AFRL launches new ADDS program
On August 28, the Air Force Research Laboratory (AFRL) launched a new broad agency announcement (BAA) for the Algorithm-Derived Decision Support (ADDS) program. “Intent to Propose” is requested by September 6 at 3:00 p.m. EST. Proposal Due Date and Time: October 16, 3:00 p.m. EST. In addition, AFRL will hold an Industry Day on September 20.
The Algorithm-Derived Decision Support (ADDS) Program seeks to develop and judiciously apply sensemaking approaches to established, classified IC crisis scenarios to demonstrate the potential for machine-generated written intelligence products capable of providing timely and relevant decision advantage.
An Industry Day is anticipated to occur on 20 September 2019 in McLean, Virginia. The purpose is to further discuss the intent of the BAA and answer any questions. Questions and answers will be handled verbally during the large group session or individually during limited small group sessions. Questions and answers will be documented and posted to FBO.gov. Limitations: Industry Day will be held at the Secret Level. Foreign nationals are not permitted to attend. For details regarding the Industry Day time, location and to RSVP, contact Anne Cybenko (email@example.com) by 1200 EST on 13 September 2019.
Current approaches for producing and disseminating written intelligence products continue to limit the speed and scale of relaying actionable information and analysis to the Intelligence Community’s (IC’s) principal customers, U.S. policymakers and warfighters. The latency inherent in traditional IC analytic production—processing disparate and often incomplete or contradictory information streams, drafting, editing, and reviewing—may prove catastrophic to U.S. and allied interests, particularly in a fast-moving crisis. Moreover, practical considerations, such as personnel ceilings, govern the number of topics that the IC can cover at any given time.
To circumvent these bottlenecks, a recent IC effort demonstrated the potential for artificial intelligence (AI) techniques to generate in seconds reasonable quality written intelligence products on a number of related topics from a variety of source documents. However, the IC has yet to apply these machine analytic techniques to realistic crisis scenarios.
Analysts within the IC—whose job is to inform and warn policymakers in a timely fashion—would benefit from greater use of AI technologies, once properly designed, vetted, and employed within their discipline. Information retrieval and management consume significant amounts of IC analysts’ time that could be better spent driving the collection of new information and performing in-depth analysis. The persistent growth in available data will only increase the amount of time spent on these information retrieval and management tasks without a Page 4 of 25 As of 28 Aug 2019 significant advancement in automated solutions for written intelligence product generation.
At present, IC analysts have tools that assist them in finding information, yet they must still integrate these myriad information streams largely without technological assistance. Given the IC’s charge to craft timely, insightful, well-written reports, the intelligence production load would particularly benefit from AI techniques that can populate a specified written intelligence product format with relevant content from a variety of sources. Such a capability could ultimately afford IC customers a parallel production avenue capable of sidestepping traditional analysis’ latency and bandwidth issues while still leveraging the IC’s underlying collection structure.
The Office of the Director of National Intelligence (ODNI)/Acquisition, Technology, and Facility’s Science and Technology (S&T) division in 2016 launched the Intelligence Ventures in Exploratory Science and Technology (In-VEST) program with the stated goal of catalyzing disruptive research approaches for addressing select IC needs. Specifically, the In-VEST program initially designed a series of prize challenges to explore opportunities identified through the technology roadmapping efforts of a complementary effort within S&T, the Intelligence Science and Technology Partnership (In-STeP) program.
The first In-VEST challenge, the ODNI-Office of the Undersecretary of Defense Intelligence (OUSD(I)) Xpress Challenge, demonstrated the potential for a sensemaking technology-based production capability. The winning submission was able to parse through approximately 15,000 documents in response to specific questions and generate reasonable-quality, IC-representative written intelligence products in a matter of seconds. Xpress prizes were awarded based on a blind review by ODNI’s Analytic Integrity and Standards (AIS)—the IC’s established body for reviewing written analytic products across the IC. ODNI’s Rating Scale for Evaluating Analytic Tradecraft Standards (RSEATS) served as the basis for evaluating solver submissions.
Supported by ODNI’s Augmenting Intelligence using Machines (AIM) initiative,2 the ADDS program will build on the Xpress Challenge by further developing these techniques and applying them to established IC crisis training scenarios. Specifically, corpora of U.S. Government-provided multi-Intelligence (Geospatial Intelligence, (GEOINT), Human Intelligence (HUMINT), Signals Intelligence (SIGINT), Open Source Intelligence (OSINT), etc.) will be used to drive evolving crisis scenarios, against which ADDS algorithms will be tasked to create a variety of written intelligence products in specified formats. ADDS performer-generated products will be evaluated based on a combination of timeliness (the time required to generate the written product) and quality (performance against how well the responses address the posed question, and their performance against the publicly available Intelligence Community Directive (ICD) 203 standards).
Full information is available here.