9607 research outputs found
Sort by
The beginning of AI-driven welfare?: A mapping of how public sector AI experiments shape the post-digital welfare state
No full textAugmented and mixed reality for rehabilitation of people living with Parkinson’s disease: a scoping review
No full textA Novel Approach to Teaching Empathic Leadership using Heuristics
No full textIntroduction: NHS failures at Mid-Staffordshire, Shrewsbury and Telford, and East Kent reached the same conclusion: leaders who fail to listen and understand put patients at risk. The NHS response has been to promote compassionate leadership. Yet compassion often entails emotional merging, which can blur boundaries and lead to fatigue. By contrast, empathic leadership, curiosity-driven perspective-taking rather than emotional merging, offers a way forward. Yet no practical method exists to teach it. We tested whether simple heuristics (“rules of thumb”) could fill this gap. Methods: Twenty-one healthcare leaders attended a structured workshop. Using established heuristic development and curriculum design methods, participants generated, refined, and prioritised rules of thumb (heuristics) for empathic leadership, then codesigned a training course. Results: The group produced 35 heuristics and prioritised 12 for teaching healthcare leaders, including “Listen first, speak last” and “Walk the shop floor.” A one-day interprofessional empathic leadership course for 25 participants was then co-designed, featuring experiential learning, roleplay, and implementation planning. Conclusion: Addressing persistent failures in healthcare leadership while avoiding compassionate leadership's limitations, this study was able to prioritise heuristics for empathic leadership and co-design an innovative course to teach them
Modular Neuro-Symbolic Knowledge Graph Completion
No full textKnowledge graph completion (a.k.a. link prediction), i.e., the task of inferring missing edges in knowledge graphs, is a widely used task in many applications, such as product recommendation and question answering. State-of-the-art approaches include knowledge graph embeddings and rule mining which are data-driven and, thus, solely based on the information contained in the input knowledge graph. This leads to unsatisfactory prediction results and ignoresdomain expertise making such solutions inefficient for domains such as healthcare and bioinformatics. To enhance the accuracy of knowledge graph completion we propose Poderoso, a modular neuro-symbolic framework that loosely integrates the data-driven power of knowledge graph embeddings with rule-based reasoning. Poderoso not only enhances the prediction accuracy with domain knowledge via rules stemming from experts but also allows users to plug their own knowledge graph embedding models and reasoning engines. In our preliminary results we show that Poderoso enhances the MRR accuracy of vanilla knowledge graph embeddings and outperforms hybrid solutions that combine knowledge graph embeddings with rule mining. We also discuss how Poderoso can be used in bionformatics, in particular how it can advance research in bacteriophage therapy
Metareferentiality as an Indicator of Procedural Poetics
No full textThe concept of poetics has been central to literature since antiquity, and is still widely used today, often with new dimensions and nuances, such as in the German idea of "Kulturpoetik." Other domains have adapted the concept, e.g., Igor Stravinsky's Poetics of Music (1947), Édouard Glissant's Poetics of Relation (1997), Albin Zak's Poetics of Rock (2001), or Paul Frosh's Poetics of Digital Media (2018). In connection with videogames, though, poetics has never been as widely used as rhetoric (see Bogost 2007) or aesthetics (see Hunicke et al. 2004). Where the term occurs, it is predominantly used to conceptualize poetic games as a category or genre (see Ensslin 2014; Mitchell et al. 2020), or to situate videogames in larger contexts like interactive drama (see Mateas 2001) or interactive narrative (see Ryan 2006). The chapter proposes—before the background of the existing, predominantly prescriptive and theoretical poetics of videogames—a descriptive, analytical poetics that explains relationships between works more nuancedly than established notions like genre. Metareferentiality is crucial for such a poetics, as moments of reflection on a work's own features and on its medium are markers and comments on the immanent poetics of a work
TensorSocket: Shared Data Loading for Deep Learning Training
No full textTraining deep learning models is a repetitive and resource-intensive process. Data scientists often train several models before landing on a set of parameters (e.g., hyper-parameter tuning) and model architecture (e.g., neural architecture search), among other things that yield the highest accuracy. The computational efficiency of these training tasks depends highly on how well the training data is supplied to the training process. The repetitive nature of these tasks results in the same data processing pipelines running over and over, exacerbating the need for and costs of computational resources.In this paper, we present TensorSocket to reduce the computational needs of deep learning training by enabling simultaneous training processes to share the same data loader. TensorSocket mitigates CPU-side bottlenecks in cases where the collocated training workloads have high throughput on GPU, but are held back by lower data-loading throughput on CPU. TensorSocket achieves this by reducing redundant computations and data duplication across collocated training processes and leveraging modern GPU-GPU interconnects. While doing so, TensorSocket is able to train and balance differently-sized models and serve multiple batch sizes simultaneously and is hardware- and pipeline-agnostic in nature.Our evaluation shows that TensorSocket enables scenarios that are infeasible without data sharing, increases training throughput by up to 100%, and when utilizing cloud instances, achieves cost savings of 50% by reducing the hardware resource needs on the CPU side. Furthermore, TensorSocket outperforms the state-of-the-art solutions for shared data loading such as CoorDL and Joader; it is easier to deploy and maintain and either achieves higher or matches their throughput while requiring fewer CPU resources
Prompt refinement or fine-tuning? best practices for using LLMs in computational social science tasks
Get PDFLarge Language Models are expressive tools that enable complex tasks of text understanding within Computational Social Science. Their versatility, while beneficial, poses a barrier for establishing standardized best practices within the field. To bring clarity on the values of different strategies, we present an overview of the performance of modern LLM-based classification methods on a benchmark of 23 social knowledge tasks. Our results point to three best practices: prioritize models with larger vocabulary and pre-training corpora; avoid simple zero-shot in favor of AI-enhanced prompting; fine-tune on task-specific data, and consider more complex forms instruction-tuning on multiple datasets only when only training data is more abundant
