My research in this field focuses on the development of frameworks and interfaces for advanced mixed-initiative interaction with ambient intelligence environments. In such environments, people face increasing difficulty with the complexity of interaction and hence require support in order to use various services of interest. Researches in human-environment interaction aim to address this issue. In general, Human-Environment Interaction is divided into two classes: explicit and implicit interaction. In implicit interaction mode, users do not determine directly when actions occur as events automatically lead the system to perform certain operations without their direct intervention. While this has some applications, in general people want to be always in control. Furthermore, due to the imprecision of sensing and context processing and the over-automation phenomenon, full-automated solutions can lead to user’s distrust and dissatisfaction. Explicit Interaction, where the user initiates all actions, bypasses the challenges of implicit interaction but typically confronts users with the full scale of the environment’s complexity and thus users become overloaded and distracted from their real tasks.
A mixed-initiative interaction paradigm, which mediates between the implicit and the explicit interaction, can therefore play a key role in this situation. On one hand, it allows the user to stay in the control loop because there is always a possibility to intervene via the explicit assistant system, while on the other hand it supports the user by the implicit system. Thus, the benefits of automation will remain while eliminating the problem of over-automation. Naturally, due the concurrent nature of this mixed-initiative approach, interaction conflicts can occur when performing unwanted and inappropriate actions, simultaneously accessing both interaction mechanisms, or executing operations with opposite semantics. To avoid such conflicts, my research proposes mechanism to coordinate and synchronize the use of implicit and explicit interaction. In addition, in order to reduce the cases of miss-automation caused by imprecise context sensing, the proposed mechanism suggests to dynamically adjusting the level of output automation based on the quality of context information. In such approach, the system also suggests users appropriate services or display related information instead of invoking them automatically in the case of lower context quality.
Overall, in order to increase user satisfaction and trust in ambient intelligence environments, my research focuses on the following:
- Direct control of smart environments, model and assess 3D based explicit interaction system and adaptive navigation metaphors to overcome the problem of intuitive device selection in unknown environments;
- Transparency and predictability, ubiquitous meta user interfaces (Ubi-MUI), visualization and animation of sensing activities, decisions, and implicit interactions of ambient intelligent systems, which allows users to understand and predict the behavior of the system.
- Mixed-initiative interaction with smart environments
- Model and assess a mixed-initiative system to interact with Ambient Intelligence environments;
- Provide metaphors to prevent and/or correct conflicting access to smart environments such as undoing or stopping wrong automatics;
- Provide coordination algorithms for distributed conflict management, mechanisms that allow users an explicit interaction with smart environments.
- Provide mechanisms for dynamically adjusting the level of automation or assistance in the environment based on the quality of context information.