User experience (UX) evaluation investigates how people feel about using products or services and is considered an important factor in the design process. However, there is no comprehensive UX evaluation method for time-continuous situations during the use of products or services. Because user experience changes over time, it is difficult to discern the relationship between momentary UX and episodic or cumulative UX, which is related to final user satisfaction. This research aimed to predict final user satisfaction by using momentary UX data and machine learning techniques. The participants were 50 and 25 university students who were asked to evaluate a service (Experiment I) or a product (Experiment II), respectively, during usage by answering a satisfaction survey. Responses were used to draw a customized UX curve. Participants were also asked to complete a final satisfaction questionnaire about the product or service. Momentary UX data and participant satisfaction scores were used to build machine learning models, and the experimental results were compared with those obtained using seven built machine learning models. This study shows that participants’ momentary UX can be understood using a support vector machine (SVM) with a polynomial kernel and that momentary UX can be used to make more accurate predictions about final user satisfaction regarding product and service usage.

Effective development of virtual reality (VR) applications is heavily reliant on the evaluation of user experience (UX). However, traditional methods such as questionnaires have inherent limitations which hinder their ability to capture nuanced behavioral responses and impede the agility of VR content creation: They are time-consuming, burdensome for users, and require significant human effort for interpretation. This study introduces an automated framework aimed at addressing the limitations of questionnaire-based evaluations to assess UX in VR. Our primary focus is to validate the concept of this framework through assessment of the sense of presence (SOP), a crucial psychological perception with significant impact on VR UX. Our proposed framework utilizes a deep neural network (DNN) to analyze patterns in multimodal behavioral cues, including facial expressions, head movements, and hand movements, to predict scores from the Igroup Presence Questionnaire (IPQ). Additionally, we introduce two statistical profiles: the Visual Entropy Profile (VEP), which offers insights into visual complexity by depicting scene entropy, and the Experiential Presence Profile (EPP), which is designed to capture users’ historical SOP levels to enable personalized baseline and sensitivity estimation. The proposed framework achieves a significant correlation between actual and predicted IPQ scores, with a Spearman’s rank correlation coefficient of 0.7303, showcasing the potential of DNNs in analyzing complex behavioral signals and automating SOP assessment. This study represents a pioneering effort in leveraging DNNs for the automatic assessment of SOP and paves the way for future advances in automatically assessing VR UX and unlocking new opportunities in the field.

In user experience evaluation (UXE), it is generally accepted that the order in which users perform tasks when using a product is often random rather than fixed. UXE based on these so-called randomly ordered tasks is challenging. Although several articles have been published on UXE, none have proposed a technique to evaluate the significance of randomly ordered tasks. In this study, we propose a new approach to predict final user satisfaction based on UX related to randomly ordered tasks. We aimed to study the importance of task order in the UX. In the main experiment, 60 participants completed questionnaires about satisfaction while performing a series of tasks on a travel agency website. Among the machine learning models tested, we found that accounting for the order or sequence of actions actually performed by users in a support vector machine (SVM) algorithm with a polynomial kernel produced the most accurate predictions of final user satisfaction (97%). These findings indicate that some machine learning techniques can comprehend participants’ randomly ordered UX data. Moreover, using random ordering, which accounts for the actual order of actions performed by users, can significantly impact the prediction of final user satisfaction.

Virtual Reality (VR) technology offers immersive experiences across various domains, but assessing user experience (UX) remains a challenge. Traditional methods, such as questionnaires, are time-consuming and fail to capture the complexity of user interactions, while automated approaches which leverage sensors and deep learning (DL) techniques show promise but have several architectural limitations. Existing frameworks often rely on data from contact-based physiological sensors. They also tend to overlook attributes of the visual stimuli, use simplistic multimodal fusion techniques, and focus on a single UX dimension, all of which reduce their ability to provide a comprehensive and accurate assessment of UX. This study proposes MUXAS-VR, a DL-based framework that overcomes these limitations by leveraging multimodal behavioral cues. These are seamlessly captured using embedded sensors in commercially available VR devices, eliminating the need for contact-based physiological sensors. MUXAS-VR integrates individual behavioral data with visual attributes, such as motion dynamics and visual complexity, to account for individual interactions and visual influences. Its Multimodal Fusion Module (MFM) selectively combines data from multiple modalities to enhance feature representation and predictive performance. Unlike conventional systems, MUXAS-VR evaluates multiple UX dimensions—cybersickness (CS), sense of presence (SOP), and overall satisfaction (OS)—to provide a comprehensive view of VR UX. To validate its effectiveness, we proposed a rigorous evaluation strategy across three scenarios—Missing Content, New Content, and New User—to ensure generalizability and practical applicability. Experimental results showed that multimodal behavioral cues significantly improved predictive performance. Combining these cues with visual attributes provided complementary information, and the MFM outperformed simpler fusion methods by capturing complex interdependencies. These comprehensive evaluations across diverse scenarios demonstrate MUXAS-VR’s usefulness in predicting multiple UX dimensions.