Adaptive Interfaces (Computer Science)

—This paper presents a design space for engineering adaptive user interfaces throughout the user interface development life cycle in order to describe any adaptation technique, adaptable or adaptive, to compare two or more techniques, and to generate new, perhaps unprecedented, techniques. Grounded in the theory of psychological perception, this design space structures the adaptation life cycle into two regulation loops between the user and the system: a perception-decision-action (PDA) loop for both the system and the user, and a learning-prediction-adaptation (LPA) for supporting the adaptation, this last being particularly expressive for adaptivity. This PDA-LPA design space enables defining properties for assessing the quality of these loops between the system and the end-user. This design space of is instantiated on two advanced adaptive user interfaces: adaptive user interfaces based on machine learning and adaptive layouts. This design space provides new insights for considering adaptivity design options. Keywords— Adaptation of user interfaces, adaptive user interfaces ; design space exploration; quality properties; intelligent user interfaces.

This paper presents a model-based approach for designing Polymodal Menus, a new type of multimodal adaptive menu for small screen graphical user interfaces where item selection and adaptivity are responsive to more than one interaction modality: a menu item can be selected graphically, tactilely, vocally, gesturally, or any combination of them. The prediction window containing the most predicted menu items by assignment, equivalence, or redundancy is made equally adaptive. For this purpose, an adaptive menu model maintains the most predictable menu items according to various prediction methods. This model is exploited throughout various steps defined on a new Adaptivity Design Space based on a Perception-Decision-Action cycle coming from cognitive psychology. A user experiment compares four conditions of Polymodal Menus (graph-ical, vocal, gestural, and mixed) in terms of menu selection time, error rate, user subjective satisfaction and user preference, when item prediction has a low or high level of accuracy. Polymodal Menus offer alternative input/output modalities to select menu items in various contexts of use, especially when graphical modality is constrained. CCS Concepts: • Human-centered computing → Systems and tools for interaction design; • Human-centered computing → Ubiquitous and mobile devices; • Software and its engineering → Graphical user interface languages KEYWORDS Adaptation, adaptive menu model, model-based approach, polymodal menu, multimodality, prediction window ACM Reference format: Sara Bouzit, Gaëlle Calvary, Denis Chêne, and Jean Vanderdonckt. 2017. Polymodal Menus: A Model-based Approach for Designing Multimodal Adaptive Menus for Small Screens.

This paper presents Cloud Menus, a split adaptive menu for small screens where the predicted menu items are arranged in a circular tag cloud with a location consistent with their corresponding position in the static menu and a font size depending on their prediction level. This layout results from a 3-step design process: (i) defining an initial design space on Bertin's 8 visual variables and 4 quality properties, (ii) identifying the most preferred layout based on agreement rate, and (iii) implementing it into Cloud Menus, a new widget for Android with circular layout. An empirical study suggests that cloud menus reduce item selection time and error rate when prediction is correct without penalizing it when prediction is incorrect, compared to two baselines: a non-adaptive static menu and an adaptive linear menu. From this study, design guidelines for cloud menus are elaborated.

The advent of advanced mobile devices in combination with new interaction modalities and methods for the tracking of contextual information, opens new possibilities in the field of context-aware user interface adaptation. One particular research direction is the automatic context-aware adaptation of input modalities in multimodal mobile interfaces. We present existing adaptive multimodal mobile input solutions and position them within closely related research fields. Based on a detailed analysis of the state of the art, we propose eight design guidelines for adaptive multimodal mobile input solutions. The use of these guidelines is further illustrated through the design and development of an adaptive multimodal calendar application.

This paper presents a design space for exploring design options of adaptive graphical menus based on Bertin's eight visual variables: position, size, shape, value, color, orientation , texture, and motion. In order to transform a traditional (static) menu into an adaptive one, at least one visual variable should be exploited to convey which menu items have been promoted or demoted depending on five characteristics: select , associative, quantitative, order, and length. The paper reviews selected adaptive menu interaction techniques belonging to each of these variables, classifies them according to the variables and characteristics and suggests not yet explored adaptive menu designs. It then defines four stability properties depending on which variables remain constant after adaptivity. A supporting software for prototyping the rendering of graphical adaptive menus is then introduced.

This paper presents two forms of adaptive menus for small devices (smart phones). Contrary to the Ephemeral appearing adaptation proposed by Findlater et al. [7], we claim for disappearing adaptation. The first form is named In Context Disappearing (ICD); the second one Out of Context Disappearing (OCD). The principle of ICD is to display predictive information in a prompting window placed above the main list. The prompting window disappears gradually while maintaining the context always visible and directly accessible. In case of low level prediction, ICD enables user to reach its target without waiting for disappearing effect. OCD principle is almost the same except that the disappearing prompting window covers the full page and thus is out of context like Findlater's approach. Our study shows that for small devices " fading out " a contextual window is better than " fading in ". We demonstrate the benefit of these new forms of adaptation through an experiment with 24 subjects. We conclude that (1) ICD and OCD adaptive lists support faster selection than Control condition when the level of prediction is high, slower in case of bad prediction, and that (2) ICD is faster than OCD in case of bad prediction.

This paper reports on the results of an experiment comparing two graphical adaptive menus for smartphones in order to improve their hierarchical navigation: " Shortcut Menu " and " Step-by-Step Menu " keep constant the actual presentation of initial menus and overlay them with a prediction window displaying the most frequently used menu items, wherever they are located in the hierarchy. In order to reach predicted items, the " Step-by-Step Menu " maintains the consistency with the initial menu through a level-by-level navigation while the " Shortcut Menu " directly moves the end user towards the predicted menu item, thus shortcutting the hierarchical navigation. Thirteen subjects performed fifty tests each on smartphones and data were collected about their item selection time and error rate. The " Step-by-Step Menu " has a positive impact on both variables, whether frequently used menu items are accurately predicted or not. The " Step-by-Step Menu " is fast, but could induce some problems when prediction is wrong.