IARPA peers inside the mind of intelligence analysts
Baku, January 22 (AZERTAC). The US Intelligence Advanced Research Projects Activity (IARPA) is seeking new ways to see how the brain processes conceptual information in a bid to develop new analysis tool and improve training for intelligence analysts and linguists.
The human brain is adept at inferring information using conceptual knowledge, such as the properties of an entity and its relationships to other entities. Given an apple, a human would know it was edible and could be found in orchards and supermarkets, for example.
Current computer systems find this difficult to replicate. Through a Broad Agency Announcement (BAA) called Knowledge Representation in Neural Systems (KRNS), IARPA is soliciting business and academia to develop new ways to better understand how the mind carries this. The agency sees this as a step toward building new analysis tools that acquire, organise and wield knowledge in a more human-like manner.
Neuroscience research uses tools such as brain scans to explain how the brain represents various types of sensory and motor information. However, it has to date fallen short of generating a general predictive theory of the neural basis of conceptual knowledge, beyond high-level concepts such as faces and places.
The KRNS programme seeks to develop and rigorously assess novel theories that explain how the human brain represents diverse types of conceptual knowledge within spatial and temporal (dynamic) patterns of neural activity. IARPA wants responders to develop systems that aim to predict patterns of neural activity associated with particular concepts and that can interpret which concepts are represented within measured patterns of neural activity. All neural activity data in KRNS will be obtained using non-invasive methods including functional magnetic resonance imaging (fMRI) and magnetoencephalography (MEG).
The KRNS programme consists of two phases. Phase 1 seeks to understand how the brain's representation of an individual concept varies as a function of semantic context, for example; how does the neural representation of "apple" differ depending on whether the subject is contemplating "the apple was delicious" versus "the boy threw the apple"?
Phase 2 will explore how combinations of multiple individual concepts are represented in the brain such as; how is the neural representation of a composite concept such as "the doctor drove the car" related to the neural representation of the individual concepts, "doctor," "drove," "car"?
Other news in this section
