Integration of Artificial Intelligence into the US Intelligence System: An Analysis of the Current Situation of AI-Assisted Intelligence Training

1. Standardize the intelligence analysis process

The U.S. intelligence community has long tried to use artificial intelligence to train intelligence personnel. According to public information, the intelligence team of the U.S. Navy Battle Group (BG) needs to conduct at least three training sessions at the Intelligence Team Trainer (ITT) of the Pacific Fleet Intelligence Training Center (FITCPAC). The trainer simulates each intelligence unit on the ship, which can not only be used for the training of intelligence personnel, but also can be used for the exercise of intelligence collection, transmission and analysis of the cooperation process of each intelligence unit.

As an auxiliary and extension of the human brain, artificial intelligence technology can break through the physiological limitations of the human brain and significantly improve the scientificity and efficiency of intelligence personnel's analysis and synthesis. Strengthening the specialization and standardization of analysts' analysis process is an effective means to improve intelligence analysis work. In reality, analysts are recruited, trained and used according to their specialties. It is difficult for those engaged in signal intelligence analysis to engage in human intelligence analysis. It is extremely difficult for analysts to fully expand from their familiar fields to other fields.

Artificial intelligence technology can integrate elements of unpredictability into training simulations, so it has special value for intelligence analysis training. In recent years, the U.S. intelligence community has developed a variety of artificial intelligence software designed to assist in the training of intelligence personnel. Research institutions led by the Intelligence Advanced Research Projects Agency (IARPA) provide immersive and realistic virtual experiences for intelligence personnel by developing software such as military games and virtual simulation environments. The CREATE program being developed by IARPA is a system for developing structured analysis techniques to improve analytical reasoning. The program is designed to train intelligence officers to better understand the evidence and assumptions that support or conflict with conclusions, while also helping users better communicate their reasoning and conclusions.

The training algorithm of artificial intelligence is generated based on repeated training of previous major events, massive data sets, and expert experience. Although it lacks critical thinking and innovation in unconventional event simulation, its ability to analyze events in traditional models far exceeds that of manpower. Therefore, the use of artificial intelligence training algorithms to assist intelligence personnel training can eliminate the understanding bias caused by human factors to a certain extent, making the analysis process of intelligence analysts more scientific and standardized.

2. Auxiliary psychological and behavioral training

Intelligence analysis is a highly complex thinking process, so intelligence analysis is still essentially the thinking activity of intelligence analysts. With the reflection and investigation of intelligence mistakes by the American intelligence community, the process of human psychological cognition has gradually become the key direction of intelligence analysis theory research. American intelligence theorists represented by Richards J. Heuer Jr. Believe that the "psychological decision-making mechanism" of intelligence analysts is limited by their personal cognitive ability, which will have a significant impact on the results of intelligence analysis. However, people's rationality is affected by the objective world they have been in contact with for a long time, and once it is formed, it is often difficult to change. This requires standardization and stimulation of a certain intensity to regulate its cognitive model and avoid cognitive bias as much as possible.

At present, the U.S. intelligence community is also exploring the use of modeled artificial intelligence and machine learning software to train analysts and help them recognize and avoid misunderstandings. Based on the theories of cognitive psychology and behavioral science, the MOSAIC project led by IARPA uses multi-modal sensing to continuously monitor changes in analysts and their surrounding environment, and finally establishes a comprehensive model to evaluate the analytical capabilities of analysts. Another IARPA project, KRNS, which studies mental behavior, has also taken a big step towards understanding how the human brain represents conceptual knowledge. It acquires, organizes and applies existing knowledge at high speed, and finally forms novel techniques for training intelligence analysts and linguists. It can be seen that the US intelligence community has achieved certain results in the research on the use of artificial intelligence to assist intelligence training. Through machine learning and repeated interpretation of a large number of historical events, artificial intelligence software can also propose alternatives to a certain extent to test or standardize the analysis process of analysts. Artificial intelligence has gradually shifted from simple simulation and data visualization to exerting an impact on human cognition and psychology.