Knowledge Elicitation: Data Collection and Knowledge Modeling for Human-Aware Systems

Abstract

Knowledge elicitation systems are important for utilizing expert domain knowledge in human-machine systems. These systems acknowledge the role that expert knowledge plays in structuring and utilizing information effectively. And yet, there remains the issue of how to “captur[e] the important information that sufficiently represents the real-world domain.” Central to this issue is the characterization of knowledge beyond the systems of data collection and knowledge modeling. We turn to feminist epistemology to approach the notion of knowledge from a critical perspective that challenges normative representational ways of knowing. Feminist theories support generative methods that influence the futuring of knowledge-driven AI systems. In this workshop, we put forward a speculative feminist knowledge elicitation system to challenge normative assumptions about knowledge. Through feminist epistemology, we consider knowledge from a feminist lens, adding a new, context-sensitive perspective to knowledge elicitation systems: a re-formulation of content, context, and ethics.

*Note. Both researchers contributed equally to this workshop.

Citation

Workshop Description: Knowledge elicitation regards methods and techniques for extracting knowledge from domain experts to create structured descriptions of a domain. It involves analyzing qualitative data, such as interview transcripts, to identify concepts and relations of the domain. An overall aim in knowledge elicitation is to develop common vocabulary and semantics to enable representing, sharing, and reusing the elicited knowledge across human-system boundaries. In the field of Artificial Intelligence (AI), such knowledge models, typically implemented in the form of ontologies, can be integrated into knowledge-based intelligent systems to enable context-aware reasoning and decision-making capabilities.

One of the main challenges of knowledge elicitation is capturing the important information that sufficiently represents the real-world domain. The social sciences and humanities often deal with complex and nuanced information of human domains, enabling insights that capture multi-dimensional views on, e.g., human, societal, socio-economic, and political considerations of the world. Incorporating knowledge elicitation tools and methods in social sciences and humanities, which have well-established methods for qualitative data collection and analysis, can support the development of knowledge formats for understanding and sharing knowledge of, e.g., social structures, cultural practices, and human behaviors across disciplines and human-system boundaries to develop human-aware intelligent systems.

Workshop Link

Resources