Adaptive Systems

Home


Adaptive Systems

Contact person: Alexandros Paramythis
 

Our current work on adaptive systems falls into the following main categories:

You may also want to have a look at our ongoing projects in this area.

 

Areas of work

Adaptive support for (collaborative) eLearning

There is a recently completed EU-funded project in this area (Adaptive Learning Spaces -- see below for more information), as well as an ongoing FWF-funded project that started on January 1st 2008 (Adaptive Support for Collaborative eLearning -- details also below).
 

Meta-adaptation and self-regulation


Self-regulation is a form of meta-adaptivity and, in simplified terms, it refers to an adaptive system's capacity to improve its adaptive behaviour through learning. For a relatively complete introduction to the field, you may want to refer to this paper:

  • Paramythis, A. (2004). Towards Self-Regulating Adaptive Systems. In Weibelzahl, S., & Henze, N. (Eds.), Proceedings of the Annual Workshop of the SIG Adaptivity and User Modeling in Interactive Systems of the German Informatics Society (ABIS04), Berlin, October 4-5 (pp. 57-63).

For an example of ways in which meta-adaptivity, and self-regulation in particular, can affect our current approaches to designing and implementing adaptive systems, you can have a look at the papers below accepted for publication in AH2006. Actually, the first of the two papers, accepted for presentation in the main track of the conference, is a short version of the second, accepted for publication in the joint A3H and A3EH workshop.

Work in this area is moving along two axes: (a) development of the formal theoretical basis for self-regulation, and (b) implementation and integration of self-regulatory bahaviour in an adaptattion "engine".
 

Development of a generic hypermedia adaptation engine

We are working on a hypermedia adaptation that is intended to be applicable to any web-based system that has an XML "pipeline", or has an XML-based representation of the document (fragments) to be returned to the user, at any stage of the request-response cycle. The engine supports adaptation "performatives", which can be thought of as adaptation "verbs" that encapsulate adaptation techniques at different levels of granularity, or manipulate the system's adaptation models.

The engine defines programmatic interfaces (based on pre-defined, yet flexible information flows) through which different modelling and decision-making components can be plugged in. Standard implementations of these components will be made available, including: (a) a simple user modelling component, supporting overlay modelling (with respect to the application's domain model(s)), as well as independent user properties; (b) a simple rule-based decision-making component; and, (c) a rather more advanced logic-based decision-making component. Alternative implementations are, of course, possible. Provided that a small set of prerequisites is met, there is also the possibility to create "adapters" that enable existing components to be plugged into the engine without modifications.

Semantic web technologies are used for the ontological specification of the application's domain model, as well as for "linking" together adaptation models, adaptation logic and adaptation performatives.

Perhaps the most important feature of the engine is a number of features intended to facilitate the implementation of self-regulating adaptive systems (i.e., systems that can modify their adaptive behaviour through "learning" based on self-evaluation). The system's self- evaluation and regulation principles will be specified in a declarative manner. Naturally, these features will depend heavily upon (and would, therefore, constrain) the approaches used in creating the dynamic adaptation models (especially the user model) and in drawing adaptation decisions.

In summary, the goals of the ongoing work are: to create an adaptation engine that can be applied to multiple application domains; to support the declarative specification of a wide range of adaptive system behaviours; to provide preliminary support for creating self-regulating adaptive systems; and, to enable the integration of alternative adaptation components. Finally, an overarching objective in the implementation of the engine is to make it orthogonal to common web application architectures, so that the only requirement for integrating it in existing systems is the one stated above: have an XML representation (with XHTML and even HTML also being possible) of the "served" document, at any stage of the request-response cycle.
 

Evaluation of adaptive systems

Our work on the evaluation of adaptive systems focuses mainly on the perspectives of deriving design feedback, and of validating common adaptive methods and techniques in different application domains. In this context we have been collaborating with Dr. Stephan Weibelzahl of The National College of Ireland towards a framework that addresses the specific problems one encounters when evaluating an adaptive system. The main idea behind the framework is identifying the different "layers" of the adaptation process, and devising methodological approaches that would allow evaluators to address each of these layers separately or in combination, depending on their evaluation goals and constraints. Some of the first outcomes of our work have been recorded in:

EASy-Hub logoAgain with Dr. Stephan Weibelzahl we have also established and are hosting EASy-Hub, a site devoted to the subject of evaluating adaptive systems. Our main goals in establishing the site were, firstly, to create a central "repository" of information and knowledge on the subject, and, secondly, to provide an interactive space to which others can contribute.

Finally, our institute has participated in the organisation of a series of workshops on "User-Centred Design and Evaluation of Adaptive Systems", with five installments thus far, as well in the organisation of the tutorial "Formative Evaluation Methods for Adaptive Systems" held in conjunction with the 11th International Conference on User Modeling (UM 2007) (the tutorial slides are available from this page).

 

Personalised searching

We are working on the Prospector system, which is a "meta-search" engine, acting as a front end to popular search engines (such Google and Yahoo), and uses individual- and group- based models to personalise search results. A brief description of the first generation of the system can be found in the following paper:

 

Recently completed and ongoing Projects

"Adaptive Learning Spaces" (ALS) (229714-CP-1-2006-1-NL-MPP) was a European collaborative research project, funded under the Socrates - Minerva Programme , overseen by the European Commission's Education, Audiovisual and Culture Agency Executive Agency (EACEA). The ALS project started on October 1st 2006, and was completed on March 31st 2008. For additional information please refer to the project's web site: http://www.als-project.org/
 

"Adaptive Support for Collaborative e-Learning" (ASCOLLA), funded by the Austrian Science Fund (Fonds zur Förderung der wissenschaftlichen Forschung FWF) is a research project undertaken by the FIM Institute. The project started on January 1st 2008 and will be completed in September 2010. The goal of the project is to provide the technological means through which a lack (or limited amounts) of face-to-face contact between learners in e-Learning can be partially compensated for, towards the goal of supporting, enhancing and facilitating on-line collaborative learning. For additional information please refer to the project's web site: http://ascolla.fim.uni-linz.ac.at


 


Last modified by Alexandros Paramythis  2009-12-05