ICSBT 2026 Abstracts

Area 1 - AI & Decision Intelligence
Full Papers Short Papers
Area 2 - FinTech, Blockchain & Digital Trust
Short Papers
Area 3 - Smart Business Transformation & Innovation
Full Papers Short Papers
Area 4 - Sustainability & ESG Technologies
Short Papers

Area 1 - AI & Decision Intelligence

Full Papers
Paper Nr: 23
Title:

Machine Learning-Based Prediction of Patient No-Shows in Resource-Constrained Healthcare Settings: Insights from Developing Countries
Authors:

Jannatul Ferdous Nishi, Mohitul Shafir, Md Zahid Hasan, Md. Shah Newaz, Partho Protim Saha and Afroza Sultana
Abstract: Setting up appointments is a key part of using resources better and making healthcare systems work better. But patients still miss a lot of appointments, which is a big problem, especially in low- and middle-income countries (LMICs) like Bangladesh, where missed appointments have a bigger effect because there aren't enough resources. This study introduces a machine learning-based predictive framework for identifying patients likely to miss scheduled appointments by analyzing retrospective healthcare data. Demographic, socioeconomic, behavioral, and environmental variables were employed to develop predictive models, including logistic regression, decision trees, random forests, and gradient boosting. After using SMOTE, the recall for the minority class went from less than 10\% to about 30\% for models based on trees. This demonstrates that imbalance-aware resampling is effective in healthcare datasets with constrained resources. An examination of feature significance revealed that age, appointment lead time, and scheduling patterns were the predominant influences on attendance behavior. The study highlights the necessity for methodological transparency, the utilization of standardized evaluation metrics (accuracy, precision, recall, and $F_1$-score), and the importance of LMIC-specific contextual factors, such as transportation accessibility and environmental conditions.The findings illustrate that automated statistical prediction systems can enhance scheduling efficiency, help with targeted interventions, and optimize resource allocation in healthcare environments characterized by limited resources, provided they are employed alongside appropriate imbalance management methodologies and robust implementation strategies. This study enhances scalable, data-driven strategies to reduce appointment no-shows in healthcare systems within low- and middle-income countries.
Short Papers
Paper Nr: 16
Title:

AI at Work before It Arrives: Task-Level Expectations about Copilot in a University Setting
Authors:

Mostafa Goudarzi, Katarina Cellarova, Mikael Collan and Joanna Kedziora
Abstract: This paper examines task-level expectations and intended use of GenAI tool Mi-crosoft Copilot during the early phase of organizational rollout. We collected data with a customized oTree-based survey among 65 staff members at a Nordic higher education institution in September–December 2025. The respondents were asked to list recurring work tasks and to report their expectations of the effort spent in com-pleting the tasks with and without generative AI tool Copilot, their expectation of how Copilot will impact the output quality, time required, and task satisfaction, and their intended use of the technology under an unlimited-availability scenario. The da-ta indicates that a substantial shift towards a lower effort level is expected with Copi-lot, with pronounced heterogeneity across tasks and application contexts. A higher ex-ante effort-need predicts stronger expected benefits with Copilot. Larger expected effort reductions are associated with more favorable expectations for quality and sat-isfaction, while time use expectations are mixed. Intended use of Copilot is high overall, but varies by primary application, suggesting that tasks embedded in differ-ent application workflows differ in how easily Copilot can be integrated and used. The study contributes with task-level pre-adoption results regarding the expectations about benefits from and use intentions of generative AI. The results serve as com-parative expectations baseline for future research on achieved benefits and can be used to help plan rollouts.
Paper Nr: 39
Title:

Microsoft Recall as a Basis for Process Mining of Computer Usage
Authors:

Heiko Rölke
Abstract: This paper describes an attempt to use screenshots from Microsoft Recall for process mining, especially for the automatic timing of user activities. Microsoft Recall is a novel feature running in the background and enabling the user to search for past activities by means of regular screenshots. Building on this feature, we define a simple workflow for copying and decrypting the screenshot images, decoding their metadata and using those for defining tasks and events. In a small self-experiment, the workflow could be carried out successfully. However, the automatic task labelling needs improvement to be useful for the goal of automatic activity timing. This is the limiting factor of the approach described here and will be tackled in future work.
Paper Nr: 46
Title:

From Raw Climate Data to Decision-Ready Insight: A Lightweight Generative AI Interface for Central Mediterranean Marine Indicators
Authors:

Joel Azzopardi
Abstract: Climate change is reshaping the Mediterranean at a faster rate than the global average, yet the data describing these shifts remain largely inaccessible to non-specialist users. Authoritative sources such as the Copernicus Marine Service expose raw products that require technical expertise to query, while existing visualisation portals present global trends without the regional statistical depth needed for local decision support. This paper presents a lightweight Generative AI platform that bridges this gap for marine climate indicators in the central Mediterranean. The platform combines four components: a RESTful service that returns pre-computed statistics and time series in JSON for sea surface temperature, sea level anomaly, chlorophyll concentration and sea surface pH; a browser-based visualisation interface that renders spatial difference plots and time series from the API output; an explanation layer that produces both plain-language and technical interpretations of the results; and a natural language query interface that maps free-text questions into structured visualisation configurations. The platform is operational at \url{https://climate.ir.mt} and is designed to run on commodity hardware without GPU acceleration. We evaluate the natural language interface against a purpose-built benchmark spanning explicit, relative, named-region and ambiguous query formulations, comparing cloud LLM APIs, locally hosted small models and a hybrid system combining rules and vector embeddings, across both accuracy and resource cost. Results indicate that natural language access to environmental data can be delivered on modest infrastructure without GPU acceleration, with a lightweight rules-and-embeddings hybrid rivalling locally hosted small models on this task and supporting decision intelligence applications in ESG analytics, sustainability reporting and public-sector climate communication.

Area 2 - FinTech, Blockchain & Digital Trust

Short Papers
Paper Nr: 19
Title:

The Multidimensional Role of Trust in Mobile App Adoption and Usage: Qualitative Insights from Australian App Users and Developers
Authors:

Michael J. Bryant, Simon J. Wilde, Golam Sorwar, Vinh Bui and William Smart
Abstract: Location Based Services (LBS) apps as an Information System (IS) have proliferat-ed since the introduction of the smartphone. LBS applications support a wide range of Internet of Things (IoT) cases by enabling real-time data collection and device in-terconnectivity. These applications may also integrate AI technologies to enhance de-cision-making and personalization. However, the increasing reliance on such tech-nologies raises concerns about cybersecurity threats and the potential misuse of user data. Despite the perceived benefits, mobile app usage constitutes ongoing trust is-sues. Furthermore, with the increasing application of Artificial Intelligence (AI) within LBS, such as social media, health, and navigation apps, trust and ethical con-cerns become elevated. To date, trust has been primarily researched through a unidi-mensional lens within the Information Communication Technology app space. How-ever, trust has been widely theorized to encompass multidimensional constructs in many related fields of research. Therefore, this research gap prompted focus group studies with Australian app users and developers to gain insights into multidimen-sional trust and LBS app adoption and usage. Both groups supported the reliability of the trust concepts under discussion. The analysis highlighted the importance of privacy, security, risk taking, transparency, and honesty trust characteristics. Thus, implications for practice are evident for app developers. A holistic trust perspective would give particular attention to transparency and honesty and to risk taking during LBS app development. Therefore, addressing the multidimensional trust elements associated with IoT, as outlined in this study, is essential for guiding future research.

Area 3 - Smart Business Transformation & Innovation

Full Papers
Paper Nr: 26
Title:

A Human-Centered, Multi-Layer Decision-Support Architecture for Remanufacturing with LLM-Based Interaction
Authors:

Corinna Huber, Lukas Rißmann, Patricia Salzinger and Sebastian Meißner
Abstract: Disassembly is a key process in the Circular Economy, enabling the recovery and remanufacturing of components from returned products. Its high variability, uncertain component conditions, and manual task characteristics require decision‑support systems that combine reliable data integration with adaptive and accessible human–machine interaction. This paper presents a human‑centered, multi‑layer decision‑support architecture that integrates real‑time data processing, simulation‑based evaluation of disassembly tasks, and a dialog‑enabled interaction layer based on a retrieval‑augmented large language model (LLM). The architecture embeds human‑in‑the‑loop mechanisms to transform model outputs into comprehensible, context‑sensitive, and multilingual work instructions while maintaining operator agency. A prototypical implementation of the interaction layer was developed and evaluated using multilingual benchmark scenarios derived from industrial disassembly procedures. Results indicate high procedural correctness, low hallucination rates, and reliable clarification and abstention behavior, demonstrating the suitability of the proposed architecture for hu-man‑centered guidance in remanufacturing environments.
Paper Nr: 28
Title:

The Transformation Readiness Index: A Framework for Assessing Regional Automotive Transformation Capacity
Authors:

Anne Häner, Steve Schumann, Bernhard Kölmel, Lukas Waidelich, Rebecca Bulander, Luk Palmen and Johannes Brunner
Abstract: The European automotive industry faces a profound transformation driven by electrification, (vehicle) automation, connectivity, and platform economy. Regional ecosystems are critical in enabling small and medium-sized enterprises to navigate this transition, yet no sector-specific instrument exists to systematically assess regional transformation readiness. This paper introduces the Transformation Readiness Index (TRI), developed using Design Science Research. The TRI integrates firm-level perceptions and regional ecosystem assessments across ten dimensions including risk exposure, transformation pressure, readiness, opportunity recognition, technology-specific regional factors, specialization, and development perspectives, comprising 89 scored indicators administered to automotive companies and business support organizations. It produces a normalized 0–100 readiness score classified into five interpretive levels. A proof-of-concept evaluation with more than 140 organizations across nine Central European regions demonstrates the instrument's feasibility, cross-regional differentiation, and diagnostic value. While serving as initial validation rather than representative assessment, results reveal persistent pressure-readiness gaps and systemic weaknesses in platform economy readiness and policy-side ecosystem support. The study contributes a novel, replicable assessment framework to the literature on regional innovation systems and industrial transformation.
Paper Nr: 31
Title:

Automating Conference Abstract Selection for Impact-Oriented Streams: A Fine-Tuned BERT Approach
Authors:

Nguyen Thanh Ha and Que Nguyet Tran
Abstract: Academic conferences increasingly receive large volumes of submissions, making manual abstract screening time-consuming and subjective. This challenge is especially pronounced for tracks focused on real-world implementation and measurable impact, where reviewers must assess not only methodological quality but also evidence of practical application and operational value. We formulate this problem as a semantic relevance classification task in which relevant abstracts are rare and often signalled by subtle contextual cues rather than explicit keywords. To address this, we propose an automated abstract selection framework based on a fine-tuned BERT classifier. The framework converts raw submissions into machine-readable text and applies preprocessing and supervised labelling. Using the EURO Making an Impact stream as a case study, we compare several text representation methods, including FastText, Word2Vec, GloVe, and fine-tuned BERT. Experiments on EURO22, EURO24, and EURO25 show that fine-tuned BERT substantially outperforms traditional embedding baselines, achieving a stronger precision–recall balance on the minority class and demonstrating practical value as a first-stage filtering tool for conference organisers.
Paper Nr: 32
Title:

Turning Operational Data into Services: Digital Twin-Supported Business Models in (Hybrid-) Electric Aviation
Authors:

Kutay Can Yinanc, Maiara Rosa Cencic and Kai Lindow
Abstract: The transition toward hybrid-electric aviation demands not only technological innovation but also new forms of data collaboration among a complex network of stake-holders. Decentralized data spaces and collaborative digital twins offer promising infrastructure for this collaboration, yet no existing framework that integrates these technical architectures with business model design in a multi-stakeholder hybrid-electric aviation context. This article presents an in-depth descriptive use case study conducted within the publicly funded DIREKT project, examining the development of the value creation network for a battery lifecycle simulation service. A qualitative, user-centred design approach was adopted, combining system and stakeholder mapping, semi-structured expert interviews with eleven professionals in the aviation in-dustry, and the sequential application of the Value Proposition Canvas (VPC), Business Model Canvas (BMC), and Data Cooperation Canvas (DCC). Stakeholder analysis identified fifteen relevant actors, including non-obvious data providers usually overlooked by conventional supply-chain-oriented data ecosystem initiatives, whose integration opens otherwise structurally invisible service opportunities. The resulting business model framework defines distinct but interdependent models for each data provider and the service twin provider, structured around subscription-based SaaS and pay-per-use revenue models. The Data Cooperation Canvas extended this firm-centric analysis to the ecosystem level, revealing that data sovereignty provides the security that makes participation in the data ecosystem rational, and monetization provides the economic incentive. The decentralized data space architecture derived from these requirements is built on the principles of local data storage, machine-readable usage policies, governed data exchange via connectors (such as EDC), and smart contract-based access control, ensuring that stakeholders can collaborate without exposing proprietary data. The findings of this study offer initial guidance for OEMs, airlines, operators, and certification authorities, and establish a foundation for empirical validation through future pilot projects.
Paper Nr: 34
Title:

Enhancing Business Intelligence for Contract Management: Semantic-Aware LLM-Based Analytics for Decision Support
Authors:

Antony Seabra, Daniel Schwabe and Sergio Lifschitz
Abstract: Business Intelligence (BI) for Contract Management requires integrating and interpreting heterogeneous information from structured systems and unstructured contractual documents. Recent advances in Large Language Models (LLMs) have enabled the development of BI systems that combine techniques such as Retrieval-Augmented Generation (RAG), Text-to-SQL, and agent-based orchestration to support complex queries over multiple data sources. However, these approaches remain fundamentally data-driven, relying on retrieval and probabilistic generation without explicit semantic grounding. As a result, they face limitations in handling semantically complex queries involving obligations, rights, and contractual responsibilities, where meaning is distributed and implicitly defined. In this paper, we propose a semantic-aware architecture for LLM-based BI in contract management that introduces a formal Semantic Layer grounded on ontology-enriched knowledge models. We present a Semantic Onboarding methodology that transforms heterogeneous contractual artifacts into structured, executable representations by combining LLM-based extraction, deterministic schema construction, and ontology alignment. These models enable direct querying of contractual semantics through logic-based reasoning, supporting more consistent, complete, and explainable decision-support capabilities. We evaluate the proposed approach against a retrieval-based baseline and show that semantic grounding significantly improves the system's ability to answer semantically grounded questions. In particular, the results demonstrate gains in semantic completeness, interpretability, and consistency when analyzing contractual obligations and enforcement relations. The proposed approach represents a shift from data integration toward knowledge construction, providing a foundation for semantically grounded decision-support systems in contract management and related domains.
Paper Nr: 37
Title:

Success Factors for Process Mining Institutionalization: A CoE Framework Validated by a Digital Transformation Use Case
Authors:

Erik Petsch, Rebecca Bulander, Frank Morelli, S. Vijayakumar Bharathi and Raghav Sandhane
Abstract: Process mining produces fact-based process transparency and supports evidence-based decision-making, yet many organizations do not progress beyond isolated proof-of-concept pilot projects. This paper investigates which organizational factors support the design and scaling of a Process Mining Center of Excellence (PM-CoE). Combining a systematic literature review of 22 publications (2010–2024) with five semi-structured expert interviews, the study consolidates ten critical success factors and organizes them into a conceptual socio-technical framework spanning strategic, structural, operational, and cultural layers, together with a three-stage maturity sequence (pilot, standardization, enterprise scale). Consistent with established work on Centers of Excellence, the interviewed practitioners associate enterprise scaling with a shift from the centralized structures typical of early pilots toward federated hub-and-spoke arrangements, and report that greater process transparency can provoke workforce resistance requiring structured change management. A single telecommunications case is used to illustrate how the factors interrelate in practice. The framework is conceptual, and the supporting evidence is qualitative, drawn from a small, predominantly practitioner-based sample and a single cross-sectional case. The findings are therefore offered as a structured synthesis and a set of propositions for future testing rather than as confirmed results, and they motivate larger, independent, longitudinal, and cross-industry investigation.
Short Papers
Paper Nr: 27
Title:

An Immersive Virtual Reality-Based Simulation Environment for Modular Construction Process
Authors:

Ali Attajer and Mael Delphin-Poulat
Abstract: The increasing adoption of modular and off-site construction methods introduces new challenges related to process coordination, logistics planning, and on-site assembly supervision. In this context, immersive technologies offer promising opportunities to enhance both operational understanding and professional training. This paper presents the design and development of a Virtual Reality (VR)–based simulation environment dedicated to the analysis of modular construction processes. The proposed platform enables users to interactively simulate the end-to-end process of prefabricated building modules, encompassing factory production stages, transportation logistics, site constraints, and crane-assisted assembly operations. The environment integrates physics-based behaviors, parametric module representations, and scenario-driven interactions to reproduce realistic construction conditions, including delays, sequencing conflicts, and unforeseen events. The model is designed as an immersive simulation and learning environment that supports experiential training, procedural analysis, and informed decision-making in Modular Integrated Construction (MiC) projects. An empirical evaluation involving 17 participants demonstrated significant improvements in MiC workflow understanding, logistics comprehension, and scheduling interpretation following VR-based training sessions, with post-training Likert scores exceeding 4.0 across all measured dimensions. The main contribution of this work lies in demonstrating how VR-based simulation can bridge the gap between theoretical construction planning and hands-on operational understanding, while providing a scalable foundation for future extensions toward data-connected and cyber-physical construction systems.
Paper Nr: 30
Title:

AI Policies in Research Administration: An NLP Analysis
Authors:

Aimee Kendall Roundtree
Abstract: This study examines U.S. research universities’ institutional guidance on the use of generative artificial intelligence (AI) in research administration. Drawing on a corpus of 3000 sentences extracted from official policy documents across 95 institutions (2024–2026), the analysis applies a mixed-methods natural language processing (NLP) approach combining supervised classification, topic modeling, coherence analysis, and linguistic inquiry using LIWC. A logistic regression classifier with TF-IDF features categorized policy statements into research administration themes. Latent Dirichlet Allocation (LDA) identified major topics, and coherence scores measured thematic consistency. The results showed that universities mainly focused on Compliance and Governance, Data Privacy, and Security. Institutional language emphasized oversight, accountability, ethics, and data protection. Innovation and AI Enablement language was more positive, confident, and future-oriented than governance-related language. LIWC analysis revealed statistically significant linguistic differences across institutional AI policy categories, particularly in emotional tone, authenticity, analytic thinking, and risk-related language. Overall, universities framed AI as both a strategic opportunity and a technology requiring careful regulation. The findings show how higher education institutions are balancing innovation with compliance and ethical responsibility in research administration.
Paper Nr: 41
Title:

The Impact of Digital Overload and Perceived Digital Surveillance on Teleworkers’ Job Performance: The Mediating Role of Technostress and Moderating Role of Digital Organizational Culture
Authors:

Kurnia Irawati and Muafi Muafi
Abstract: This study examines the relationships between digital overload, perceived digital surveillance, technostress, and job performance among teleworkers in the Special Region of Yogyakarta, while also exploring the mediating role of technostress and the moderating role of digital organizational culture. A quantitative survey was con-ducted with 120 respondents using a structured questionnaire measured on a five-point Likert scale, and the data were analyzed using Partial Least Squares Structural Equation Modeling (PLS-SEM) [1]. The findings indicate that digital overload has a positive and significant effect on technostress, supporting prior studies [2], [3]. However, digital overload does not significantly affect job performance [4]. Per-ceived digital surveillance is also found to have no significant effect on job perfor-mance or technostress, in line with previous research [5], [6]. In addition, tech-nostress does not significantly influence job performance [7], [8]. The study further reveals that digital organizational culture does not significantly moderate the relation-ship between technostress and job performance, and technostress does not mediate the examined relationships [9]. Overall, the results suggest that teleworkers tend to adapt to digital demands and monitoring practices.
Paper Nr: 47
Title:

The Business Model Landscape of Predictive Maintenance Scenarios
Authors:

Matthias Pohl, Christian Haertel, Daniel Staegemann and Klaus Turowski
Abstract: Predictive maintenance (PdM) has emerged as a commercially significant application of the Industrial Internet of Things. The translation of its technical capabilities into viable business models remains fragmented across engineering, information systems, and operations management research. This study presents a structured literature review at the intersection of PdM and business model research, examining how PdM capabilities are commercialized and what technological, economic, and organizational factors shape their design. The analysis synthesizes the reviewed corpus into an integrative framework that characterizes the operational path from data acquisition to value realization and maps the landscape of business model configurations through which PdM is delivered. Enabling technologies, implementation barriers, economic evidence, and sustainability implications are examined as interconnected dimensions of business model viability rather than isolated factors. The review concludes by discussing the configurational and ecosystem-level nature of PdM business models and outlines a research agenda addressing open questions in archetype transitions, value capture, and the gap between sustainability potential and practice.
Paper Nr: 38
Title:

AGI-Driven Policy Making for Net-Zero Cities: Integrating Data Intelligence into Urban Mobility Governance
Authors:

Mohammad Rezaei, Sadashivayya Chandrashekerayy and Omid Fatahi Valilai
Abstract: Urban mobility systems are a major source of carbon emissions in cities and therefore play a crucial role in achieving net-zero targets. While artificial intelligence has increasingly been used to improve transport efficiency, many existing approaches focus on isolated operational problems and remain weakly connected to policy-level decisionmaking. This study addresses this limitation by proposing an AGI-driven decision-support reference architecture for urban mobility governance. In this paper, AGI-driven refers to an architecture designed around AGIlevel capabilities, such as cross-domain reasoning, adaptive learning, scenario generation, and explanatory decision support, rather than to the implementation of a fully realized AGI system. The proposed architecture integrates AGI-supported policy reasoning, reinforcement-learningbased optimisation, simulation-based scenario testing, and key performance indicator evaluation within a unified decision-support structure. In the current version, the AGI-level functions are approximated through coordinated AI components, including large language model-based policy reasoning, reinforcement learning, simulation, and KPI-based assessment. The framework enables consistent comparison between baseline mobility conditions and policy-driven interventions. The architecture is evaluated through a simulation-based proof-of-concept experiment. Five scenarios are tested, including a baseline, public transport shift, peakdemand management, low-emission priority, and a combined policy scenario. Results show that the combined policy with reinforcement-learning optimisation achieves the strongest performance in the synthetic evaluation, reducing simulated CO2 emissions and congestion relative to the baseline. Rather than predicting exact real-world outcomes, the evaluation demonstrates how the proposed architecture can support transparent comparison, trade-off analysis, and informed decision-making for long-term net-zero urban mobility planning.

Area 4 - Sustainability & ESG Technologies

Short Papers
Paper Nr: 29
Title:

Digital Support Services in Vegan Practice Transition: A Quantitative Study of Sustainable Food Practices
Authors:

Thomas Neifer and Kerstin Salewski
Abstract: Digital support services have become increasingly relevant for sustainable food consumption and vegan practice transition. However, there is still limited quantitative evidence on how different forms of digital support vary across transition duration and how they relate to perceived benefits and barriers in everyday life. This study addresses that gap with a survey of 211 participants who identified as vegan or as being in transition to veganism. Drawing on data collected in May 2020, we analyse media use both across duration groups and in relation to perceived benefits and barriers, including moderation by duration. The results suggest that digital support in vegan transition is best understood not as one uniform media effect, but as a differentiated support ecology. Some media types are more strongly associated with early orientation and uncertainty, whereas others remain relevant for the ongoing organisation of everyday feasibility, including shopping, cooking, eating out, and nutritional confidence. In addition, community-facing media appear less as straightforward barrier-removal tools than as environments of social learning, interpretive support, and legitimacy. Rather than proposing a new framework for vegan transition, the study provides quantitative evidence that substantiates prior qualitative and practice-theoretical work. The findings further suggest that vegan-support services are better understood as practical support infrastructures for sustainable consumption than as generic persuasion tools.