Artificial Intelligence (AI) and Machine Learning (ML) excel in processing vast amounts of data swiftly. During open-source intelligence (OSINT) investigations, you almost always gather information in overwhelming volumes. AI algorithms can efficiently sift through this data, identifying patterns and relevant information significantly quicker than a human ever could.
For example, let’s say you’ve collected 800 PDF documents, each containing at least 300 pages. This means you have at least 240,000 pages of data. But what if you are only interested in seeing the documents that contain a specific name or phone number, or if that name is associated with a specific location?
No OSINT analyst in his or her right mind (or at least I hope so) would manually comb through each of the 240,000 pages to find the information that they are looking for. AI and ML have the power to analyze thisdata at a super fast scale—they are huge time saving tools that will free you up to focus on the analysis and reporting stages of your investigation. AI can also automate repetitive and mundane tasks, such as data collection, sorting and preliminary analysis. Again, this automation is a huge time saver that enables human analysts to focus on more complex aspects of their investigations.
The OODA loop with AI/ML integration for OSINT
The OODA loop (Observe, Orient, Decide, Act) is a critical framework that encapsulates the decision-making process—it also plays a very important role in OSINT. Shortening this loop is essential for quicker, more effective intelligence analysis, but the key to success is building and maintaining a robust knowledge base. A well-structured and comprehensive knowledge base will empower analysts to quickly observe and pivot, providing instant access to historical data, contextual information and a repository of past analyses. Add AI and ML capabilities to this base, and the potential for deep and insightful analysis increases exponentially. These algorithms can process and analyze data at a scale and speed that no human analyst alone could ever achieve. This enables analysts to ask complex questions and glean insights that they may have missed or would have taken longer to uncover.
This can be especially useful for analysts who may lack expertise in data science or database management, as it allows them to focus on formulating the right questions and interpreting the results, rather than the frustrating technicalities of data processing. This approach also broadens the scope of who can participate in and contribute to the intelligence analysis process. Not everyone on the team needs to be super tech savvy or have programming/data science skills. As analysts engage with the system, they contribute new insights and data back into the knowledge base, thus continuously enriching it. This cyclical process of learning and updating ensures that the knowledge base remains dynamic and increasingly valuable over time, continually shortening the OODA loop.
The Role of Human Investigators
While AI can recognize patterns, it lacks the nuanced understanding of context that humans possess. Human investigators can understand cultural, social and political nuances that might influence the interpretation of data. Human investigators can adapt and apply intuition, which AI currently cannot replicate. This intuition is crucial in making sense of ambiguous or contradictory information.
AI also lacks the ability to make ethical and moral judgments, which are often needed in intelligence work. Human investigators can weigh the ethical implications of their findings and actions.
Humans excel at creative/critical thinking and problem-solving, especially in novel situations. AI and ML, on the other hand, are limited to the data they have been trained on and can struggle with new, unforeseen scenarios.
AI versus Humans in OSINT Research and Analysis
When it comes to scaling the data, AI can perform quantitative analysis, focusing on data-driven insights. Humans, conversely, can combine quantitative analysis with qualitative insights, considering the broader context and implications. AI algorithms are generally less flexible and require structured data, while human investigators can work with unstructured data and adapt their methods as needed.
Human investigators can also draw upon a wide range of disciplines and experiences, allowing for a more holistic approach to OSINT. With technology like ChatGPT, on the other hand, we can achieve somewhat similar holistic goals. Meaning, the data that these Large Language Models (LLMs) are trained on literally have decades of knowledge in them. With the correct and targeted prompt engineered question, we could get that broad overview of experiences and disciplines.
Humans are Better at Analyzing OSINT Data
Humans can integrate various types of information – social, cultural, political and emotional – to form a comprehensive understanding of the intelligence gathered. We are capable of the critical thinking and healthy skepticism that is essential when analyzing the credibility and relevance of information. This is something that AI and ML cannot yet do.
The final decision-making in OSINT investigations often requires a strategic approach that considers the long-term implications and ethical concerns, an area in which humans excel. Humans can be each other's “devil's advocate,” pointing out assumptions, fallacies and biases.
Of course, we can ask LLM algorithms to proofread our report and point out biases, but what if that model was not trained to point out a specific form of bias? What if it can only read English text? What if it doesn’t understand cultural habits? The current LLMs out there, like ChatGPT, Bard and Claude, are known for pulling information that is factually inaccurate out of thin air.
This is why, in the end, humans will always make the final decision. We understand nuances, we can read between the lines and see the broad(er) picture. The nuanced understanding, ethical judgement, creative problem-solving and emotional intelligence of human investigators are irreplaceable in analyzing and making strategic decisions based on OSINT data. Therefore, the most effective approach is a symbiotic one, leveraging the strengths of both AI/ML and human intelligence to achieve the best outcomes in OSINT investigations.
The ShadowDragon Perspective
On our podcast, the ShadowDragon team often dives into timely and sometimes contentious topics that tie back to OSINT. Listen at the link below, and please get in touch with us if you would like to know more about how we are supporting investigators with OSINT tools. The list of positive outcomes from the use of OSINT (what we call, “OSINT for Good”) continues to grow.