T2: Knowledge Discovery

An important aspect of the vision of UC is the possibility, to access knowledge any place and any time. With the rapid growth of web-2.0-applications the amount of knowledge which is self-created by the end users has also increased strongly. For the automation of many applications – e.g. the detection of the user context during the interaction with the aim to adapt the behavior of the search engine or the recommender-system to the context – the knowledge is not well enough structured.

In the project, procedures are developed which can learn from the content, the structure and the ontologies of the user behavior of such systems and fill them with instances. Based on the created knowledge, base mechanisms for personal navigation and search are developed.

In order to recognize the networking with other users methods are developed to extract thematically and geographically defined user groups by the content and the special structure of the systems (e.g. folksonomies) and the geo-location of users.

It also plays a role how the results are presented to the user. Thereby methods of social network analysis which have so far been used only in manual analyses by scientists (especially sociologists) are automated and made usable for the untrained end user.

From the legal point of view these aspects could lead to new risks in data protection that demand new solutions because the data protection law currently protects only the single affected individual.

The effects on the development of trust by different approaches for knowledge exploration in UC-systems are hardly comparable with all known types of IT-based interactions relations. Therefore new opportunities as well as risks could arise which then must  be addressed.

With regard to the assessment of design proposals the lawyers will especially examine, how the analysis of personal user data in communities of preferences and context can be assessed.

To obtain design rules it will be examined how far knowledge exploration can be carried out successfully even in anonymized and pseudo-anonoymized data and how the transparency and acceptance of recommender systems can be carried out by technical and organizational measures. 

To provide the user situational and on an adequate abstraction level with information, it is necessary to estimate his actual status (e.g. informational and motor stress) and hence deduce his needs and possibilities of interactions.

This can be based on the analysis of the user behavior or on the detection with additional sensors (e.g. eye-tracking analyses and heart rate for the above-mentioned kinds of stress).