africa (10.64%) and Europe (14.89%). Oceania and North America each account for 4.26%, indicating a restricted research scope in these areas. Additionally, 62% of the studies employed multiple algorithms to enhance the reliability of the results. Moreover, most studies focused on construction costs rather than total project costs or total capital investment (project investment) and on residential and educational property types. The findings suggest that extensive testing and applications are necessary to gain a comprehensive understanding of global perspectives, particularly outside Asia, and in commercial properties such as retail and office buildings.
Faith Dowelani
Open Access
Article
Conference Proceedings
A social media site for social well-being? The curious case of BeReal
Authentic self-presentation on social media has been identified as a potential protective factor for users’ wellbeing, yet empirical research has largely focused on highly curated platforms such as Instagram and TikTok. The present study is among the first to systematically investigate BeReal, a platform explicitly designed to promote authenticity through the daily posting of spontaneous and unedited images. The study examines how BeReal use and posting behaviours are associated with emotional intelligence (EI) and psychological wellbeing. Twenty-eight BeReal users aged 18 to 25 completed validated measures of online self-presentation, wellbeing, and EI, and provided app usage data. Participants’ daily BeReal posts were also collected over a 30-day period and categorised using an objective coding scheme capturing image type, location, user activity and appearance, posting lateness, and image retakes. BeReal use was substantially lower than use of other social media platforms. Although participants posted on most days, the majority of posts were late and image retakes were common, suggesting continued selective self-presentation. Posting behaviour showed no association with wellbeing or EI in relation to lateness, retakes or image content. Overall, findings suggest that even on authenticity-focused platforms, selective self-presentation persists, and that platform design alone may be insufficient to support psychological wellbeing.
Mark Turner, Heather Balding
Open Access
Article
Conference Proceedings
Methods and Tools for Optimizing to Avoid Similarity in Graphic Design Content Generated by Artificial Intelligence
With the widespread application of generative artificial intelligence in the field of graphic design, the high production efficiency it brings is accompanied by significant risks of content homogenization and stylistic convergence. The fundamental nature of AI models, trained on large-scale datasets, often results in similarities in the composition, elements, and style of their output, severely constraining the originality and commercial distinctiveness of designs. This research focuses on the core challenge of avoiding similarity in AI-generated graphic design content, aiming to develop a systematic solution encompassing both generation methods and tool optimization.The study first provides an in-depth analysis of the technical root causes of similarity, including training data bias, the limitations of prompts, inherent model patterns, and the convergence of generation parameters. Building on this foundation, the paper proposes and explores multi-dimensional avoidance methods. At the technical level, it advocates for a "hybrid generation" strategy that combines different modal models (e.g., image-text, video generation models) for cross-inspiration and re-creation. It also introduces mechanisms for controllable randomness and noise injection to disrupt the model's inherent output patterns, alongside developing deep style transfer and element recombination algorithms based on semantic understanding. At the management and strategic level, it emphasizes building high-quality, diverse, and domain-specific refined datasets and promotes an iterative "human-machine collaboration" workflow. This workflow deeply integrates designers' aesthetic judgment and creative intervention at critical nodes within the generation chain.To effectively implement these methods, the research further outlines pathways for tool optimization. The focus is on developing intelligent "de-homogenization" plugins and platform features. These would integrate modules for multi-model comparative generation, style entropy analysis, and element deconstruction and fusion suggestions, providing designers with real-time similarity assessment and decision support for differentiation adjustments. The goal of tool optimization is to simplify the complex technical process of avoiding similarity into a visual, controllable, and user-friendly design assistant. This aims to ensure and stimulate the uniqueness and innovativeness of designs while enhancing efficiency. This study offers theoretical and practical references for the evolution of AI-assisted design from a mere "production tool" to a "creative partner," contributing positively to the healthy and diversified development of the design industry.Keywords: AI-generated design; content similarity; de-homogenization; method avoidance; tool optimization; human-machine collaboration
Cai Yanhao, Zhao Yulin, Chen Jinpeng
Open Access
Article
Conference Proceedings
Understanding Individual Differences in Adolescents’ Emotional Responses to Social Media Through Human-Centered Causal and Dynamic Modeling
Social media platforms shape adolescents’ daily social experiences by making peer feedback, visibility and social comparison highly salient. Yet research on social media and adolescent well-being remains inconsistent, partly because many studies rely on broad between-person measures that obscure within-person dynamics and developmental differences. This paper proposes a human-centered computational framework using Ecological Momentary Assessment (EMA) to examine how adolescents’ emotional responses to online validation vary within individuals and across developmental stages. Focusing on early adolescents (ages 13–15) and later adolescents (ages 16–20), the framework analyzes longitudinal EMA data on online validation events, emotional state and contextual factors using a three-part modeling pipeline: (1) multilevel within-person models to estimate immediate changes relative to individual baselines, (2) Dynamic Structural Equation Modeling (DSEM) to capture temporal dependencies and carryover effects across prompts and (3) causal forest modeling to estimate heterogeneous effects and identify profiles of sensitivity to validation. As an initial methodological step, the framework is validated using a synthetic dataset designed to reflect realistic EMA-style patterns of behavior, missingness and emotional dynamics before real-world deployment. The proposed pipeline provides a reusable and developmentally sensitive approach for studying adolescent digital well-being with greater precision than coarse between-person measures of social media use.
Anzara Ausaf, Kazi Ruslan Rahman, Sanchita Ghose
Open Access
Article
Conference Proceedings
Biomechanical plausibility of generative AI models: A validation methodology for studying cultural motor accents.
The rapid advancement of AI video generation models presents new opportunities to reduce financial and logistical barriers in biomechanical research in diverse cultural contexts. However, the capacity of these tools to generate movement with physical fidelity remains insufficiently validated. This study evaluated the performance of Google VEO 3.1 Fast in representing a culturally defined motor task: the Japanese seiza sitting posture. Kinematic measures of actual human performance were compared with AI-generated sequences using OpenPose for pose estimation purposes. AI sequences were generated from reference frames extracted from the original performance, including virtual camera rotations produced using Qwen AI Image Edit to enable novel view synthesis. All kinematic trajectories were temporally normalized before analysis. The results indicated that while the model achieved high temporal correspondence for knee flexion (r = 0.936), it introduced substantial discrepancies in postural control. In particular, AI-generated movements exhibited a tendency toward postural regularization, reducing trunk flexion toward a more idealized vertical orientation, with mean absolute errors of approximately 10° relative to the original performance. Furthermore, subtle motor variations, such as eccentric control phases and micro-pauses, were systematically smoothed. These findings indicate that current generative video models are not yet capable of faithful biomechanical reconstruction, as they homogenize motor execution and compromise the preservation of cultural motor accents essential for rigorous quantitative analysis.
Louis Poague, Alexandre Anibal Campos Bonilla
Open Access
Article
Conference Proceedings
Evaluating Public Art in Commercial Complexes: A Dual-Channel Emotion Recognition Framework Fusing Facial Micro-Expressions and Semantic Analysis
The development of public art in commercial spaces increasingly emphasizes emotional experience, but traditional evaluation methods struggle to objectively capture the real-time dynamics and inherent complexity of user emotions. This study proposes a dual-channel emotion recognition framework integrating facial micro-expression analysis and semantic understanding. We define seven emotional categories and construct a dataset of facial expressions and speech data. Feature-level fusion between DeepFace-processed facial data and DeepSeek-R1-analyzed semantic data generates a unified recognition model. An empirical study analyzing five types of public art media with 2,100 speech transcripts and 1,200 facial images reveals that dual-channel fusion achieves 93.2% accuracy, significantly outperforming single-modal approaches. Interactive art generates the strongest emotional stimulation, platform-based art enables efficient communication through social attributes, and functional art provides unique emotional buffering. This research offers a quantifiable methodology for commercial public art evaluation.
Jianing Hu, Xiayun He, Lianmei Dong
Open Access
Article
Conference Proceedings
A Unified Taxonomy of Deep Learning Optimizers for Scalable and Efficient AI Systems
The rapid advancement of artificial intelligence (AI), particularly large language models (LLMs), has created a significant socio-technical divide. The immense computational resources required for AI training increasingly limit participation to a few well-funded entities, hindering the democratization of AI research and raising concerns about environmental sustainability. While optimization algorithms are critical to reducing these resource barriers, the current landscape is highly fragmented, offering limited practical guidance for practitioners in resource-constrained environments. To address this accessibility gap, we present a unified taxonomy of deep learning optimizers that systematically organizes methods by their order of information: zeroth, first, and second order, while integrating emerging, IO-aware and Flash attention paradigms. Instead of merely enumerating algorithms, our approach emphasizes cost-efficiency, memory usage, and hardware constraints as pivotal factors for equitable AI development. Our synthesis of the literature reveals that system-level considerations, particularly IO efficiency, are essential not just for computational performance, but for making large-scale AI accessible. We introduce a decision-oriented framework that translates theoretical insights into practical guidelines, establishing a structured foundation for broader communities to train and deploy human-centered AI systems sustainably and efficiently.
Carlos Villarreal, Jonathan Luzuriaga, Emilio Quinga, Nicolas Reinoso, Bryan Morales, Diana Martinez-Mosquera
Open Access
Article
Conference Proceedings