Pentagon Invests in Sentiment Analysis Tools But How Useful Will They Be?

There are signs that the U.S. Department of Defense (DoD) is increasing its efforts to understand sentiment in support of information operations. A recent article from Mark Pomerleau of C4ISRNET highlighted a nearly $1 billion contract awarded by DoD to Peraton to support four Combatant Commands with influence planning, content development, dissemination of messages, and the assessment of those messages. During interviews at the annual Association of the U.S. Army conference in Washington, DC, earlier this month, leaders and industry emphasized the importance of sentiment analysis tools to support both targeting strategies and evaluation of the success of messaging operations. As the author noted, “Understanding the effectiveness of messaging will allow commanders to tweak and adjust their information operations campaigns and strategies to be more effective.”

While there are many tools that will be useful to crafting, targeting, and assessing influence and information operations, this article does not delve deeply into some key considerations for integrating new tools such as sentiment analysis capabilities into real-world workflows. The promise of the technology overshadows the real-world challenges. Like with the use of other artificial intelligence/machine learning (AI/ML)-type tools, there should be a framework for considering how to select and use the right tool for the job (and ensure they don’t inadvertently inject new, unconscious biases into the algorithmic process). Without understanding the technology under the hood, it is unclear how good or bad the specific sentiment analysis tool is (for example, are we employing a Ferrari or a Yugo, or something in between; having something may be better than having nothing, but there is no way to determine the extent of the advantage this might offer). Having a simple framework for teasing out these issues would be helpful to the operator, as well as the acquisition decision maker.

Additionally, how will such tools be integrated into the overall workflow (and training needs) for the MISO operator?

If we continue to lack the cultural or linguistic understanding of the target environments, will these tools be able to overcome that disadvantage, or will we simply paper over those deficiencies and assume a false sense of security that the technology makes up for the human shortcomings?