Let's go back to the discussion, only mentioned in a previous article, relating to the role of Artificial Intelligence in the field of intelligence analysis.
The question we ask ourselves and which we will try to answer is the following: what is the impact of Artificial Intelligence on intelligence analysis?
First of all, we recall, in favor of all, the definition of intelligence as "the product resulting from the collection, processing, integration, analysis, evaluation and interpretation of available information (and data) concerning foreign countries or regions, or information and knowledge about an opponent obtained through observation, investigation, analysis or understanding".
Of course, this definition has its raison d'etre also in areas other than the confrontation between parties (militarily or politically speaking), in fact it is possible and often useful to make intelligence on a "friendly" element or on oneself, what matters is to define correctly the elements to investigate.
Again to facilitate the discussion, let us recall what is meant by "Artificial Intelligence". Since there is no univocally recognized definition, I will refer to that of Professor John McCarthy, of Stanford University for whom Artificial Intelligence "is the science and engineering of making intelligent machines, especially intelligent computer programs. It is related to the similar task of using computers to understand human intelligence, but AI does not have to confine itself to methods that are biologically observable."
As it is easy to understand, this is not really a definition as to define Artificial Intelligence, it uses the term "intelligence". Therefore, Professor McCarthy therefore adds that by "intelligence" we mean: "the computational part of the ability to achieve goals in the world. Varying kinds and degrees of intelligence occur in people, many animals and some machines", and warns that there is no clear and recognized definition for "intelligence", explaining that when a behavior (human in our case) is well known it is possible to create machines that behave in such a way as to imitate the behavior. However, when there is no clear understanding of the process under consideration, it is difficult to build a machine that mimics human behavior.
It therefore seems logical to me to think that in order to understand how Artificial Intelligence can help in the intelligence analysis process, it is first necessary to understand what this process consists of. Only later will it be possible to identify possible areas in which AI can somehow help.
One of the best known (and used) models in the world of Intelligence is the so-called "Intelligence Cycle", which is based on five phases:
- Planning and Direction. In this phase it is necessary to identify the initial needs in terms of data to be collected and the final intelligence products necessary for decision makers to support them in their decisions. Direction is usually provided by decision makers or government bodies, often in the form of questions.
- Collection. It consists in the collection of raw data and information, necessary to produce intelligence, using all possible sources (among those available or authorized for the specific case). Among the sources generally most used are open sources, but they are not the only ones. Technological development has in fact allowed the collection of data through electronic surveillance devices (sensors), for example through satellite photography or the collection of radio signals or Internet traffic.
- Processing. This phase consists of converting and normalizing non-standard raw data and information into a form that can be used by analysts. For example, the Collection phase may have involved newspapers written in languages not known by the analysts, in this case, in the Processing phase, the translation into one or more known languages must therefore be carried out.
- Analysis and production. In this phase, data and information are transformed into intelligence. The analyst (or rather, analysts), an expert in the sector, must consider the reliability of the information source, its validity and relevance according to the objective (contextualization) and the future implications (to participate in this way to the realization of a partial situational awareness).
- Dissemination. The last phase of the cycle consists in distributing the finished products to those who have requested them (or who need them), in principle these decision makers are the same ones who started the cycle through the initial requests.
Sometimes it happens that decision makers are not yet able to make decisions so they can start a new cycle.
Now, being clearer what the (generic) intelligence cycle consists of, it is possible to try to understand where and how AI can help. I also add that each organization has its own specificities and the intelligence cycle used is not always perfectly identical to the one illustrated. This means that if we want to study how Artificial Intelligence can help a specific organization in the field of intelligence analysis, we must first study their internal processes and verify their intelligence cycle in every detail.
In our case we can say, at first glance, that AI can support the intelligence cycle in the phases of Collection e Processing. At the stage of Collection, AI tools can be used for the selection and identification of data sources and the data to be collected. At the stage of Processing, AI can help in labeling, cataloging and indexing data. The use of systems based on technologies of Machine Learning, which already exist, can be all the more effective the greater the amount of data to be collected, processed and correlated, freeing the operator from performing repetitive and relatively simple tasks, time that can be more usefully employed for true analysis and its own or to improve the preparation of analysts, as also indicated in the Deloitte study "The future of intelligence analysis".
With the use of technologies such as Machine Learning and in particular with the Deep Learning, it is possible to take another step forward. With the Deep Learning it is in fact possible to use AI systems also in the phases of Analysis and production and probably also undergoing Dissemination, particularly. The power of Deep Learning consists in effectively processing and correlating text, images, video and audio without necessarily having to carry out textual conversions. Furthermore the Deep Learning allows us to access predictive capabilities, which as we have seen in the previous article are the last piece of Situational Awareness.
In the next article we will try to understand how.
Alessandro Rugolo, Giorgio Giacinto
To learn more:
- Kwasi Mitchell, Joe Mariani, Adam Routh, Akash Keyal, and Alex Mirkow. The future of intelligence analysis A task-level view of the impact of artificial intelligence on intel analysis. THE DELOITTE CENTER FOR GOVERNMENT INSIGHT. 2019.