A New Framework to Predict and Visualize Technology Acceptance: A Case Study of Shared Autonomous Vehicles

TL;DR

The study addresses the challenge of understanding public acceptance of Shared Autonomous Vehicles (SAVs) by moving beyond traditional SEM to a predictive, data-driven framework that combines Random Forests with chord diagram visualizations. It demonstrates that Attitude, Perceived Risk, Perceived Usefulness, Trust, and Perceived Ease of Use are key predictors of SAV adoption, with non-linear relationships captured by the ML model and interpretable via item- and factor-level chord diagrams. The approach yields insights into adopter vs. non-adopter perspectives and provides practical guidance for targeted strategies to improve acceptance, such as emphasizing societal and personal benefits and enhancing reliability and user experience. The framework is positioned as broadly applicable to other technology-acceptance problems, offering a principled way to quantify predictor importance and communicate complex interdependencies to researchers and stakeholders.

Abstract

Public acceptance is critical to the adoption of Shared Autonomous Vehicles (SAVs) in the transport sector. Traditional acceptance models, primarily reliant on Structural Equation Modeling, may not adequately capture the complex, non-linear relationships among factors influencing technology acceptance and often have limited predictive capabilities. This paper introduces a framework that combines Machine Learning techniques with chord diagram visualizations to analyze and predict public acceptance of technologies. Using SAV acceptance as a case study, we applied a Random Forest machine learning approach to model the non-linear relationships among psychological factors influencing acceptance. Chord diagrams were then employed to provide an intuitive visualization of the relative importance and interplay of these factors at both factor and item levels in a single plot. Our findings identified Attitude as the primary predictor of SAV usage intention, followed by Perceived Risk, Perceived Usefulness, Trust, and Perceived Ease of Use. The framework also reveals divergent perceptions between SAV adopters and non-adopters, providing insights for tailored strategies to enhance SAV acceptance. This study contributes a data-driven perspective to the technology acceptance discourse, demonstrating the efficacy of integrating predictive modeling with visual analytics to understand the relative importance of factors in predicting public acceptance of emerging technologies.

Figures (19)

• Figure 1: Comparison of Factor Relationship Visualization: Subfigures \ref{['fig:Lee_model']} and \ref{['fig:Dichabeng_model']} depict previous models by lirui-036 and Dichabeng_Factors_2021, respectively. Subfigure \ref{['fig:cir_plot_compare']} showcases the Chord Diagram derived from the current case study, emphasizing the methodological advancements and findings by applying the proposed framework in this study, as detailed in Section \ref{['sec: Framework']} and \ref{['sec: Methodology']}.
• Figure 2: Proposed Framework for Analyzing and Visualizing Technology Acceptance. This figure outlines a framework for predicting and visualizing the acceptance of new technologies. It encompasses the stages of data collection, Machine Learning analysis, and visualization through chord diagrams, offering a novel approach for acceptance studies across various technological domains.
• Figure 3: Implementation of the Proposed Framework in the SAV Acceptance Case Study. This figure illustrates the practical application of the framework in analyzing and visualizing data specific to the acceptance of SAVs. It highlights the key steps undertaken in the case study, from questionnaire design to advanced data analysis and visualization techniques, to derive insights into SAV acceptance.
• Figure 4: Example of Computing Relative Importance of the Questions and Factors When Predicting BI4. (a) The predictor importance was first computed using the 'vip' function from the best-trained Random Forest regression trees. (b) Then, the relative importance for each predictor item was calculated by computing the weight of each item, ensuring the sum of all predictor items' relative importance is 100%. (c) Next, the relative importance was transformed into the chord diagram, which shows the predictors (start nodes), the target variable (end nodes), and the relative importance of each item (chords). (d) The sum of all items' relative importance was computed to demonstrate the overall contribution of the predictor.
• Figure 5: Demographic and Travel Modes Distribution of the Collected Data.