Aim/Purpose Generative artificial intelligence (GAI) tools based on Large Language Models (LLMs) are increasingly used to enrich knowledge, support decision-making, and generate new content. However, as beneficial as these tools may be, they also store vast amounts of information, which might jeopardize users’ privacy and data protection. Thus, users who wish to maximize the benefits of GAI tools must relinquish some of their privacy, a phenomenon known in the literature as the “privacy paradox.” Background This research aims to determine whether users experience privacy concerns when using GAI tools by examining the relationships among privacy concerns, trust in AI operators, and actual usage patterns among students. It considers different theoretical models (Privacy Calculus, TAM, and UTAUT) to deter-mine if perceived benefits outweigh the risks of information disclosure. Methodology To examine this issue, a quantitative pilot study was conducted using closed-ended questionnaires distributed to 121 students in Israel. Data were analyzed using Pearson correlations and a stepwise multivariate linear regression to identify predictors of GAI tool usage. Contribution This study contributes to the body of knowledge by being the first to examine GAI privacy perceptions within Israeli academia. It challenges the universal applicability of the privacy paradox by demonstrating that, in this specific context, trust and proficiency are more decisive than privacy concerns. Findings The findings indicate that AI-related privacy concerns have no significant correlation with the amount of GenAI tool usage. However, higher levels of trust in the operators of GenAI tools were found to positively affect the amount of GenAI tool use and Internet proficiency. Recommendations for Practitioners Operators of GAI tools should prioritize transparency regarding data collection and usage to build user trust, which directly influences adoption rates. Educational institutions should integrate AI literacy programs to increase students’ proficiency, thereby fostering more confident and effective use of AI tools. Recommendation for Researchers Researchers should move beyond simple correlational studies to explore causal links between emotional engagement and data disclosure. It is also recommend-ed to use validated, multidimensional scales to measure privacy concerns rather than broad self-reports. Impact on Society The findings suggest that as users feel safe and knowledgeable about digital risks, their intent to utilize GAI increases, potentially accelerating the integration of AI into work and study processes. Future Research Future studies should employ larger, more diverse samples across multiple institutions to test the generalizability of these findings. Additionally, a mixed-methods approach, including qualitative interviews, could provide a deeper understanding of the privacy paradox.
This systematic literature review paper explores perspectives on the ideal metaverse from user experience, business, and national levels, considering both academic and industry viewpoints. The study examines the metaverse as a sociotechnical imaginary, enabled collectively by virtual reality (VR), augmented reality (AR), and mixed reality (MR) technologies. Through a systematic literature review, n=144 records were included and by employing grounded theory for analysis of data, we developed three research models, which can guide researchers in examining the metaverse as a sociotechnical future of information technology. Designers can apply the metaverse user experience maturity model to develop more user-friendly services, while business strategists can use the metaverse business readiness model to assess their firms' current state and prepare for transformation. Additionally, policymakers and policy analysts can utilize the metaverse national competitiveness model to track their countries' competitiveness during this paradigm shift. The synthesis of the results also led to the development of practical assessment tools derived from these models that can guide researchers
Aim/Purpose With the technology of artificial intelligence (AI) improving every day it is important to find ways to harness AI in the software development life cycle (SDLC). This research demonstrates how AI tools were incorporated into an upper division Computer Science course to assist with development of various memory games. Background Since ChatGPT’s release in 2022, other companies have released rival chatbots each competing for a piece of the new market. With the plethora of AI options now available, it is important for a developer to learn to use AI as an assistant within the development of a custom project. Methodology The research presented is a multi-case, cross-analysis of four student researchers in a required, senior level Computer Science course. All students were tasked with collecting mixed-methods data on two AI assistants, throughout design and development a unique memory app; then these four students pooled data and conducted a cross-comparative analysis. To prepare for cross analysis, standardized Likert rankings and thematic categories were developed and consistently used during data collections. AI assistants evaluated: Claude, Copilot, ChatGPT Free, and ChatGPT Paid. Throughout the development process, each student provided both of their AI assistants with the same initial queries, the results of which were given a Likert ranking and notes were kept regarding AI accuracy. Individual datasets were examined, then pooled and the combined dataset was used to finalize hypothesis findings. The four student-researchers presented their multi-case, mixed-methods analysis as a snap-shot in time regarding the value of AI as assistants in the development of their projects. Contribution This paper builds on prior research focusing both on student experience and instructional methods in capstone-like courses. This study examines using AIs as assistants as a current trend in Computer Science education. Findings During multi-case analysis, two hypotheses were analyzed against the data of the four student-researchers. The cross examination of data found no statistical significance between the helpfulness of paid vs. free AI as course project assistants; while non-IDE AI assistants performed significantly better than IDE assistants across 7 out of 8 usage type categories. Recommendations for Practitioners Technology instructors can use this research to incorporate AI assistants into advanced courses that focus on building custom software, with cautions that foundational coding skills and knowledge should be in place prior to attempting complex projects. Companies that are researching how AI can be integrated into the software development process can use this research to see preferred strengths of various AI’s, with cautions for use with proprietary data. Recommendations for Researchers Researchers can observe how different AI’s can assist with application development. Further research is encouraged as AI capabilities will continue to evolve. Impact on Society The researchers’ findings show AI in light of its current abilities and limitations in the software development life cycle. While AI assistants excelled in simple to medium complexity debugging tasks, there were many complex tasks where a human coder was preferred over the AI assistants; however, this is expected to change over time. Future Research As future technology strengthens AI some aspects of the study may become historical; however, the core of the research, that of using AI as assistants in development of software projects is expected to remain pertinent to education for some time.
Aim/Purpose To explore the potential of Federated Machine Learning (FML) in developing predictive models while ensuring data privacy and security. Background The rise of data-driven technologies has led to an increased focus on privacy concerns associated with centralized data storage. FML offers a decentralized approach, allowing organizations to collaboratively train models without sharing sensitive data (McMahan et al., 2017). Methodology This study employs a FML framework, utilizing local model training on decentralized datasets, followed by aggregation of model updates to create a global model. Privacy-preserving techniques, such as differential privacy, are also implemented (Dwork & Roth, 2014). Contribution This research contributes to the field of machine learning by demonstrating the efficacy of FML in predictive modeling, highlighting its potential for secure and privacy-conscious applications. Findings The study indicates that FML can effectively enhance model performance while maintaining the privacy of individual data sources. Recommendations for Practitioners Practitioners are encouraged to adopt FML techniques in applications requiring high data security, particularly in sectors such as healthcare and finance. Recommendation for Researchers Future research should explore advanced aggregation methods and evaluate the scalability of FML in diverse settings. Impact on Society The findings of this research have implications for the broader application of machine learning in sensitive areas, promoting data privacy while harnessing the power of collaborative intelligence. Future Research Further investigations should focus on the robustness of FML against adversarial attacks and its applicability in real-world scenarios.
Aim/Purpose This paper explores how large incumbent organizations adopt the newly pro-posed data management approach “Data Mesh”. Particularly, this paper explores to which extent data ownership and data governance are shifting from a centralized to a decentralized approach and whether companies take different paths in this transition. Background Large incumbent organizations suffer from high complexity and often centralized infrastructure of data management that lead not only to high operating costs but also slow down innovation projects and make them more expensive. As a managerial complement of decentralized data management technology such as Data Fabric, Data Mesh was introduced as a management approach to address the organizational challenges of centralization and complexity. Methodology We studied how ten large Swiss incumbent organizations adapted Data Mesh. We interviewed positions such as chief information officers, chief data officers, head of information management, head of group IT. We developed a conceptual framework to position their data management approaches and how much they decentralized data governance and data ownership. Findings and Contribution All large incumbent companies adopt Data Mesh principles, but in different ways and to a different extent. While data governance continues to be increasingly centralized, the degree of data ownership centralization mostly remains unchanged, according to the degree of regulation and other contextual factors. Recommendations for Practitioners Data Mesh is regarded as beneficial by many organizations and thus a relevant option for corporate data management. But there is no “silver bullet” adoption process – instead the degree and way of adoption depends on a certain number of contextual factors, and several reference adoption models exist. Recommendations for Researchers and Future Research The study proposes several hypotheses on Data Mesh adoption that need to be validated in other geographies, other types of organizations (non-incumbents, government, small companies) to better understand adoption paths, adoption barriers, and adoption economies.
