Explaining Model Confidence Using Counterfactuals

Authors: Thao Le, Tim Miller, Ronal Singh, Liz Sonenberg

AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We therefore evaluate our explanation to know whether counterfactual explanations can improve understanding, trust, and user satisfaction in two user studies using existing methods for assessing understanding, trust and satisfaction.
Researcher Affiliation Academia School of Computing and Information Systems, The University of Melbourne thaol4@student.unimelb.edu.au, {tmiller, rr.singh, l.sonenberg}@unimelb.edu.au
Pseudocode No The paper describes algorithms conceptually but does not provide structured pseudocode or an algorithm block.
Open Source Code No The paper refers to a third-party tool's website (Gurobi Optimizer) but does not provide specific access to the authors' own source code for the described methodology.
Open Datasets Yes The data used for the income prediction task is the Adult Dataset published in UCI Machine Learning Repository (Dua and Graff 2017) that includes 32561 instances and 14 features. In the second domain, we use the IBM HR Analytics Employee Attrition Performance dataset published in Kaggle (Pavansubhash 2017), which includes 1470 instances and 34 features.
Dataset Splits No The paper describes the datasets used and the selection of features, but it does not specify the training, validation, or test dataset splits (e.g., percentages or counts) used for the machine learning models.
Hardware Specification No The paper does not explicitly describe the specific hardware (e.g., GPU/CPU models, memory) used to run the experiments.
Software Dependencies No The paper mentions using 'Gurobi Optimization' but does not specify a version number for this or any other software dependency.
Experiment Setup No The paper describes the choice of logistic regression model and details of the human-subject experiment setup, but it does not specify concrete hyperparameters or system-level training settings for the models (e.g., learning rate, batch size, epochs).