We’re looking to answer a few key questions in our research:
- Which data structures would allow individuals to incorporate their personal knowledge into a decentralized knowledge graph?
- Which behaviors would the interface need to facilitate in order to integrate and maximize the accessibility and iterability of knowledge?
- How can we build upon implicit metadata for a frictionless user experience?
Much of the data on the web is unstructured or poorly structured. While there may be agreed-upon schemas under which people operate, it’s a challenge to get users to input metadata in a consistent way. These pre-existing structures are implicit metadata — the significance of which is revealed through user behavior. For example, the number of times a paper has been cited acts as an implicit measure of support for that particular paper. However, these citation counts miss critical information about whether citing authors supported or opposed the findings, or how claims in the paper play into the area of the discourse graph that authors are to contribute to. If we can identify and understand the rationale behind these agreed-upon structures, we can build upon them in a way that does not feel intrusive to the user and allows them to maximize their potential.
Could the interface to the knowledge graph support positive user behaviors? For example, Twitter’s 280-character limit promotes knowledge compression and unqualified hot takes. We first hope to understand how this would impact an individual user, but a more interesting question comes from the potential of this strategy to disrupt how we conceptualize data and iterate on a global knowledge base. The current body of research is limited by its inability to be queried and updated. Nearly every graduate student has completed a review of the literature at some point, but their labor will likely go unrecognized unless it is published by a journal, leaving future researchers to start at the beginning. Can we create a decentralized knowledge graph that encourages synthesis through its design? Can this be developed while still maintaining an individual user’s privacy?
Our success will be measured by our ability to accurately map out the space, utilize prior research, and understand the techniques and strategies employed by professional knowledge workers and those building tools for thought. We seek to gain an understanding of how implicit metadata already plays a role in these systems, and understand the strategies and searches that underlie these decisions. In the long term, we hope our research will be used in the creation of a collaborative, decentralized knowledge graph.