Examples of Applications

Our specialists in applied distributed artificial intelligence have developed various network analysis tools, each with its own practical applications. These include the following:

● A tool that reduces human labor in choosing and parametrizing AI algorithms through feedback between artificial intelligence modules. Modules are rated with tests specific to the module and with the tests set by their consumers, among others. For some classes of AI programs (for example, unsupervised algorithms such as clusterers and vector spaces), these multiple weak tests measure effectiveness better than any one strong test.

● A natural language tool to reduce the human labor of putting data into applications. This tool will interpret natural language texts (such as medical research papers) in a way that is both understandable by humans and needed by downstream AIs. It is designed to nudge unsupervised clusterers into a human-designed ontology through seeding with a few exemplars, rather than with the large lists required by supervised learning techniques, using networked relations.

● A tool to test social policy that emulates micro-level social psychological phenomena (such as cognitive dissonance and symbolic interactionism) in AI agents to explore how these micro behaviors create and react to macro social patterns. Treatment policies for social ills are applied to individual agents, where we can observe the effects on agent interactions and explore treatments. This tool has been applied to develop defense strategies against hybrid warfare campaigns that cause polarization in populations and to develop policies that alleviate corruption in societies, and has done so via award-winning analyses.

● A tool to combine the outputs of multiple disparate simulated realities into a single coherent whole using an intelligent fabric of probabilistic ontologies that automate the entry of moves in each reality and run models ahead in a gametree to evaluate the results of moves. This tool was applied to an award-winning analysis of large social systems and is useful for any data fusion application.

● A market-testing tool that incorporates adaptive economic agents. In work conducted for an insurance company, our researchers used this tool to test the effectiveness of payment innovations in increasing the quality of healthcare in America under the Affordable Care Act and to find the best pricing and offerings for new businesses in particular markets, via analyses that include higher-order effects.

● A tool to convert real-world data into a form that can be played as a game and optimized by artificial intelligence techniques. This tool was applied to personalized medicine using healthcare claims data to map out the likely effects of treatments, combining the accuracy of deep neural networks with the ability of epidemiological applications to tease out causal links in data.