University of Liechtenstein

Innovation is a crucial driver of competitiveness, yet traditional metrics fail to capture its full scope. Therefore, holistic approaches encompassing a wider variety of innovation types are needed. We conceptualize perceived innovativeness as an externally attributed judgment formed under information asymmetry, where financial analysts act as informed intermediaries translating firms’ disclosures into credible innovation signals. Following this concept, this paper extends an existing LDA method by introducing a supervised approach (sLDA), trained with Research & Development (R&D) intensity. This variation addresses the gap between the dynamic, multifaceted nature of innovation and the limitations of prevailing metrics. We analyzed 1606 analyst reports from 499 European companies to test our approach. Our findings show that, compared to unsupervised approaches, the sLDA approach yields more accurate results in modelling perceived innovativeness. Furthermore, we demonstrate robust results over the period of 2019–2022: the supervised model achieves consistently lower mean absolute rank errors compared to the unsupervised baseline and, in most industries, than R&D-based benchmarks. We also identify influential topics and find fixed-effects-robust associations with patent indicators, with rank-based evaluation reducing sensitivity to outliers. This study contributes to the ongoing discourse on innovation assessment and benchmarking by linking language-based perception with signaling theory and offering a method that captures perceived innovation as expressed in analyst language. It provides managers with a tool to support strategic planning, investors with an additional signal of perceived innovation and regulators with insights into innovation narratives complementing R&D and patent disclosures. • Presents a supervised LDA (sLDA) approach to quantify firm perceived innovativeness. • Uses firm R&D intensity to guide topic modeling in analyst reports. • Captures innovation signals overlooked by traditional metrics. • Validates sLDA scores against R&D intensity and patent-based innovation metrics. • Provides a benchmarking tool for managers and researchers.

Zusammenfassung Der europäische Finanzsektor befindet sich in einer tiefgreifenden Transformation, die durch Digitalisierung, steigenden Wettbewerbsdruck und wachsende regulatorische Anforderungen getrieben wird. Angesichts dieser Komplexität entwickelt sich organisationale Resilienz, die Fähigkeit, schnell, stabil und sicher auf unerwartete Ereignisse zu reagieren, zu einem strategischen Erfolgsfaktor. Traditionelle, langwierige Prozesse in Banken werden dabei durch veraltete Strukturen und funktionale Silos ausgebremst. Die vorliegende qualitative Fallstudie untersucht, wie Continuous Integration/Continuous Delivery (CI/CD) und Development, Security und Operations (DevSecOps) zur agil-digitalen Transformation und zur Stärkung der organisationalen Resilienz in einer europäischen Bank beitragen. Die Analyse zeigt, dass die Einführung dieser Praktiken Silostrukturen auflöst, indem cross-funktionale Teams gebildet werden. Der „Shift-Left“-Ansatz integriert Sicherheits- und Compliance-Prüfungen frühzeitig in den Entwicklungsprozess, was Risiken proaktiv minimiert und die regulatorische Konformität effizienter erfüllt. Die Studie leistet damit einen Beitrag, sowohl für Wissenschaft als auch für Praxis. Sie zeigt empirisch, wie technologische, organisatorische und kulturelle Mechanismen zusammenwirken, um Transformation und Resilienz zu stärken, und identifiziert zentrale Erfolgsfaktoren wie cross-funktionale Zusammenarbeit, frühzeitige Integration von Sicherheitsprüfungen und lernförderliche Kultur. Für Unternehmen lassen sich daraus konkrete Handlungsempfehlungen ableiten, etwa zur Gestaltung von Teams, Prozessen und Führung, um agil-digitale Transformationen und organisationale Resilienz effektiv zu fördern.

We analyze 52 Bitcoin hard forks using minute-level local projections to assess their impact on Bitcoin’s returns, volume, and volatility following the planned split. Trading volume shows significant effects, while returns and volatility remain unaffected. Our findings suggest that governance splits trigger immediate trading activity, while price and volatility consequences likely emerge over longer horizons. These results remain robust to the omission of major forks. This study provides insights for future debates on controversial protocol changes, such as those concerning quantum resistance and efforts to make Bitcoin more energy efficient. • Insights prepare investors for future blockchain governance issues. • We provide insights into Bitcoin’s microstructure dynamics following its forks. • We estimate impulse response functions using local projections. • A statistically significant effect was found for volume.

To obtain insights from event data, advanced process mining methods assess the similarity of activities to incorporate their semantic relations into the analysis. Here, distributional similarity that captures similarity from activity co-occurrences is commonly employed. However, existing work for distributional similarity in process mining adopt neural network-based approaches as developed for natural language processing, e.g., word2vec and autoencoders. While these approaches have been shown to be effective, their downsides are high computational costs and limited interpretability of the learned representations. In this work, we argue for simplicity in the modeling of distributional similarity of activities. We introduce count-based embeddings that avoid a complex training process and offer a direct interpretable representation. To underpin our call for simple embeddings, we contribute a comprehensive benchmarking framework, which includes means to assess the intrinsic quality of embeddings, their performance in downstream applications, and their computational efficiency. In experiments that compare against the state of the art, we demonstrate that count-based embeddings provide a highly effective and efficient basis for distributional similarity between activities in event data.

Abstract Estimating porosity distribution is essential for hydrocarbon exploration, carbon capture and storage, and geothermal energy. Traditional workflows, which rely on seismic inversion and rock physics equations, are time-intensive, parameter-dependent, and sensitive to variations in mineralogy and fluid composition. On the other hand, workflows based on machine learning algorithms in the state of the art are also limited by their reliance on specific rock physics equations that may oversimplify geologic complexity. In addition, there are assumptions of data availability in the time domain and requirements for multifrequency seismic inputs, which are not commonly accessible. This study proposes a workflow that uses 1D convolutional neural networks to directly estimate porosity from poststack seismic data, addressing the aforementioned traditional limitations. The workflow implements a custom loss function during training that incorporates porosity derived from well logs. Our proposed workflow outperformed alternative machine learning approaches for porosity estimation from seismic data in both the time and depth domains, using synthetic and field data. Experiments on synthetic data demonstrated its superior predictive capability compared to conventional and state-of-the-art methods. Moreover, evaluations on field data achieved a determination coefficient (R2) of 0.860 on the test set, an outcome reinforced by five-fold cross-validation. This performance shows the network's robust generalization, resilience to data variability, and stability under moderate noise, yielding interpretable results. The network was trained efficiently in less time by requiring fewer trainable parameters, significantly reducing the time required compared to conventional and state-of-the-art workflows.

Background: Maxillary sinus floor augmentation (MSFA) is commonly used to increase posterior maxillary bone volume prior to implant placement. Although generally successful, late complications can impact long-term outcomes. The purpose of the study was to estimate the incidence and timing of atypical late complications following (MSFA) using bovine xenohybrid bone grafts. The study also aimed to evaluate whether preoperative bone volume is associated with the risk of complications. Methods: This retrospective cohort study was conducted at the Center of Oral and Maxillofacial Surgery, Danube Private University, Krems-Stein, Austria, and included patients who underwent MSFA with bovine xenohybrid bone grafts and either simultaneous or staged implant placement between January 2020 and December 2023. Preoperative bone volume of the posterior maxilla measured via cone beam computed tomography (CBCT) in the planned implant insertion position. The primary endpoint was the time (days) from MSFA to the occurrence of a graft-related complication (defined as atypical if occurring more than 6 months after MSFA and not related to peri-implantitis) The covariates included subjects’ age, sex, the quantity of graft used for MSFA, timing of dental implant insertion (simultaneous vs. staged) and implant dimensions. Kaplan–Meier analysis and Cox proportional hazards regression were used to evaluate time-to-event data. Only one graft site per patient was analyzed. Results: Atypical complications occurred in 9 out of 47 patients (19.1%), with an average time to onset of 645 days. In a multivariable analysis, a lower preoperative bone volume was found to be an independent predictor of an increased risk of complications (hazard ratio [HR]: 0.972; 95% confidence interval [CI]: 0.925–1.021; p = 0.252). However, the quantity of graft used for MSFA was not found to be a predictor (p = 0.46). Conclusions: Within the limitations of a retrospective study, reduced native bone volume appears to increase the risk of atypical late complications following MSFA with bovine xenohybrid grafts. This makes closer clinical and radiologic follow-up of patients over a longer period very necessary.

Cyber threats have become increasingly prevalent and sophisticated. Prior work has extracted actionable cyber threat intelligence (CTI), such as indicators of compromise, tactics, techniques, and procedures (TTPs), or threat feeds from various sources: open source data (e.g., social networks), internal intelligence (e.g., log data), and “first-hand” communications from cybercriminals (e.g., underground forums, chats, darknet websites). However, “first-hand” data sources remain underutilized because it is difficult to access or scrape their data. In this work, we analyze (i) 6.6 million posts, (ii) 3.4 million messages, and (iii) 120,000 darknet websites. We combine NLP tools to address several challenges in analyzing such data. First, even on dedicated platforms, only some content is CTI-relevant, requiring effective filtering. Second, “first-hand” data can be CTI-relevant from a technical or strategic viewpoint. We demonstrate how to organize content along this distinction. Third, we describe the topics discussed and how “first-hand” data sources differ from each other. According to our filtering, 20% of our sample is CTI-relevant. Most of the CTI-relevant data focuses on strategic rather than technical discussions. Credit card-related crime is the most prevalent topic on darknet websites. On underground forums and chat channels, account and subscription selling is discussed most. Topic diversity is higher on underground forums and chat channels than on darknet websites. Our analyses suggest that different platforms may be used for activities with varying complexity and risks for criminals.

Abstract This study examines how humans interact with large language models (LLMs) in real-world, unconstrained settings, focusing on potential shifts in users’ mental models. Initially, many users approach LLMs as traditional software tools, employing structured, machine-like prompts. However, after their first interaction, a notable shift occurs, such as increased politeness, more natural language phrasing, and shorter, more contextually nuanced prompts. That is, users increasingly adopt conversational behaviors typical of human-to-human communication and, in turn, this suggests a cognitive transition in the way users perceive and engage with AI systems. Analyzing over 200,000 conversations with computational linguistics methods, we find initial indications supporting this change. These insights have implications for AI design, trust, and ethical concerns, highlighting the need for further research beyond a mostly computational perspective to strengthen our findings on how users cognitively frame their interactions with AI.

This study examines the development of "liquid organisations" within the context of work in the metaverses, focusing on the ways spatial and virtual reality technologies profoundly reorganise the topology of the organisation and supersede traditional hierarchical frameworks.Using the concepts of spatial sociology and organisational theory, we construct a transformation framework that outlines significant aspects of change which include reconfiguration of space, mechanisms of visibility, dynamic arrangements, and fluid representation.Our analysis shows how metaverse environments eliminate physical proximity as a determining factor of organisational domination and influence, create reciprocal visibility that defies traditional surveillance backlash, permit agile team formation through topological responsiveness, and allow disruption of status representations through plastic representation.All these changes collectively constitute dissolving stagnant bureaucratic frameworks for more flexible ones.These changes are profoundly impactful for fundamental concerns of coordination and control in post-physical environments and identity situational to the organisation.This work enhances the understanding of the spatial redesign in metaverse workspaces as an enabling factor to structurally reimagine organisational design while simultaneously posing grave hardships to governance, equity, and inclusion within virtually performed enacted organisational structures.