NSF HDR Ecosystem and the Spatial Web

The NSF HDR Ecosystem Conference 2024 was held at UIUC on Sept 9-12, 2024

NSF Harnessing the Data Revolution (HDR) enables new modes of data-driven discovery that address fundamental questions at the frontiers of science and engineering.   An objective of the conference was to advance a vision of how the HDR ecosystem is an integral part of a broader coherent ecosystem of AI and data-intensive research.  The notes below were prepared for my participation in the Community Engagement Panel of the HDR conference:

• Two themes of my comments during the HDR community engagement panel:

  • My perspective on Community Engagement is the use of open consensus standards in order to bring HDR research results to the broader engineering developments, in particular making the HDR insights available to commercial product developers.  This is of particular importance for AI commercial product development which could benefit from a more scientific grounding.

  • My perspective on HDR comes from the I-GUIDE Institute which is focused on harnessing the geospatial data revolution based on   Space and Time as the organizing principles.

• NSF funded projects - both the current HDR projects as well as previous projects - have produced foundational science that has been the basis of open geospatial standards development.

  • The previous NSF-funded NCGIA conducted research in geographic topology that defined spatial operators that are now included in open standards like OGC's Simple Features standards and other standards that define geospatial queries.

  • I-GUIDE developments including work with the Open Geospatial Consortium (OGC) on interoperability regarding CyberGIS are relevant to open standards development. Results of I-GUIDE involvement in OGC Testbed 19 and 20 could result in HDR research results becoming part of new OGC standards development.

• HDR research results on Artificial Intelligence (AI) and Data Science (DS) as presented in the conference are relevant to the development of the IEEE Spatial Web as led by the Spatial Web Foundation (SWF). 

  • Expanding Space into Hyperspace 

    • AI and DS computational tools for high-dimensional spaces represent and calculate semantics relations in ways not prevviously possible.  During the HDR conference, the A3D3 Institute presented an example of this high-D analysis in  GWAK: Gravitational-Wave Anomalous Knowledge with Recurrent Autoencoders.   

    • Graph networks bring non-coordinate type relationships to the spatial relations between entities of interest.  During the HDR conference, I-GUIDE Telecoupling research results represented economic dependencies between entities and activities located on different parts of the globe using graph networks.

    • Knowledge Representation in different formats brings increased capabilities to AI.  The HDR conference keynote presentation on Knowledge-Guided Machine Learning KGML showed how both scientific knowledge and data in ML frameworks can achieve better generalizability, scientific consistency, and explainability of results. 

    • The IEEE Spatial Web standard provides a definition of Spatial Web Hyperspace that accommodates these expanded notions of space coming from AI and DS.  The HDR research provides additional basis for the specification of hyperspace. Further work in SWF is using category theory to develop a common interface across the various forms knowledge representation in hyperspace.  

  • AI agents engage in activities across time

    • Much of the recent success of AI relates attaching static labels to entities in images.  Similarly, GIS is powerful in representing static features.  Effort is needed to expand geographic modeling to the behavior of dynamic features.  AI Agent development requires a richer modeling and information exchange associated with dynamic entities.

    • AI is developing of methods to examine video and apply labels to the dynamic behavior.  During the HDR conference, the  Imageomics Institute presented results of analyzing baboon activities in the wild and attaching labels of behavior traits.

    • The IEEE Spatial Web standard defines an ontology for autonomous agents.  The multi-layer Spatial Web ontology sits on a base layer of hyperspace including time; next up is a layer of entities that persist through time such as geographic features; then the ontology includes a layer of dynamic agents, activities and contracts; and capping off with an overarching layer of governance and norms.  The HDR research regarding AI agents and behavior traits is relevant to maturing the Spatial Web ontology.

For additional information on the Spatial Web, see: AI is changing how we think about Space: Hyperspace in the IEEE P2874 Spatial Web