A carregar...
Publicação
AutoUsability : previsão da usabilidade de sistemas interativos a partir de sinais fisiológicos
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 5.2 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
To gather information about users’ perceived usability when using interactive applications, standard questionnaires developed for this purpose are generally used. However, these approaches pose some problems that can compromise the data and, in turn, lead to erroneous results. The delay and subjectivity of responses can lead to less rigorous testing due to user saturation, fatigue, or difficulty in understanding certain nuances of the questionnaires. Automating the collection of users’ perceived usability based on their physiological signals would lead to a greater focus on task completion, making the process faster and more enjoyable, and minimising the problems mentioned above. This work investigates the relationship between the usability of the interface perceived by users and their physiological signals while interacting with applications and creates machine learning models to predict usability from this signals. To do this, we used an annotated dataset composed of physiological signals (e.g., EEG, PPG) and results from standard usability questionnaires (e.g., SUS, NASA-TLX), processed the signals, and extracted features. With the features extracted from the physiological signals and the results of the standard questionnaires organised into categories (positive, negative, and neutral), we built machine learning models to make six distinct predictions across four areas (overall perception, usability, UX, and emotional component). The experimental evaluation with users revealed that predictions over time are the ones with the best results, achieving accuracy rates of 59% and 64.6% for general perception and emotional component, respectively. Regarding predictions at the end of the task, the highest accuracy rate (49.4%) belongs to the prediction of the emotional component. These results show that we were able to measure users’ overall perception and emotional response to the applications. On the other hand, predictions of more specific dimensions, such as usability and UX, are not as effective.
Descrição
Tese de Mestrado, Informática, 2025, Universidade de Lisboa, Faculdade de Ciências
Palavras-chave
Usability Physiological Signals User Experience Questionnaires Prediction