Directive Explanations for Monitoring the Risk of Diabetes Onset: Introducing Directive Data-Centric Explanations and Combinations to Support What-If Explorations

Abstract

Explainable artificial intelligence is increasingly used in machine learning (ML) based decision-making systems in healthcare. However, little research has compared the utility of different explanation methods in guiding healthcare experts for patient care. Moreover, it is unclear how useful, understandable, actionable and trustworthy these methods are for healthcare experts, as they often require technical ML knowledge. This paper presents an explanation dashboard that predicts the risk of diabetes onset and explains those predictions with data-centric, feature-importance, and examplebased explanations. We designed an interactive dashboard to assist healthcare experts, such as nurses and physicians, in monitoring the risk of diabetes onset and recommending measures to minimize risk. We conducted a qualitative study with 11 healthcare experts and a mixed-methods study with 45 healthcare experts and 51 diabetic patients to compare the different explanation methods in our dashboard in terms of understandability, usefulness, actionability, and trust. Results indicate that our participants preferred our representation of data-centric explanations that provide local explanations with a global overview over other methods. Therefore, this paper highlights the importance of visually directive data-centric explanation method for assisting healthcare experts to gain actionable insights from patient health records. Furthermore, we share our design implications for tailoring the visual representation of different explanation methods for healthcare experts.

EXMOS: Explanatory Model Steering through Multifaceted Explanations and Data Configurations

“…Over the past decade, the field of XAI has witnessed numerous studies being conducted to measure the efficacy of various explanation methods for increasing the transparency of ML systems across multiple application domains, such as healthcare [8,16,18,60], finance [9,13,24], and law enforcement [72,79,87]. Along with making black-box ML models more transparent, XAI methods have also aimed to make these systems more understandable and trustworthy [8,49,54]. Explanation methods have been categorised as model-specific or model-agnostic based on the degree of specificity [2,5].…”

“…Data-centric explanations, on the other hand, aim to find insights from the training data to justify the behaviour of prediction models [5]. Recent works have shown that data-centric explanations can justify the failure of ML models by revealing bias, inconsistencies and quality of the training data [5,7,8]. Examples of data-centric explanation approaches include summarisation of the training data using descriptive statistics, disclosing the bias in training data by showing the distribution of the data across various demographic parameters and revealing the potential issues that can impact the data quality [5,7,8].…”

“…Previous studies have shown that local data-centric explanations are more effective than local model-centric explanations for increasing the trust and understandability of prediction models by justifying predicted outcomes with reference to the training data [8,21]. However, these studies have focused on the efficacy of these explanations solely in the context of prediction justification of individual data instances (i.e., local explanations) rather than the working of the whole model (i.e., global explanations).…”

“…In our implementation, the feature importance scores were generated using the Python SHAP module [57]. The design principles of Bhattacharya et al [8] can be followed to distinguish between actionable and non-actionable features while presenting the feature importances.…”

Towards Directive Explanations: Crafting Explainable AI Systems for Actionable Human-AI Interactions

“…Research Study 1 -In his first research study [7], the author designed and developed an explanation dashboard for monitoring the risk of diabetes onset. This interactive dashboard provides explanations for a diabetes risk prediction model by combining data-centric, feature importance and example-based local explanations.…”