1,053 research outputs found
Meaningful human control: actionable properties for AI system development
No full textHow can humans remain in control of artificial intelligence (AI)-based systems designed to perform tasks autonomously? Such systems are increasingly ubiquitous, creating benefits - but also undesirable situations where moral responsibility for their actions cannot be properly attributed to any particular person or group. The concept of meaningful human control has been proposed to address responsibility gaps and mitigate them by establishing conditions that enable a proper attribution of responsibility for humans; however, clear requirements for researchers, designers, and engineers are yet inexistent, making the development of AI-based systems that remain under meaningful human control challenging. In this paper, we address the gap between philosophical theory and engineering practice by identifying, through an iterative process of abductive thinking, four actionable properties for AI-based systems under meaningful human control, which we discuss making use of two applications scenarios: automated vehicles and AI-based hiring. First, a system in which humans and AI algorithms interact should have an explicitly defined domain of morally loaded situations within which the system ought to operate. Second, humans and AI agents within the system should have appropriate and mutually compatible representations. Third, responsibility attributed to a human should be commensurate with that human’s ability and authority to control the system. Fourth, there should be explicit links between the actions of the AI agents and actions of humans who are aware of their moral responsibility. We argue that these four properties will support practically minded professionals to take concrete steps toward designing and engineering for AI systems that facilitate meaningful human control.Interactive IntelligenceDesign AestheticsCyber SecurityHuman-Robot InteractionEthics & Philosophy of TechnologyHuman Information Communication DesignWeb Information System
Explainable Artificial Intelligence for human-AI collaboration
No full textAs a society, we have come to notice the influence and impact Artificially Intelligent (AI) agents have on the way we live our lives. For these AI agents to support us both effectively and responsibly, we require an understanding on how they make decisions and what the consequences are of these decisions. The research _field of Explainable Artificial Intelligence (XAI) aims to develop AI agents that can explain its own functioning to provide this understanding. In this thesis we defined, developed, and evaluated a core set of explanations an AI agent can provide to support their collaboration with humans.Interactive Intelligenc
Code Smells for Machine Learning Applications
No full textThe popularity of machine learning has wildly expanded in recent years. Machine learning techniques have been heatedly studied in academia and applied in the industry to create business value. However, there is a lack of guidelines for code quality in machine learning applications. In particular, code smells have rarely been studied in this domain. Although machine learning code is usually integrated as a small part of an overarching system, it usually plays an important role in its core functionality. Hence ensuring code quality is quintessential to avoid issues in the long run. This paper proposes and identifies a list of 22 machine learning-specific code smells collected from various sources, including papers, grey literature, GitHub commits, and Stack Overflow posts. We pinpoint each smell with a description of its context, potential issues in the long run, and proposed solutions. In addition, we link them to their respective pipeline stage and the evidence from both academic and grey literature. The code smell catalog helps data scientists and developers produce and maintain high-quality machine learning application code. ACM Reference Format: Haiyin Zhang, Luís Cruz, and Arie van Deursen. 2022. Code Smells for Machine Learning Applications. In 1st Conference on AI Engineering - Software Engineering for AI (CAIN'22), May 16-24, 2022, Pittsburgh, PA, USA. ACM, New York, NY, USA, 12 pages. https://doi.org/10.1145/3522664.3528620 Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.Software EngineeringSoftware Technolog
Creative AI for HRI Design Explorations
No full textDesign fixation, a phenomenon describing designers' adherence to pre-existing ideas or concepts that constrain design outcomes, is particularly prevalent in human-robot interaction (HRI), for example, due to collectively held and stabilised imaginations of what a robot should look like or behave. In this paper, we explore the contribution of creative AI tools to overcome design fixation and enhance creative processes in HRI design. In a four weeks long design exploration, we used generative text-to-image models to ideate and visualise robotic artefacts and robot sociotechnical imaginaries. We exchanged results along with reflections through a digital postcard format. We demonstrate the usefulness of our approach to imagining novel robot concepts, surfacing existing assumptionsand robot stereotypes, and situating robotic artefacts in context.We discuss the contribution to designerly HRI practices and conclude with lessons learnt for using creative AI tools as an emerging design practice in HRI research and beyond.Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.Design Aesthetic
Sustainability of Edge AI at Scale: An empirical study on the sustainability of Edge AI in terms of energy consumption
No full textEdge AI is an architectural deployment tactic that brings AI models closer to the user and data, relieving internet bandwidth usage and providing low latency and privacy. It remains unclear how this tactic performs at scale, since the distribution overhead could impact the total energy consumption. We identify four architectural scalability factors that could impact the energy consumption of AI: environment, optimisation, throughput, and overhead. The latter consists of downloading, verification, and updating the model over time. This work performs an empirical study on the sustainability of Edge AI compared to Cloud AI at scale in terms of energy consumption. For the environment variable, energy consumption measurement experiments are run on a cloud device and multiple edge devices, various quantized models for optimisation, and various throughput levels per hour. We simulate the distribution overhead and combine the results with the measurements to find the holistic energy efficiency of each architectural strategy. We find that all four variables impact energy consumption, but the main contributors are environment, throughput, and overhead. We observe that Edge AI is most energy-efficient in low-distribution, low-demand scenarios, whereas in high-distribution, high-demand scenarios Cloud AI is better optimised and outperforms Edge AI in energy efficiency. This means that developers depending on their use case and the project’s scalability need to consider these quality attributes for the most sustainable architectural solution.https://zenodo.org/records/11065939 Reproducability package https://github.com/rvandernoort/local-vs-cloud Repository of code used for this studyComputer Scienc
Loswal Noord veraf of verdiept: Onderzoek naar de losmogelijkheden van baggerspecie
No full textHet rond Loswal Noord en de Rijnmond verrichte onderzoek reheeft van eerste meting t o t en met het laatste berekeningsresultaat van het driedimensionale sedimenttransport bijgedragen tot een sterk verbeterd inzicht in de natuurlijke processen en de invloed van de mens daarop. Het onderzoek heeft opties aangereikt, waardoor de hoeveelheid baggerwerk jaarlijks kan verminderen. Behalve dat dit financiële voordelen heeft, zijn er enkele zowel positieve als negatieve effecten op het milieu aanwijsbaar. Het gaat er nu om te kiezen tussen een loswal 'veraf', een loswal 'verdiept' en mogelijke combinaties van 'veraf' en 'verdiept'. Ook kan worden besloten Loswal Noord voorlopig te laten wat het is. Voorlopig is daar nog ruimte voor de baggerspecie. De beslissing zal niet alleen genomen kunnen worden op basis van de onderzoeksresultaten. Ook andere (met name sociaal-economische) factoren, die niet in getallen zijn te vatten, spelen immers een rol in het ai dan niet verplaatsen van de Loswal
Mapping Value Sensitive Design onto AI for Social Good Principles
No full textValue Sensitive Design (VSD) is an established method for integrating values into technical design. It has been applied to different technologies and, more recently, to artificial intelligence (AI). We argue that AI poses a number of challenges specific to VSD that require a somewhat modified VSD approach. Machine learning (ML), in particular, poses two challenges. First, humans may not understand how an AI system learns certain things. This requires paying attention to values such as transparency, explicability, and accountability. Second, ML may lead to AI systems adapting in ways that ‘disembody’ the values embedded in them. To address this, we propose a threefold modified VSD approach: 1) integrating a known set of VSD principles (AI4SG) as design norms from which more specific design requirements can be derived; 2) distinguishing between values that are promoted and respected by the design to ensure outcomes that not only do no harm but also contribute to good; and 3) extending the VSD process to encompass the whole life cycle of an AI technology in order to monitor unintended value consequences and redesign as needed. We illustrate our VSD for AI approach with an example use case of a SARS-CoV-2 contact tracing app
Respect as a Lens for the Design of AI Systems
No full textCritical examinations of AI systems often apply principles such as fairness, justice, accountability, and safety, which is reflected in AI regulations such as the EU AI Act. Are such principles sufficient to promote the design of systems that support human flourishing? Even if a system is in some sense fair, just, or 'safe', it can nonetheless be exploitative, coercive, inconvenient, or otherwise conflict with cultural, individual, or social values. This paper proposes a dimension of interactional ethics thus far overlooked: The ways AI systems should treat human beings. For this purpose, we explore the philosophical concept of respect: if respect is something everyone needs and deserves, shouldn't technology aim to be respectful? Despite its intuitive simplicity, respect in philosophy is a complex concept with many disparate senses. Like fairness or justice, respect can characterise how people deserve to be treated; but rather than relating primarily to the distribution of benefits or punishments, respect relates to how people regard one another, and how this translates to perception, treatment, and behaviour. We explore respect broadly across several literatures, synthesising perspectives on respect from Kantian, post-Kantian, dramaturgical, and agential realist design perspectives with a goal of drawing together a view of what respect could mean for AI. In so doing, we identify ways that respect may guide us towards more sociable artefacts that ethically and inclusively honour and recognise humans using the rich social language that we have evolved to interact with one another every day. Green Open Access added to TU Delft Institutional Repository 'You share, we take care!' - Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.Human Information Communication Desig
Tough Decisions? Supporting System Classification According to the AI Act
No full textThe AI Act represents a significant legislative effort by the European Union to govern the use of AI systems according to different risk-related classes, linking varying degrees of compliance obligations to the system's classification. However, it is often critiqued due to the lack of general public comprehension and effectiveness regarding the classification of AI systems to the corresponding risk classes. To mitigate those shortcomings, we propose a Decision-Tree-based framework aimed at increasing robustness, legal compliance and classification clarity with the Regulation. Quantitative evaluation shows that our framework is especially useful to individuals without a legal background, allowing them to improve considerably the accuracy and significantly reduce the time of case classification.Organisation & GovernanceInformation and Communication Technolog
A comparison of computational models with and without genotyping for prediction of response to second-line HIV therapy
No full textOBJECTIVES: We compared the use of computational models developed with and without HIV genotype vs. genotyping itself to predict effective regimens for patients experiencing first-line virological failure. METHODS: Two sets of models predicted virological response for 99 three-drug regimens for patients on a failing regimen of two nucleoside/nucleotide reverse transcriptase inhibitors and one nonnucleoside reverse transcriptase inhibitor in the Second-Line study. One set used viral load, CD4 count, genotype, plus treatment history and time to follow-up to make its predictions; the second set did not include genotype. Genotypic sensitivity scores were derived and the ranking of the alternative regimens compared with those of the models. The accuracy of the models and that of genotyping as predictors of the virological responses to second-line regimens were compared. RESULTS: The rankings of alternative regimens by the two sets of models were significantly correlated in 60-69% of cases, and the rankings by the models that use a genotype and genotyping itself were significantly correlated in 60% of cases. The two sets of models identified alternative regimens that were predicted to be effective in 97% and 100% of cases, respectively. The area under the receiver-operating curve was 0.72 and 0.74 for the two sets of models, respectively, and significantly lower at 0.55 for genotyping. CONCLUSIONS: The two sets of models performed comparably well and significantly outperformed genotyping as predictors of response. The models identified alternative regimens predicted to be effective in almost all cases. It is encouraging that models that do not require a genotype were able to predict responses to common second-line therapies in settings where genotyping is unavailable.AD Revell, MA Boyd, D Wang, S Emery, B Gazzard, P Reiss, AI van Sighem, JS Montaner, HC Lane and BA Larde
- …
