Soar Technology, Inc. (SoarTech) and Georgia Tech Research Institute (GTRI) will research and develop a digital assistant ("Ada") that can help NASA experts search for and effectively use scientific and technical information (STI) in their work. Ada can provide cognitive support in high-effort search tasks, like examining details of more sources, and use tasks, like applying search results to produce a work product.
Our team has already created a search tool to help examine and use STI in complex information environments, which will enable us to quickly prototype a digital assistant with search at its center. We propose to transform the search tool into a digital assistant by adding NASA-specific knowledge structures, tailored user interactions that make relevant details within STI accessible, and cognitive support dialogs to help users accomplish work. The increased interactivity and self-improvement are key to make Ada into a research partner rather than a tool.
Enhancing search with cognitive support functions is made possible by innovative machine intelligence -- cognitive modeling that reasons about tasks and user needs, as well as structured machine learning (ML) that uses symbolic knowledge structures to make deep learning more efficient and accurate. The proposal team have worked together and in parallel on related science and technology, allowing Ada to leverage millions of dollars in funded research progress.
Our proposed Phase I research will produce a software prototype that helps examine and use search results, a demonstration of structured ML feasibility, and designs for adding new interactivity and cognitive support. Phase II will implement, deploy, and test a full digital assistant that can build on the knowledge in search products to help the user accomplish a work task with interactive, tailored cognitive support.
Within NASA research centers, Ada could be useful to a lab-based engineer who has a concrete goal to understand the state of the art and needs to explore alternatives, tradeoffs, or lessons learned. In addition, Ada could also support a basic scientist who is expert in her own field but needs to interact with other fields to find connections that enable application. Ada could also be a digital assistant for patent lawyers who need to search across many fields of expertise and communicate effectively with experts to differentiate prior work.
Ada could be offered as a service to academics or industry researchers. As a more capable and interactive system, Ada could also be licensed to large textbook companies or others who offer knowledge, education, or search products. Ada could help meet a need in competitive modern industry to keep a specialized workforce trained in critical thinking using up-to-date information.