Conformity Assessment, Fairness Metrics or Transparency? Users’ Perceptions on Fairness of AI Algorithms

<p>This Dataverse contains datasets associated with the research study <b>"Conformity Assessment, Fairness Metrics, or Transparency? Users’ Perceptions on Fairness of AI Algorithms."</b> The study investigates how different attributes of AI decision-making systems, such as transparency, conformity assessment, and fairness metrics, influence user perceptions of fairness across various decision contexts, including hiring, credit scoring, and recidivism prediction. The data collected through <b>Adaptive Choice-Based Conjoint (ACBC)</b> analysis provides valuable insights into user preferences and the trade-offs they are willing to make between these attributes.</p>

<p><b>The Dataverse includes the following datasets:</b></p>

<ul><li><b>Raw Data</b>: Contains the raw data from the online survey, including response metadata, page time data, context information, survey responses, control variables, and quiz question data.</li>

<li><b>ACBC Counts</b>: Provides detailed counts and summaries of respondent preferences across different attributes and levels, along with individual-level data and level frequencies.</li>

<li><b>ACBC Test Design</b>: Documents the test run of the study design, including information about level appearances, standard errors, warnings, and the d-efficiency of the design.</li>

<li><b>Utility Report Main Analysis</b>: Contains utility analysis results estimated using the Hierarchical Bayes method, including average utility values, importances, and histogram data for both the overall sample and individual respondents.</li>

<li><b>Utility Report Otter's Method</b>: Provides utility analysis results estimated using the Task-Specific Scale Factors Hierarchical Bayes method (Otter’s method) to assess the robustness of the findings, with data similar to the Main Analysis utility report.</li></ul>