2021/09/12 by Óscar Gómez, Gomez, Oscar, Steffen Holter +5 · 1 citation
Computer Science · #Artificial Intelligence (cs.AI) #Data Analysis with R #Data Visualization and Analytics #Explainable Artificial Intelligence (XAI) #FOS: Computer and information sciences #H.5 #Human-Computer Interaction (cs.HC) #I.2 #Machine Learning (cs.LG)
paper · pdf · doi:10.48550/arxiv.2109.05629
openalex publication_date 2021/09/12 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Rapid improvements in the performance of machine learning models have pushed\nthem to the forefront of data-driven decision-making. Meanwhile, the increased\nintegration of these models into various application domains has further\nhighlighted the need for greater interpretability and transparency. To identify\nproblems such as bias, overfitting, and incorrect correlations, data scientists\nrequire tools that explain the mechanisms with which these model decisions are\nmade. In this paper we introduce AdViCE, a visual analytics tool that aims to\nguide users in black-box model debugging and validation. The solution rests on\ntwo main visual user interface innovations: (1) an interactive visualization\ndesign that enables the comparison of decisions on user-defined data subsets;\n(2) an algorithm and visual design to compute and visualize counterfactual\nexplanations - explanations that depict model outcomes when data features are\nperturbed from their original values. We provide a demonstration of the tool\nthrough a use case that showcases the capabilities and potential limitations of\nthe proposed approach.\n