67471 research outputs found
Sort by
Mining Hidden Prompt Engineering Patterns with Formal Concept Analysis and Association Rules
Full text linkDesigning effective prompts to guide generative artificial intelligence (GAI) systems, or prompt engineering, has become a crucial skill. However, the underlying prompt patterns have not yet been thoroughly examined. This paper introduces a novel analytical method that combines formal concept analysis (FCA) and association rule mining. This approach is used to systematically analyze prompt engineering behaviors within an empirical dataset of human–AI interactions. Findings reveal hidden prompt patterns linking prompts to GAI outputs, providing insights that traditional analyses cannot offer. Furthermore, we demonstrate that prompting guides, especially those with examples, facilitate more sophisticated prompt engineering behavior and improve GAI output quality. Our work contributes to information systems theory by demonstrating the value of FCA-based structural analysis in human–GAI contexts and to the practice of prompt engineering by offering evidence-based guidance on improving prompt design and prompt engineering skill development
Considering the Material-discursive Practice: Enacting the Unspoken Goal
Full text linkGoals are commonly recognized through their materialization in explicit statements. By applying a sociomaterial lens to the practice of goal formation, the scope of the analysis is widened beyond pre-defined social and material actors. Hence, considering also the mundane, presumed and overlooked spatial and material dimensions. Through observations of workshops in a home care service quality development initiative in a local Swedish government, it is shown how goals are in becoming and performatively configured given conditioned possibilities of material-discursive arrangements. This study contributes to IS research by demonstrating the applicability of a sociomaterial lens in understanding social phenomena. This is done by showing how goals are formed not only by what actors say and do, but also where, when, how and with what they do it. Through downplaying language and increasing tentativeness toward significant material dimensions, this study shows how goals can be perceived when not necessarily spoken or conscious
Aggregate Modeling of Air-Conditioner Loads Under Packet-based Control with Both On and Off Grid Access Requests
Full text linkCoordination of distributed energy resources (DERs) can engender flexibility necessary to improve grid reliability. Packetized Energy Management (PEM) is a method for coordinating DERs, such as thermostatically controlled loads (TCLs) and electric vehicles, within customer quality-of-service (QoS) limits. In PEM, a DER uses local information to offer flexibility by sending a request to the DER coordinator to turn-ON or turn-OFF. Much work has focused on modeling and analyzing aggregations of DERs under PEM with fixed packet durations and only turn-ON requests. Different recent efforts to enable variable packet lengths have shown an increase in available flexibility and ramping capability, but have not been modeled in aggregate, which limits systematic analyses. To address this issue, this paper presents a new aggregate bin-based (macro) model of PEM loads that incorporates both turn-ON and turn-OFF request features, enabling the model to accurately characterize the capability of the fleet of DERs to track a power reference signal, population temperature dynamics, aggregate request rates, and variable packet lengths. Simulation-based validation is performed against an agent-based (micro) model to evaluate robustness and quantify model accuracy. Finally, the distribution of variable packet lengths from macro-model simulations are applied to inform past work on PEM with randomized packet lengths
Towards Quantifying Compliance with the EU AI Act
Full text linkAs AI systems proliferate in high-risk domains, assessing their compliance with emerging regulatory standards has become imperative. The EU AI Act outlines ethical requirements across five dimensions: explainability, fairness, privacy, robustness, and social and environmental well-being. However, existing evaluation approaches lack a unified methodology to quantitatively operationalize these principles. In this paper, we propose a structured, score-based framework that translates the Act’s pillars into 22 interpretable metrics, enabling reproducible, model-agnostic compliance assessments. Applied to three benchmark tabular classification tasks using a standardized deep learning model, our framework captures how dataset characteristics shape ethical performance. The results reveal key trade-offs: models with high predictive accuracy do not necessarily meet compliance expectations, and larger datasets tend to improve robustness but increase vulnerability to privacy leakage. Correlation analyses expose metric redundancy in fairness and explainability, suggesting potential for simplification. Privacy metrics, by contrast, remain essential and diverse. Social and environmental measures emerge as least mature, underscoring the need for novel, bounded metrics in future research
Collaborative Use of Information Systems: Joint IT Use Analysis Framework and Typology
Full text linkInformation systems (IS) often involve joint use of information technology (IT), named joint IT use, where multiple users share an IT interface. However, IS literature suggests gaps in understanding group-level joint IT use patterns and their antecedents, consequences, and mechanisms. This paper proposes a holistic framework of joint IT use, offering a typology based on technical elements (system and task) and action levels (user and user group). The framework, inspired by the Input-Mediators-Output-Input model, consists of three layers. (1) Inputs, including characteristics at system, task, individual, group, and organization levels, plus IS triggers; (2) Mediators, made of individual-level, group-level, and cross-level configurations of emotions, cognitions, and behaviors, system attributes, and task configurations; (3) Outcomes, made of consequences of the Mediators layer, which influence the Inputs layer. This framework supports future research in IT use, including on multilevel view of IT use and real-time IT use patterns
Familiarity with and Attitudes Towards Chatbots: Findings from a Three-Wave National Surveys of U.S. Adults Before and After ChatGPT
Full text linkGenerative AI technologies like ChatGPT have transformed how people interact with information and services. However, it is unclear (1) what the public knew about chatbots before ChatGPT, (2) how those understandings have evolved, and (3) whether digital divides exist in these understandings. To explore this, we conducted a three-wave national online survey. Wave 1 data were collected just before ChatGPT’s 2022 release; Waves 2 and 3 followed one and two years later. Each wave assessed chatbot familiarity (awareness, use, frequency), and Waves 2 and 3 included generative AI. We also measured trust in, support for, and intentions to use chatbots. We analyzed changes over time and examined differences by age, education, and income. Results suggest that people are more familiar with chatbots post-ChatGPT and use of these technologies is associated with more positive attitudes toward them. We further find evidence of digital divides across age and, increasingly, education and income
Expert-Informed Design of an AI-Augmented Preventive Health App for Young Adults
Full text linkPreventive digital health solutions (DHSs) are critical for addressing chronic conditions such as prediabetes, which is a growing concern, particularly among young adults (aged 18–34). Existing wellness apps suffer high dropout rates due to poor usability and inclusivity. MiCARE, an expert-informed, AI-augmented progressive web app (PWA), addresses these gaps through a novel multi-theoretical framework integrating Self-Determination Theory (SDT), CARE (“Compassion”, “Assistance”, “Respect”, “Empathy”), User-Centered Design (UCD) and Inclusive Design. For instance, its empathetic chatbot, using closed-domain natural language processing (NLP) with clinically verified and multilingual responses, promotes culturally adaptive engagement. MiCARE was developed using the Design Science Research Methodology (DSRM) and a WCAG 2.1-compliant prototype was refined through multidisciplinary expert feedback, adhering to La Trobe University ethics approval. A pilot study is planned to evaluate usability, usefulness, and satisfaction using an integrated Task-Technology Fit (TTF)- Unified Theory of Acceptance and Use of Technology (UTAUT) framework. MiCARE offers a replicable, theory-driven blueprint for designing and evaluating inclusive DHSs, validated through expert review
Historical Homologation in AI Algorithmic Computation: When the Past Decides Your Future
Full text linkThis paper introduces “historical homologation”, the systematic tendency of algorithms to make future decisions match past patterns regardless of contemporary evidence. Analyzing the 2020 UK A-level grading controversy, where algorithms downgraded 40% of teacher assessments, we demonstrate how the Ofqual DCP algorithm was designed to protect grade distributions rather than predict individual achievement. Through analysis of 55,000 schools, we identify three core mechanisms. Historical anchoring transformed 2017-19 grade averages into computational rules functioning as hard ceilings. Individual erasure compressed all achievement data into class averages, systematically disadvantaging high-achievers in lower-performing schools. Temporal smoothing operated as a low-pass filter, pulling trajectories back toward historical means. These mechanisms interact synergistically to create computational determinism, the structural necessity that algorithms reproduce rather than transcend historical patterns. This reveals historical homologation as an ontological constraint where historically anchored algorithms shape social futures by overlooking the very changes and exceptions that systems should prioritiz
An Institutional Analysis of Litigation Growth in Brazil's Electronic Justice System
Full text linkThis study examines an unexpected consequence of judicial digitalisation in Brazil: the substantial increase in new legal cases following the implementation of the Court Case Management System (CCMS). While digital technologies were introduced to improve judicial efficiency, reduce backlogs, and lower costs, Brazil’s experience shows that such technologies can also generate unforeseen effects, specifically, a rise in both legitimate and predatory litigation. Drawing on the Technology Enactment Framework (Fountain, 2001), this paper employs qualitative research, including interviews with legal actors and an analysis of official data, to examine how institutional arrangements influence the outcomes of IT implementation in the judiciary. Findings reveal that while CCMS expanded access to justice, it also enabled abusive litigation practices that strain judicial resources. The study highlights the importance of institutional context in shaping technology use and offers critical insights for designing more resilient and adaptive e-justice systems