Adaptation des interfaces utilisateurs aux émotions

Abstract : User interfaces adaptation by using emotions.Perso2U, an approach to personalize user interfaces with user emotions.User experience (UX) is nowadays recognized as an important quality factor to make systems or software successful in terms of user take-up and frequency of usage. UX depends on dimensions like emotion, aesthetics or visual appearance, identification, stimulation, meaning/value or even fun, enjoyment, pleasure, or flow. Among these dimensions, the importance of usability and aesthetics is recognized. So, both of them need to be considered while designing user interfaces (UI).It raises the question how designers can check UX at runtime and improve it if necessary. To achieve a good UI quality in any context of use (i.e. user, platform and environment), plasticity proposes to adapt UI to the context while preserving user-centered properties. In a similar way, our goal is to preserve or improve UX at runtime, by proposing UI adaptations. Adaptations can concern aesthetics or usability. They can be triggered by the detection of specific emotion, that can express a problem with the UI.So the research question addressed in this PhD is how to drive UI adaptation with a model of the user based on emotions and user characteristics (age & gender) to check or improve UX if necessary.Our approach aims to personalize user interfaces with user emotions at run-time. An architecture, Perso2U, has been designed to adapt the UI according to emotions and user characteristics (age and gender). Perso2U includes three main components: (1) Inferring Engine, (2) Adaptation Engine and (3) Interactive System. First, the inferring engine recognizes the user’s situation and in particular him/her emotions (happiness, anger, disgust, sadness, surprise, fear, contempt) plus neutral which are into Ekman emotion model. Second, after emotion recognition, the best suitable UI structure is chosen and the set of UI parameters (audio, Font-size, Widgets, UI layout, etc.) is computed based on such detected emotions. Third, this computation of a suitable UI structure and parameters allows the UI to execute run-time changes aiming to provide a better UI. Since the emotion recognition is performed cyclically then it allows UI adaptation at run-time.To go further into the inferring engine examination, we run two experiments about the (1) genericity of the inferring engine and (2) UI influence on detected emotions regarding age and gender.Since this approach relies on emotion recognition tools, we run an experiment to study the similarity of detecting emotions from faces to understand whether this detection is independent from the emotion recognition tool or not. The results confirmed that the emotions detected by the tools provide similar emotion values with a high emotion detection similarity.As UX depends on user interaction quality factors like aesthetics and usability, and on individual characteristics such as age and gender, we run a second experimental analysis. It tends to show that: (1) UI quality factors (aesthetics and/or usability) influences user emotions differently based on age and gender, (2) the level (high and/or low) of UI quality factors seem to impact emotions differently based on age and gender. From these results, we define thresholds based on age and gender that allow the inferring engine to detect usability and/or aesthetics problems.
Complete list of metadatas

Cited literature [249 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-02297775
Contributor : Abes Star <>
Submitted on : Thursday, September 26, 2019 - 2:14:06 PM
Last modification on : Saturday, September 28, 2019 - 1:02:44 AM

File

GALINDO_LOSADA_2019_archivage....
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-02297775, version 1

Collections

STAR | LIG | UGA

Citation

Julian Galindo Losada. Adaptation des interfaces utilisateurs aux émotions. Robotique [cs.RO]. Université Grenoble Alpes, 2019. Français. ⟨NNT : 2019GREAM021⟩. ⟨tel-02297775⟩

Share

Metrics

Record views

95

Files downloads

31