This paper introduces and contextualises Climate Futures, an experiment in which AI was repurposed as a ‘co-author’ of climate stories and a co-designer of climate-related images that facilitate reflections on present and future(s) of living with climate change. It converses with histories of writing and computation, including surrealistic ‘algorithmic writing’, recombinatory poems and ‘electronic literature’. At the core lies a reflection about how machine learning’s associative, predictive and regenerative capacities can be employed in playful, critical and contemplative goals. Our goal is not automating writing (as in product-oriented applications of AI). Instead, as poet Charles Hartman argues, ‘the question isn’t exactly whether a poet or a computer writes the poem, but what kinds of collaboration might be interesting’ (1996, p. 5). STS scholars critique labs as future-making sites and machine learning modelling practices and, for example, describe them also as fictions. Building on these critiques and in line with ‘critical technical practice’ (Agre, 1997), we embed our critique of ‘making the future’ in how we employ machine learning to design a tool for looking ahead and telling stories on life with climate change. This has involved engaging with climate narratives and machine learning from the critical and practical perspectives of artistic research. We trained machine learning algorithms (i.e. GPT-2 and AttnGAN) using climate fiction novels (as a dataset of cultural imaginaries of the future). We prompted them to produce new climate fiction stories and images, which we edited to create a tarot-like deck and a story-book, thus also playfully engaging with machine learning’s predictive associations. The tarot deck is designed to facilitate conversations about climate change. How to imagine the future beyond scenarios of resilience and the dystopian? How to aid our transition into different ways of caring for the planet and each other?
DOCUMENT
Machine learning models have proven to be reliable methods in classification tasks. However, little research has been done on classifying dwelling characteristics based on smart meter & weather data before. Gaining insights into dwelling characteristics can be helpful to create/improve the policies for creating new dwellings at NZEB standard. This paper compares the different machine learning algorithms and the methods used to correctly implement the models. These methods include the data pre-processing, model validation and evaluation. Smart meter data was provided by Groene Mient, which was used to train several machine learning algorithms. The models that were generated by the algorithms were compared on their performance. The results showed that Recurrent Neural Network (RNN) 2performed the best with 96% of accuracy. Cross Validation was used to validate the models, where 80% of the data was used for training purposes and 20% was used for testing purposes. Evaluation metrices were used to produce classification reports, which can indicate which of the models work the best for this specific problem. The models were programmed in Python.
DOCUMENT
Background: Although physical activity (PA) has positive effects on health and well-being, physical inactivity is a worldwide problem. Mobile health interventions have been shown to be effective in promoting PA. Personalizing persuasive strategies improves intervention success and can be conducted using machine learning (ML). For PA, several studies have addressed personalized persuasive strategies without ML, whereas others have included personalization using ML without focusing on persuasive strategies. An overview of studies discussing ML to personalize persuasive strategies in PA-promoting interventions and corresponding categorizations could be helpful for such interventions to be designed in the future but is still missing. Objective: First, we aimed to provide an overview of implemented ML techniques to personalize persuasive strategies in mobile health interventions promoting PA. Moreover, we aimed to present a categorization overview as a starting point for applying ML techniques in this field. Methods: A scoping review was conducted based on the framework by Arksey and O’Malley and the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) criteria. Scopus, Web of Science, and PubMed were searched for studies that included ML to personalize persuasive strategies in interventions promoting PA. Papers were screened using the ASReview software. From the included papers, categorized by the research project they belonged to, we extracted data regarding general study information, target group, PA intervention, implemented technology, and study details. On the basis of the analysis of these data, a categorization overview was given. Results: In total, 40 papers belonging to 27 different projects were included. These papers could be categorized in 4 groups based on their dimension of personalization. Then, for each dimension, 1 or 2 persuasive strategy categories were found together with a type of ML. The overview resulted in a categorization consisting of 3 levels: dimension of personalization, persuasive strategy, and type of ML. When personalizing the timing of the messages, most projects implemented reinforcement learning to personalize the timing of reminders and supervised learning (SL) to personalize the timing of feedback, monitoring, and goal-setting messages. Regarding the content of the messages, most projects implemented SL to personalize PA suggestions and feedback or educational messages. For personalizing PA suggestions, SL can be implemented either alone or combined with a recommender system. Finally, reinforcement learning was mostly used to personalize the type of feedback messages. Conclusions: The overview of all implemented persuasive strategies and their corresponding ML methods is insightful for this interdisciplinary field. Moreover, it led to a categorization overview that provides insights into the design and development of personalized persuasive strategies to promote PA. In future papers, the categorization overview might be expanded with additional layers to specify ML methods or additional dimensions of personalization and persuasive strategies.
DOCUMENT
In het ziekenhuis kan elke fout een leven kosten. Zo kan al een kleine bereidingsfout bij het klaarmaken van intraveneuze medicijnen (IV) leiden tot levensbedreigende omstandigheden voor de patiënt. Bereiding van dit type medicijnen gebeurt in de apotheek en op de verpleegafdeling. Met name op de verpleegafdeling is het een drukke en onvoorspelbare setting. Wereldwijd komen in deze setting ernstige bereidingsfouten nog te frequent voor. Om deze menselijke fouten te reduceren, wordt in deze KIEM aanvraag een proof-of-concept ‘slim oog’ ontwikkeld die vlak voor de toediening detecteert of de juiste dosis aanwezig is, of het type medicijn correct is en geen vervuiling aanwezig is. Het slimme oog maakt gebruik van hyperspectrale technologie en artificial intelligence, en is een samenwerking tussen de Computer Vision & Data Science afdeling van NHL Stenden Hogeschool, de automatische medicijncontrole specialist ZiuZ, en het Tjongerschans ziekenhuis. De unieke combinatie tussen nieuwe AI-technieken, hyperspectrale techniek en de toepassing op intraveneuze medicijnen is voor dit consortium technisch nieuw, en is nog niet eerder ontwikkeld voor de toepassing aan het bed of in de medicijnkamer op de verpleegafdeling. De onvoorspelbare setting en de urgentie aan het bed maakt dit onderzoek technisch uitdagend. Tevens moet het uiteindelijke device klein en draagbaar en snel werkzaam zijn. Om de grote verscheidenheid aan mogelijke gebruik scenario's en menselijke fouten te vangen in het algoritme, wordt een door NHLS ontwikkelde simulatie procedure gevolgd: met nabootsing van de praktijksituatie in samenwerking met zorgverleners, met opzettelijke fouten, en computer gegenereerde beeldmanipulatie. Het project zal geïntegreerd worden in het onderwijs volgens de design-based methode, met teams bestaande uit domein experts, bedrijven, docent-onderzoekers en studenten. Het uiteindelijke doel is om met een proof-of-concept aan-het-bed demonstrator een groot consortium van ziekenhuizen, ontwikkelaars en eindgebruikers enthousiast te maken voor een groter vervolgproject.
Huntington’s disease (HD) and various spinocerebellar ataxias (SCA) are autosomal dominantly inherited neurodegenerative disorders caused by a CAG repeat expansion in the disease-related gene1. The impact of HD and SCA on families and individuals is enormous and far reaching, as patients typically display first symptoms during midlife. HD is characterized by unwanted choreatic movements, behavioral and psychiatric disturbances and dementia. SCAs are mainly characterized by ataxia but also other symptoms including cognitive deficits, similarly affecting quality of life and leading to disability. These problems worsen as the disease progresses and affected individuals are no longer able to work, drive, or care for themselves. It places an enormous burden on their family and caregivers, and patients will require intensive nursing home care when disease progresses, and lifespan is reduced. Although the clinical and pathological phenotypes are distinct for each CAG repeat expansion disorder, it is thought that similar molecular mechanisms underlie the effect of expanded CAG repeats in different genes. The predicted Age of Onset (AO) for both HD, SCA1 and SCA3 (and 5 other CAG-repeat diseases) is based on the polyQ expansion, but the CAG/polyQ determines the AO only for 50% (see figure below). A large variety on AO is observed, especially for the most common range between 40 and 50 repeats11,12. Large differences in onset, especially in the range 40-50 CAGs not only imply that current individual predictions for AO are imprecise (affecting important life decisions that patients need to make and also hampering assessment of potential onset-delaying intervention) but also do offer optimism that (patient-related) factors exist that can delay the onset of disease.To address both items, we need to generate a better model, based on patient-derived cells that generates parameters that not only mirror the CAG-repeat length dependency of these diseases, but that also better predicts inter-patient variations in disease susceptibility and effectiveness of interventions. Hereto, we will use a staggered project design as explained in 5.1, in which we first will determine which cellular and molecular determinants (referred to as landscapes) in isogenic iPSC models are associated with increased CAG repeat lengths using deep-learning algorithms (DLA) (WP1). Hereto, we will use a well characterized control cell line in which we modify the CAG repeat length in the endogenous ataxin-1, Ataxin-3 and Huntingtin gene from wildtype Q repeats to intermediate to adult onset and juvenile polyQ repeats. We will next expand the model with cells from the 3 (SCA1, SCA3, and HD) existing and new cohorts of early-onset, adult-onset and late-onset/intermediate repeat patients for which, besides accurate AO information, also clinical parameters (MRI scans, liquor markers etc) will be (made) available. This will be used for validation and to fine-tune the molecular landscapes (again using DLA) towards the best prediction of individual patient related clinical markers and AO (WP3). The same models and (most relevant) landscapes will also be used for evaluations of novel mutant protein lowering strategies as will emerge from WP4.This overall development process of landscape prediction is an iterative process that involves (a) data processing (WP5) (b) unsupervised data exploration and dimensionality reduction to find patterns in data and create “labels” for similarity and (c) development of data supervised Deep Learning (DL) models for landscape prediction based on the labels from previous step. Each iteration starts with data that is generated and deployed according to FAIR principles, and the developed deep learning system will be instrumental to connect these WPs. Insights in algorithm sensitivity from the predictive models will form the basis for discussion with field experts on the distinction and phenotypic consequences. While full development of accurate diagnostics might go beyond the timespan of the 5 year project, ideally our final landscapes can be used for new genetic counselling: when somebody is positive for the gene, can we use his/her cells, feed it into the generated cell-based model and better predict the AO and severity? While this will answer questions from clinicians and patient communities, it will also generate new ones, which is why we will study the ethical implications of such improved diagnostics in advance (WP6).
