What I Cannot Predict, I Do Not Understand: A Human-Centered Evaluation Framework for Explainability Methods.
Ontology highlight
ABSTRACT:
SUBMITTER: Colin J
PROVIDER: S-EPMC10544769 | biostudies-literature | 2022
REPOSITORIES: biostudies-literature
Publications
What I Cannot Predict, I Do Not Understand: A Human-Centered Evaluation Framework for Explainability Methods.
Colin Julien J Fel Thomas T Cadène Rémi R Serre Thomas T
Advances in neural information processing systems 20220101
A multitude of explainability methods has been described to try to help users better understand how modern AI systems make decisions. However, most performance metrics developed to evaluate these methods have remained largely theoretical - without much consideration for the human end-user. In particular, it is not yet clear (1) how useful current explainability methods are in real-world scenarios; and (2) whether current performance metrics accurately reflect the usefulness of explanation method ...[more]
Similar Datasets
Project description:Summary Rapid advances in artificial intelligence (AI) and availability of biological, medical, and healthcare data have enabled the development of a wide variety of models. Significant success has been achieved in a wide range of fields, such as genomics, protein folding, disease diagnosis, imaging, and clinical tasks. Although widely used, the inherent opacity of deep AI models has brought criticism from the research field and little adoption in clinical practice. Concurrently, there has been a significant amount of research focused on making such methods more interpretable, reviewed here, but inherent critiques of such explainability in AI (XAI), its requirements, and concerns with fairness/robustness have hampered their real-world adoption. We here discuss how user-driven XAI can be made more useful for different healthcare stakeholders through the definition of three key personas—data scientists, clinical researchers, and clinicians—and present an overview of how different XAI approaches can address their needs. For illustration, we also walk through several research and clinical examples that take advantage of XAI open-source tools, including those that help enhance the explanation of the results through visualization. This perspective thus aims to provide a guidance tool for developing explainability solutions for healthcare by empowering both subject matter experts, providing them with a survey of available tools, and explainability developers, by providing examples of how such methods can influence in practice adoption of solutions. The bigger picture Adoption of AI tools in practical settings, such as for research/clinical tasks, has been hampered by a lack of transparency/interpretability of the models. After performing a review of different types of AI explainability (XAI) methods developed to better understand the predictions made by a model, we also develop a taxonomy to better classify the different approaches. We think that these XAI techniques are not sufficient to enhance practical implementations and illustrate via an example how user-driven XAI can be useful for different stakeholders in the healthcare domain. We identify and define three key personas involved in healthcare—data scientists, clinical researchers, and clinicians—and present an overview of the different approaches that can address their needs. The ultimate goal of adopting AI in medical practice and patient care goes beyond explainability and will need the development of extra layers of security and confidence, in particular regarding AI trustworthiness, as XAI transparent systems become prone to attacks that may reveal confidential information, and AI fairness, as systems developed and tested in diverse environments need to be expanded to real-world situations. A significant amount of research has focused on making artificial intelligence (AI) techniques more interpretable and post hoc explanations more accessible. This has been achieved through the development of a wide variety of so-called explainability methods. We here propose a taxonomy and review the most popular AI explainability (XAI) methods, illustrated with examples from healthcare and life sciences. We also propose that user-centered explainability is necessary for further practical adoption of AI approaches and illustrate this using healthcare-centered user examples.
| S-EPMC9122967 | biostudies-literature
Project description:ObjectiveThe aim of this study is to examine patients' experiences in integrated care (IC) settings.DesignQualitative study using semistructured interviews.SettingsTwo IC sites in Toronto, Canada: (1) a community-based primary healthcare centre, supporting patients with hepatitis C and comorbid mental health and substance use issues; and (2) an integrated bariatric surgery programme, an academic tertiary care centre.ParticipantsThe study included patients (n=12) with co-occurring mental and physical health conditions. Seven participants (58%) were female and five (42%) were male.MethodsTwelve indepth semistructured interviews were conducted with a purposeful sample of patients (n=12) with comorbid mental and physical conditions at two IC sites in Toronto between 2017 and 2018. Data were collected and analysed using grounded theory approach.ResultsFour themes emerged in our analysis reflecting patients' perspectives on patient-centred care experience in IC: (1) caring about me; (2) collaborating with me; (3) helping me understand and self-manage my care; and (4) personalising care to address my needs. Patients' experiences of care were primarily shaped by quality of relational interactions with IC team members. Positive interactions with IC team members led to enhanced patient access to care and fostered personalising care plans to address unique needs.ConclusionThis study adds to the literature on creating patient-centredness in IC settings by highlighting the importance of recognising patients' unique needs and the context of care for the specific patient population.
| S-EPMC7389773 | biostudies-literature
Project description:BackgroundContinuous monitoring of vital signs is critical for ensuring patient safety in intensive care units (ICUs) and is becoming increasingly relevant in general wards. The effectiveness of health information technologies such as patient-monitoring systems is highly determined by usability, the lack of which can ultimately compromise patient safety. Usability problems can be identified and prevented by involving users (ie, clinicians).ObjectiveIn this study, we aim to apply a human-centered design approach to evaluate the usability of a remote patient-monitoring system user interface (UI) in the ICU context and conceptualize and evaluate design changes.MethodsFollowing institutional review board approval (EA1/031/18), a formative evaluation of the monitoring UI was performed. Simulated use tests with think-aloud protocols were conducted with ICU staff (n=5), and the resulting qualitative data were analyzed using a deductive analytic approach. On the basis of the identified usability problems, we conceptualized informed design changes and applied them to develop an improved prototype of the monitoring UI. Comparing the UIs, we evaluated perceived usability using the System Usability Scale, performance efficiency with the normative path deviation, and effectiveness by measuring the task completion rate (n=5). Measures were tested for statistical significance using a 2-sample t test, Poisson regression with a generalized linear mixed-effects model, and the N-1 chi-square test. P<.05 were considered significant.ResultsWe found 37 individual usability problems specific to monitoring UI, which could be assigned to six subcodes: usefulness of the system, response time, responsiveness, meaning of labels, function of UI elements, and navigation. Among user ideas and requirements for the UI were high usability, customizability, and the provision of audible alarm notifications. Changes in graphics and design were proposed to allow for better navigation, information retrieval, and spatial orientation. The UI was revised by creating a prototype with a more responsive design and changes regarding labeling and UI elements. Statistical analysis showed that perceived usability improved significantly (System Usability Scale design A: mean 68.5, SD 11.26, n=5; design B: mean 89, SD 4.87, n=5; P=.003), as did performance efficiency (normative path deviation design A: mean 8.8, SD 5.26, n=5; design B: mean 3.2, SD 3.03, n=5; P=.001), and effectiveness (design A: 18 trials, failed 7, 39% times, passed 11, 61% times; design B: 20 trials, failed 0 times, passed 20 times; P=.002).ConclusionsUsability testing with think-aloud protocols led to a patient-monitoring UI with significantly improved usability, performance, and effectiveness. In the ICU work environment, difficult-to-use technology may result in detrimental outcomes for staff and patients. Technical devices should be designed to support efficient and effective work processes. Our results suggest that this can be achieved by applying basic human-centered design methods and principles.Trial registrationClinicalTrials.gov NCT03514173; https://clinicaltrials.gov/ct2/show/NCT03514173.
| S-EPMC8957007 | biostudies-literature
Project description:BackgroundThorough holistic development of eHealth can contribute to a good fit among the technology, its users, and the context. However, despite the availability of frameworks, not much is known about specific research activities for different aims, phases, and settings. This results in researchers having to reinvent the wheel. Consequently, there is a need to synthesize existing knowledge on research activities for participatory eHealth development processes.ObjectiveThe 3 main goals of this review are to create an overview of the development strategies used in studies based on the CeHRes (Center for eHealth Research) Roadmap, create an overview of the goals for which these methods can be used, and provide insight into the lessons learned about these methods.MethodsWe included eHealth development studies that were based on the phases and/or principles of the CeHRes Roadmap. This framework was selected because of its focus on participatory, iterative eHealth design in context and to limit the scope of this review. Data were extracted about the type of strategy used, rationale for using the strategy, research questions, and reported information on lessons learned. The most frequently mentioned lessons learned were summarized using a narrative, inductive approach.ResultsIn the included 160 papers, a distinction was made between overarching development methods (n=10) and products (n=7). Methods are used to gather new data, whereas products can be used to synthesize previously collected data and support the collection of new data. The identified methods were focus groups, interviews, questionnaires, usability tests, literature studies, desk research, log data analyses, card sorting, Delphi studies, and experience sampling. The identified products were prototypes, requirements, stakeholder maps, values, behavior change strategies, personas, and business models. Examples of how these methods and products were applied in the development process and information about lessons learned were provided.ConclusionsThis study shows that there is a plethora of methods and products that can be used at different points in the development process and in different settings. To do justice to the complexity of eHealth development, it seems that multiple strategies should be combined. In addition, we found no evidence for an optimal single step-by-step approach to develop eHealth. Rather, researchers need to select the most suitable research methods for their research objectives, the context in which data are collected, and the characteristics of the participants. This study serves as a first step toward creating a toolkit to support researchers in applying the CeHRes Roadmap to practice. In this way, they can shape the most suitable and efficient eHealth development process.
| S-EPMC8832261 | biostudies-literature
Project description:IntroductionSpiking neural networks (SNNs) are a model of computation that mimics the behavior of biological neurons. SNNs process event data (spikes) and operate more sparsely than artificial neural networks (ANNs), resulting in ultra-low latency and small power consumption. This paper aims to adapt and evaluate gradient-based explainability methods for SNNs, which were originally developed for conventional ANNs.MethodsThe adapted methods aim to create input feature attribution maps for SNNs trained through backpropagation that process either event-based spiking data or real-valued data. The methods address the limitations of existing work on explainability methods for SNNs, such as poor scalability, limited to convolutional layers, requiring the training of another model, and providing maps of activation values instead of true attribution scores. The adapted methods are evaluated on classification tasks for both real-valued and spiking data, and the accuracy of the proposed methods is confirmed through perturbation experiments at the pixel and spike levels.Results and discussionThe results reveal that gradient-based SNN attribution methods successfully identify highly contributing pixels and spikes with significantly less computation time than model-agnostic methods. Additionally, we observe that the chosen coding technique has a noticeable effect on the input features that will be most significant. These findings demonstrate the potential of gradient-based explainability methods for SNNs in improving our understanding of how these networks process information and contribute to the development of more efficient and accurate SNNs.
| S-EPMC10565802 | biostudies-literature
Project description:Historically, two primary criticisms statisticians have of machine learning and deep neural models is their lack of uncertainty quantification and the inability to do inference (i.e., to explain what inputs are important). Explainable AI has developed in the last few years as a sub-discipline of computer science and machine learning to mitigate these concerns (as well as concerns of fairness and transparency in deep modeling). In this article, our focus is on explaining which inputs are important in models for predicting environmental data. In particular, we focus on three general methods for explainability that are model agnostic and thus applicable across a breadth of models without internal explainability: "feature shuffling", "interpretable local surrogates", and "occlusion analysis". We describe particular implementations of each of these and illustrate their use with a variety of models, all applied to the problem of long-lead forecasting monthly soil moisture in the North American corn belt given sea surface temperature anomalies in the Pacific Ocean.
| S-EPMC10187774 | biostudies-literature
Project description:We benchmarked deconvolution algorithms on human brain gene expression data. The data deposited here include A) mixtures on which algorithms were benchmarked, as well as B) signatures of pure brain cell-type expression
2021-08-28 | GSE175772 | GEO
Project description:BackgroundThe silent transmission of COVID-19 has led to an exponential growth of fatal infections. With over 4 million deaths worldwide, the need to control and stem transmission has never been more critical. New COVID-19 vaccines offer hope. However, administration timelines, long-term protection, and effectiveness against potential variants are still unknown. In this context, contact tracing and digital contact tracing apps (CTAs) continue to offer a mechanism to help contain transmission, keep people safe, and help kickstart economies. However, CTAs must address a wide range of often conflicting concerns, which make their development/evolution complex. For example, the app must preserve citizens' privacy while gleaning their close contacts and as much epidemiological information as possible.ObjectiveIn this study, we derived a compare-and-contrast evaluative framework for CTAs that integrates and expands upon existing works in this domain, with a particular focus on citizen adoption; we call this framework the Citizen-Focused Compare-and-Contrast Evaluation Framework (C3EF) for CTAs.MethodsThe framework was derived using an iterative approach. First, we reviewed the literature on CTAs and mobile health app evaluations, from which we derived a preliminary set of attributes and organizing pillars. These attributes and the probing questions that we formulated were iteratively validated, augmented, and refined by applying the provisional framework against a selection of CTAs. Each framework pillar was then subjected to internal cross-team scrutiny, where domain experts cross-checked sufficiency, relevancy, specificity, and nonredundancy of the attributes, and their organization in pillars. The consolidated framework was further validated on the selected CTAs to create a finalized version of C3EF for CTAs, which we offer in this paper.ResultsThe final framework presents seven pillars exploring issues related to CTA design, adoption, and use: (General) Characteristics, Usability, Data Protection, Effectiveness, Transparency, Technical Performance, and Citizen Autonomy. The pillars encompass attributes, subattributes, and a set of illustrative questions (with associated example answers) to support app design, evaluation, and evolution. An online version of the framework has been made available to developers, health authorities, and others interested in assessing CTAs.ConclusionsOur CTA framework provides a holistic compare-and-contrast tool that supports the work of decision-makers in the development and evolution of CTAs for citizens. This framework supports reflection on design decisions to better understand and optimize the design compromises in play when evolving current CTAs for increased public adoption. We intend this framework to serve as a foundation for other researchers to build on and extend as the technology matures and new CTAs become available.
| S-EPMC8919989 | biostudies-literature
Project description:Explainable Artificial Intelligence (XAI) aims to provide insights into the inner workings and the outputs of AI systems. Recently, there's been growing recognition that explainability is inherently human-centric, tied to how people perceive explanations. Despite this, there is no consensus in the research community on whether user evaluation is crucial in XAI, and if so, what exactly needs to be evaluated and how. This systematic literature review addresses this gap by providing a detailed overview of the current state of affairs in human-centered XAI evaluation. We reviewed 73 papers across various domains where XAI was evaluated with users. These studies assessed what makes an explanation "good" from a user's perspective, i.e., what makes an explanation meaningful to a user of an AI system. We identified 30 components of meaningful explanations that were evaluated in the reviewed papers and categorized them into a taxonomy of human-centered XAI evaluation, based on: (a) the contextualized quality of the explanation, (b) the contribution of the explanation to human-AI interaction, and (c) the contribution of the explanation to human-AI performance. Our analysis also revealed a lack of standardization in the methodologies applied in XAI user studies, with only 19 of the 73 papers applying an evaluation framework used by at least one other study in the sample. These inconsistencies hinder cross-study comparisons and broader insights. Our findings contribute to understanding what makes explanations meaningful to users and how to measure this, guiding the XAI community toward a more unified approach in human-centered explainability.
| S-EPMC11525002 | biostudies-literature
Project description:BACKGROUND:Poor EHR design adds further challenges, especially in the areas of order entry and information visualization, with a net effect of increased rates of incidents, accidents, and mortality in ICU settings. OBJECTIVE:The purpose of this study was to propose a novel, mixed-methods framework to understand EHR-related information overload by identifying and characterizing areas of suboptimal usability and clinician frustration within a vendor-based, provider-facing EHR interface. METHODS:A mixed-methods, live observational usability study was conducted at a single, large, tertiary academic medical center in the Southeastern US utilizing a commercial, vendor based EHR. Physicians were asked to complete usability patient cases, provide responses to three surveys, and participant in a semi-structured interview. RESULTS:Of the 25 enrolled ICU physician participants, there were 5(20%) attending physicians, 9 (36%) fellows, and 11 (44%) residents; 52% of participants were females. On average, residents were the quickest in completing the tasks while attending physician took the longest to complete the same task. Poor usability, complex interface screens, and difficulty to navigate the EHR significantly correlated with high frustration levels. Significant association were found between the occurrence of error messages and temporal demand such that more error messages resulted in longer completion time (p?=?.03). DISCUSSION:Physicians remain frustrated with the EHR due to difficulty in finding patient information. EHR usability remains a critical challenge in healthcare, with implications for medical errors, patient safety, and clinician burnout. There is a need for scientific findings on current information needs and ways to improve EHR-related information overload.
| S-EPMC6584062 | biostudies-literature