67471 research outputs found
Sort by
Tool-Augmented LLMs for Rapid Data Insights: Empowering Non-Expert Users in Open Government Data Contexts
Full text linkOpen Government Data (OGD) initiatives aim to foster transparency and innovation, yet actual usage remains low due to limited user resources, low data literacy, and a lack of supportive tools. Adopting a Design Science Research (DSR) approach, this study explores how systems must be designed to enable non-expert users to effectively interact with OGD. We propose a design theory comprising design requirements, design principles, and design features, which we instantiate in a prototypical system based on the ChatGPT platform. The core design integrates large language models (LLMs) with tool augmentation techniques to enable fully automated data retrieval, analysis, visualization, and interpretation through natural language interaction. Initial formative evaluations indicate that tool-augmented LLMs can substantially lower interaction barriers for non-expert users, while limitations in accuracy and reliability remain. Our study contributes prescriptive design knowledge and practical guidance for developing advanced natural language interfaces for OGD platforms
Beyond Zero-Shot: Enhancing LLM Financial Complaint Classification with Relevancy-Driven RAG-Based Few-Shot Prompting
Full text linkLarge language models (LLMs) have shown significant promise in natural language processing (NLP) tasks, yet their efficacy in real-world consumer complaint classification without fine-tuning remains a challenge. Zero-shot classification offers a valuable solution for categorizing consumer complaints, particularly for handling new and dynamic financial issues, as it allows models to classify data without prior labeled training. However, the nuanced and often overlapping nature of financial complaint categories makes this task particularly difficult. This study explores both zero-shot and a novel few-shot prompting approach for classifying consumer complaints submitted to the Consumer Financial Protection Bureau (CFPB). We compared traditional zero-shot prompting with two few-shot methods: one using randomly selected classified examples and another leveraging the top 5 most relevant classified examples with semantic similarity for in-context learning. Our results consistently demonstrated superior performance with our relevancy-driven, retrieval-augmented generation (RAG) prompting. To validate these findings and ensure they weren't due to chance, we replicated our experiments across several leading LLM models, including GPT-4o, QWEN, Deepseek V3, and Anthropic Claude Sonnet 4.0. Across all tested models, the relevancy-based few-shot approach yielded consistently better results, which we rigorously validated using accuracy, precision, recall, and F1-score. Furthermore, when benchmarked against traditional machine learning models including a fine-tuned RoBERTa, SVM, and logistic regression, our relevancy-driven few-shot approach demonstrated markedly superior performance, validating its effectiveness for this complex classification task. This research highlights the significant potential of carefully curated, relevant examples in enhancing LLM performance for complex text classification tasks in the financial domain
Regulating Deceptive Design: A Comparative Analysis of U.S. and EU Laws on Dark Patterns
Full text linkWith the expansion of digital interactions, users are increasingly subject to dark patterns—interface designs intended to manipulate decision-making. Regulatory responses from the European Union (EU) and the United States (US) aim to mitigate these practices, yet protections remain inconsistent across jurisdictions. Despite growing concern, there is limited research comparing how EU and US regulations approach dark patterns. This paper conducts a thematic analysis and word frequency test of relevant regulations, finding that EU frameworks insufficiently address specific technological risks and lack extended safeguards for vulnerable groups. In contrast, US regulations fall short in governing gatekeepers’ use of manipulative designs and lack robust oversight of artificial intelligence (AI) systems. The paper concludes with recommendations, including expanding protections for vulnerable users in the EU and strengthening gatekeeper and AI-related provisions in the US
Unlocking Value in Digital Ecosystems: A Framework of Five Analytical Lenses
Full text linkResearch on value creation in digital ecosystems is extensive but remains fragmented, lacking a cohesive structure to organize its diverse findings. This study addresses this gap by conducting a systematic bibliometric analysis of 366 articles to map the field's intellectual structure. We ask: What are the dominant analytical lenses through which value creation in digital ecosystems is researched? The analysis reveals that the literature can be understood through five distinct lenses: (1) Resilience and Trust, (2) Sustainability and Socio-Ecological Impact, (3) Knowledge and Integration, (4) Commercialization and Value Capture, and (5) Foundational Technology and Interoperability. This framework provides a novel typology for the field, highlighting critical trade-offs and integrating disparate research streams. It offers a more structured approach for both scholars and practitioners to analyze and navigate the complex dynamics of creating and sustaining value in the digital economy
Reframing Talent Management with AI: The Organizing Vision of Skills Intelligence
No full textSkills intelligence is rapidly redefining talent management approaches, fueling the shift from credential-based to skills-based strategies. Drawing on organizing vision theory, this paper conceptualizes skills intelligence as an emergent socio-technical phenomenon enabled by AI. Through qualitative data analysis of the practitioner discourse, we identify the key components, functions, and affordances of skills intelligence, and argue that it has profound implications for employees, organizations, and labor markets. Our findings reveal that although industry narratives highlight the positive outcomes of efficiency and agility associated with AI-enabled skills systems, they frequently neglect critical perspectives on their unintended consequences. The paper concludes with a call to action to mobilize information systems (IS) and talent management scholars toward a critical and future-oriented approach to shaping the trajectory of this evolving technology
Ensuring the Emergence of Collective Intelligence in Virtual Group Work
Full text linkThis research addresses how technology-based group work interventions can facilitate the emergence of collective intelligence—the measurable ability of a group to perform consistently well across tasks. Specifically, this study uses coordination theory to theoretically and empirically examine the conditions under which collective intelligence emerges and shows how existing technology-based process structuring techniques should allow groups to work more consistently across tasks
Anomaly Detection in Multivariate Time Series: Combining LSTM Autoencoders with Contrastive Learning
Full text linkWe propose a Long Short-Term Memory(LSTM)-based autoencoder model for multivariate time series anomaly detection that incorporates contrastive learning tailored to time-series characteristics. By leveraging contrastive representation learning, the model effectively pulls normal data closer to the original representation while pushing anomalous data further away, enhancing detection performance. To generate positive and negative pairs, the model applies time series-specific augmentations by sampling overlapping segments, preserving contextual integrity. It combines both instance and temporal contrastive learning to capture richer representations. Training is guided by a joint loss function that integrates weighted contrastive loss with reconstruction loss. Experimental results demonstrate that the proposed method improves F1-score by 2–6% over baseline models. This work highlights that even with a simple LSTM-based autoencoder architecture, significant gains in anomaly detection can be achieved by incorporating contrastive learning strategies suited for time series data
Introduction to the Minitrack on Trustworthy Artificial Intelligence and Machine Learning
No full textKip-Agenge Knowledge: Governing Indigenous Medicinal Knowledge for Future Generations in Africa
Full text linkAfrican communities are known for local and traditional medicine; the urgency of digitizing and governing Indigenous knowledge on traditional medicine has become increasingly evident due to the impending loss of Indigenous knowledge resulting from COVID-19, climate change, and aging Indigenous knowledge holders. The study examines the governance of traditional medicine among the Keiyo community of Kenya. A qualitative approach was adopted, employing thematic analysis and the Governing Knowledge Commons framework to guide the mapping of local and Indigenous medical commons, evaluation of policy frameworks, knowledge cafés, and discussions with Indigenous knowledge holders and practitioners from diverse backgrounds. Ethical considerations were integral to the research, incorporating an epistemic justice framework reflective of the author's positionality. The research highlights the impact of the GKC framework on Indigenous knowledge practices in sustainable resource management, as well as its contribution to the governance of Indigenous medicine
Structured Pixels: Satellite Imagery as the Cause in Causal Effect Estimation
Full text linkWe present Structured Pixels (SP), a causal inference model that positions satellite imagery as a cause/treatment in a causal graph, rather than merely a proxy for outcomes or confounders. Built on the generalized Robinson decomposition and a two-step, R-learner-inspired algorithm, SP uses learned latent representations to partial out confounding influences and isolate the causal effect. Its modular training pipeline supports integration with diverse machine learning models across domains. We evaluate SP using semi-synthesized datasets on two tasks: the impact of environmental conditions on mosquito populations and the influence of coastal characteristics on dark vessel prevalence. SP consistently outperforms baseline methods, and its learned representations capture meaningful environmental patterns. We further demonstrate SP’s applicability by re-examining the relationship between deforestation and agricultural productivity with real-world data; the results align with prior work. These findings highlight SP’s potential to advance GeoAI for environmental monitoring and resource management