For Hybrid Intelligence systems to have effective and appropriate interactions with various users, these systems will need to be sensitive to and aware of the diversity of the users and the perspectives they might have. Knowledge Graphs provide representation formats and mechanisms to represent multiple perspectives yet questions remain how multi-perspective or polyvocal information can be elicited in Hybrid Intelligence scenarios through interaction with various end users.
This research project within the larger Hybrid Intelligence programme is a collaboration between VU and UTwente. The project will be situated in a number of ongoing digital cultural heritage and digital humanities collaborations such as Pressing Matter, InTaVia and Clariah. The PD will be able to collaborate with heritage partners through ongoing collaborations that are maintained in the context of the ICAI Cultural AI lab and similar collaborations spearheaded by UTwente.
Within the project, we will focus on the domain of cultural heritage where several existing Knowledge Graphs describing tangible and intangible heritage data are available. Your research will investigate the effectiveness of
various interaction design patterns to elicit polyvocal annotations of existing objects and enrichments of existing Knowledge Graphs in that domain. Here, the variety and the perspective of the users (cultural, geographical, gender etc) is to be taken into account. The research question on how to elicit such information in an ethical and responsible manner is a crucial part of this research.
If you are passionate about user interaction design patterns, cultural heritage information, and Hybrid Intelligence and curious to join us in this exciting research, then we are looking for you!
Your dutiesIn the project, you will investigate a generic method for eliciting polyvocal Knowledge Graphs enrichments in Hybrid Intelligence scenarios through interaction designs. To this end, graph pattern extraction methods will be employed on existing KGs and -where possible- combined with natural language processing on text to extract some polyvocal knowledge automatically. During the interactions with different groups of users, a system can prompt/suggest enrichments that might be consistent or inconsistent with users' view. This way, we can connect different views to different groups of users. To determine where enrichments or annotations are useful, interaction patterns for user dialogs will need to be established to elicit this information. The research questions therefore are threefold:
- How can graph pattern mining and user context be combined to identify graph enrichment tasks?
- What are effective and ethical interaction patterns to elicit user context and perspective?
- What are effective and ethical interaction patterns to elicit graph enrichments, while retaining context and perspective?
The method will be generic, but should be implemented in a variety of modalities: in a web application, an interactive VR application or an interactive voice response system (IVR). The research will result in the generic patterns, supporting tooling to implement this in the various modalities and evaluations of these two.
Your day-to-day duties will be to:
- write scientific publications and attend international conferences and workshops to present your work
- collaborate with colleagues in and outside the department on shared research challenges
- provide limited supporting educational activities within the computer science department