Development of Hybridized Interactive AI Solutions for Healthcare

David Meyer, Sarah Foust, Franco Abott
11:15 AM CST

AI solutions can be improved and expanded when human decision-making is included in the process.

AI solutions are often a blunt instrument. Our presentation proposes three ways by which interaction between the user and AI might render it more useful. The first idea describes a method for steering a user’s path through an AI system’s voluminous findings so that only those relevant to each stage of a given task get presented. The second idea proposes querying the user whenever an AI process encounters a stumbling block of competing answers; a decision for which could permit the process to continue rather than forcing an exit – thus delivering nothing. Our third idea concerns querying the user for certain types of information which are either not easily encodable, or were for other reasons never encoded within the vast data upon which the AI was constructed.
Business Value: 
The innovations we are describing would permit AI solutions to be developed for certain types of problems where a fully-automated process would either be too expensive, would require more data than are available, or would otherwise be unfeasible. Such would also help us to enhance the effectiveness of some of nThrive’s current AI processes and improve new ones that we are creating, such as in Charge Capture, Denials, and Oncology Data Management.