Evaluation of trustworthy artificial intelligent healthcare applications using multi-criteria decision-making approach

Expert Systems with Applications, Год журнала: 2024, Номер 246, С. 123066 - 123066

Опубликована: Янв. 21, 2024

The purpose of this paper is to propose a novel hybrid framework for evaluating and benchmarking trustworthy artificial intelligence (AI) applications in healthcare by using multi-criteria decision-making (MCDM) techniques under new fuzzy environment. To develop such framework, decision matrix has been built, then integrated with q-ROF2TL-FWZIC (q‐Rung Orthopair Fuzzy 2‐Tuple Linguistic Fuzzy-Weighted Zero-Inconsistency) q-ROF2TL-CODAS Combinative Distance-Based Assessment). In integration, utilized assigning the weights evaluation attributes AI, while employed AI applications. Findings show that method effectively attributes. transparency attribute receives highest importance weight (0.173566825), whereas human agency oversight criterion lowest (0.105741901). remaining are distributed between. Moreover, alternative_4 rank order (score 7.370410417), alternative_13 −4.759794397). evaluate validity proposed systematic ranking sensitivity analysis assessments were employed.

Язык: Английский

---

Trustworthy AI in the public sector: An empirical analysis of a Swedish labor market decision-support system

Technology in Society, Год журнала: 2024, Номер 76, С. 102471 - 102471

Опубликована: Янв. 26, 2024

This paper investigates the deployment of Artificial Intelligence (AI) in Swedish Public Employment Service (PES), focusing on concept trustworthy AI public decision-making. Despite Sweden's advanced digitalization efforts and widespread application sector, our study reveals significant gaps between theoretical ambitions practical outcomes, particularly context AI's trustworthiness. We employ a robust framework comprising Institutional Theory, Resource-Based View (RBV), Ambidexterity to analyze challenges discrepancies implementation within PES. Our analysis shows that while promises enhanced decision-making efficiency, reality is marred by issues transparency, interpretability, stakeholder engagement. The opacity neural network used agency assess jobseekers' need for support lack comprehensive technical understanding among PES management contribute achieving transparent interpretable systems. Economic pressures efficiency often overshadow ethical considerations involvement, leading decisions may not be best interest jobseekers. propose recommendations enhancing trustworthiness services, emphasizing importance engagement, involving jobseekers process. advocates more nuanced balance use technologies leveraging internal resources such as skilled personnel organizational knowledge. also highlight improved literacy both effectively navigate integration into processes. findings ongoing debate AI, offering detailed case bridges gap exploration application. By scrutinizing PES, we provide valuable insights guidelines other sector organizations grappling with their

Язык: Английский

---

Multimodal Healthcare AI: Identifying and Designing Clinically Relevant Vision-Language Applications for Radiology

Опубликована: Май 11, 2024

Recent advances in AI combine large language models (LLMs) with vision encoders that bring forward unprecedented technical capabilities to leverage for a wide range of healthcare applications. Focusing on the domain radiology, vision-language (VLMs) achieve good performance results tasks such as generating radiology findings based patient's medical image, or answering visual questions (e.g., "Where are nodules this chest X-ray?"). However, clinical utility potential applications these is currently underexplored. We engaged an iterative, multidisciplinary design process envision clinically relevant VLM interactions, and co-designed four use concepts: Draft Report Generation, Augmented Review, Visual Search Querying, Patient Imaging History Highlights. studied concepts 13 radiologists clinicians who assessed valuable, yet articulated many considerations. Reflecting our findings, we discuss implications integrating more generally.

Язык: Английский

---

Towards Clinically Useful AI: From Radiology Practices in Global South and North to Visions of AI Support

ACM Transactions on Computer-Human Interaction, Год журнала: 2025, Номер unknown

Опубликована: Фев. 5, 2025

Despite recent advancements, real-world use of Artificial Intelligence (AI) in radiology remains low, often due to the mismatch between AI offerings and situated challenges faced by healthcare professionals. To bridge this gap, we conducted a field study at nine medical sites Denmark Kenya with two goals: (1) understand radiologists during chest X-ray practice; (2) envision alternative futures that align collaborative clinical work. This uniquely grounds design insights comprehensive characterisation diagnostic work across multiple geographical institutional contexts. Building on ideas articulated interviewed (N=18), conceptualised five visions transcend traditional notions support. These emphasise usefulness AI-based systems depends their configurability flexibility three dimensions: type site, expertise professionals, situational patient Addressing these dependencies requires expanding space envisioning rooted realities practice rather than solely following trajectory development.

Язык: Английский

---

Introduction to the Special Issue on Human-Centred AI in Healthcare: Challenges Appearing in the Wild

ACM Transactions on Computer-Human Interaction, Год журнала: 2023, Номер 30(2), С. 1 - 12

Опубликована: Апрель 30, 2023

The emerging concept of Human-Centred Artificial Intelligence (HCAI) involves the amplification, augmentation, empowerment, and enhancement individuals. goal HCAI is to ensure that AI meets our needs while also operating transparently, ...

Язык: Английский

---

"It depends": Configuring AI to Improve Clinical Usefulness Across Contexts

Designing Interactive Systems Conference, Год журнала: 2024, Номер unknown, С. 874 - 889

Опубликована: Июнь 29, 2024

Artificial Intelligence (AI) repeatedly match or outperform radiologists in lab experiments. However, real-world implementations of radiological AI-based systems are found to provide little no clinical value. This paper explores how design AI for usefulness different contexts. We conducted 19 sessions and interventions with 13 from 7 sites Denmark Kenya, based on three iterations a functional prototype. Ten sociotechnical dependencies were identified as crucial the radiology. conceptualised four technical dimensions that must be configured intended context use: functionality, medical focus, decision threshold, Explainability. present recommendations address pertaining knowledge, clinic type, user expertise level, patient context, situation condition configuration these dimensions.

Язык: Английский

---

Early experiences of integrating an artificial intelligence-based diagnostic decision support system into radiology settings: a qualitative study

Journal of the American Medical Informatics Association, Год журнала: 2023, Номер 31(1), С. 24 - 34

Опубликована: Сен. 25, 2023

Abstract Objectives Artificial intelligence (AI)-based clinical decision support systems to aid diagnosis are increasingly being developed and implemented but with limited understanding of how such integrate existing work organizational practices. We explored the early experiences stakeholders using an AI-based imaging software tool Veye Lung Nodules (VLN) aiding detection, classification, measurement pulmonary nodules in computed tomography scans chest. Materials methods performed semistructured interviews observations across adopter deployment sites clinicians, strategic decision-makers, suppliers, patients long-term chest conditions, academics expertise use diagnostic AI radiology settings. coded data Technology, People, Organizations, Macroenvironmental factors framework. Results conducted 39 interviews. Clinicians reported VLN be easy little disruption workflow. There were differences patterns between experts novice users critically evaluating system recommendations actively compensating for limitations achieve more reliable performance. Patients also viewed positively. contextual variations performance different hospital cases. Implementation challenges included integration information systems, protection, perceived issues surrounding wider sustained adoption, including procurement costs. Discussion Tool was variable, affected by into workflows divisions labor knowledge, as well technical configuration infrastructure. Conclusion The socio-organizational affecting under-researched require attention further research.

Язык: Английский

---

A Human–AI interaction paradigm and its application to rhinocytology

Artificial Intelligence in Medicine, Год журнала: 2024, Номер 155, С. 102933 - 102933

Опубликована: Июль 22, 2024

This article explores Human-Centered Artificial Intelligence (HCAI) in medical cytology, with a focus on enhancing the interaction AI. It presents Human-AI paradigm that emphasizes explainability and user control of AI systems. is an iterative negotiation process based three strategies aimed to (i) elaborate system outcomes through steps (Iterative Exploration), (ii) explain system's behavior or decisions (Clarification), (iii) allow non-expert users trigger simple retraining model (Reconfiguration). exploited redesign existing AI-based tool for microscopic analysis nasal mucosa. The resulting tested rhinocytologists. discusses results conducted evaluation outlines lessons learned are relevant medicine.

Язык: Английский

---

Adaptive Human-LLMs Interaction Collaboration: Reinforcement Learning driven Vision-Language Models for Medical Report Generation

Опубликована: Апрель 23, 2025

Язык: Английский

---

Opportunities for incorporating intersectionality into biomedical informatics

Journal of Biomedical Informatics, Год журнала: 2024, Номер 154, С. 104653 - 104653

Опубликована: Май 10, 2024

Many approaches in biomedical informatics (BMI) rely on the ability to define, gather, and manipulate data support health through a cyclical research-practice lifecycle. Researchers within this field are often fortunate work closely with healthcare public systems influence generation capture have access vast amount of data. informaticists also expertise engage stakeholders, develop new methods applications, policy. However, research policy that explicitly seeks address systemic drivers would more effectively health. Intersectionality is theoretical framework can facilitate such research. It holds individual human experiences reflect larger socio-structural level privilege oppression, cannot be truly understood if these examined isolation. accounts for interrelated nature providing lens which examine challenge inequities. In paper, we propose intersectionality as an intervention into how conduct BMI We begin by discussing intersectionality's history core principles they apply BMI. then elaborate potential stimulate Specifically, posit our efforts improve should five key considerations: (1) oppression shape health; (2) upstream drivers; (3) nuances outcomes groups; (4) problematic power-laden categories assign people society; (5) inform social change.

Язык: Английский

---