This is part one in a two-part series.

The largest land war since World War II remains underway. Multi-front conflicts continue in the Middle East, the collapse of the Assad regime in Syria shuffles alliances and competition across the region, and ships play cat and mouse across the South China Sea. Geopolitical struggles are primed to continue as the world reorients towards a multi-polar system. Understanding quantifiable signals that presage the initiation of armed conflict is more important now than at any time in recent history.

What if there were a temperature gauge that showed policymakers and world leaders when the stories that are told might trigger action?

This paper builds on previous work quantifying sentiment and narrative shifts as potential signals that a party is preparing the narrative battlefield in advance of steps towards kinetic activity. This approach would have significantly challenged consensus among foreign policy professionals that a Russian invasion of Ukraine was unlikely in 2022, highlighting significant negative sentiment shift and use of dehumanizing language as a warning signal. It also showed even greater Russian narrative positioning prior to the announcement that the Russian military would call up an additional 200,000 soldiers in January 2023. This analysis can serve as a valuable temperature gauge to understand where narrative conflicts may boil over to physical ones.

What’s the temperature now? Updated analysis argues for the following high-level conclusions:

• Heating Up? Sentiment in Korean language publications in South Korea discussing North Korea is now at its most negative level in the data set, lower than during the THAAD crisis of 2016-17, closure of the reunification village in October 2023, or missile tests to protest US Vice President Harris’ visit to South Korea in Sepember 2022. This occurs as North Korea ratified the mutual defense treaty with Russia and sent thousands of North Korean soldiers to fight with Russia in Ukraine.
• Cooling off? Sentiment in Mandarin-language publications in China discussing Taiwan continues to trend more positively. This is consistent with a view that China is not preparing activities that would risk significant Taiwanese casualties.
• Mixed? Sentiment in Mandarin-language publications in China discussing the United States turned more positive in the days immediately following the US election but have since moved lower. Sentiment in China towards the US is more positive than sentiment in the US towards China.

Theoretical roadmap

The idea that states leverage historical context to create a story of self, that this story of self is then used as a domestic and international political tool, and that challenges to this story can drive conflict is a well-trodden road in both academia and foreign policy circles. These concepts will be briefly described below to provide a backdrop for the extension of these concepts to the quantification of sentiment as a tool to better understand potential future actions. The current Russia/Ukraine conflict is used as a case study to illustrate major ideas.

R.D. Laing was the first to develop the concept that individuals have a need to define their sense of self as much as they have a need to defend their physical selves, which he termed “identity security.” Jennifer Mitzen later extended this concept to state-level behavior by highlighting the risk of conflict when a state’s identity security is under pressure. Brent Steele provided the additional insight that “critical situations” may arise which force state action to maintain identity security, even when doing so sacrifices the state’s physical security.

Ayşe Zarakol argues that the collapse of the Soviet Union served as a critical event in Russia’s historical memory (akin to Steele’s “critical situation”), and that subsequent status changes drove ontological insecurity. Identity insecurity, or “status anxiety” in Zarakol’s language, drives actions to address this ontological discomfort, even at the expense of material interests. This insecurity, driven by reduced status is also a topic covered by Brent Steele in the context of Britain’s entry into World War I. This helps explain why analysts who focused purely on realpolitik dynamics misforecast the risk of a Russian invasion of Ukraine. These analysts focused on the deterrent effect of risks to Russia’s physical—inclusive of economic—security, while ignoring the potential restoration of ontological security as a potential offset. This is an important insight for formulating deterrence policies moving forward.

Joseph Nye originated the idea that “soft power” can be used to both establish precedent for actions as well as frame them as justified post-fact. Jade McGlynn updated these ideas in the context of modern Russia, showing how historical propaganda narratives were used to set the stage for the invasion of Ukraine. Peter Pomerantsev lays out how these narratives are used to prepare public opinion, particularly if significant causalities are expected. Language used to specifically dehumanize a population has been recognized as a precursor to violence by authors ranging from Hannah Arendt to Herbert Kelman.

A methodology to quantify narrative and tones to better understand the stories that a country conveys to multiple specific audiences can be leveraged to more thoroughly understand when stories transition to action. 

Methodology

The theoretical analysis described above first analyzed sentiment using a publication as a proxy for target audience, examining different stories told to China’s civilian population (through People’s Daily) vs. military populations (through People’s Liberation Army Daily). Key conclusions of this work highlighted that negative sentiment expressed in China toward the United States began at a low level in early 2018 as the US administration initiated trade sanctions against China, but accelerated aggressively after the beginning of the COVID-19 pandemic—therefore COVID changed the narrative environment in meaningful ways. In addition, negative stories regarding the US told to China’s civilian population (a potential sign of scapegoating) were significantly more aggressive than those targeting China’s military population (thus providing little evidence that the military was being narratively prepared for conflict).

Subsequent work expanded audience targeting through the use of publication language as a proxy for offshore (English) vs. onshore (domestic language) messaging. Key conclusions centered on the fact that while Chinese English-language publications had seen a deterioration of sentiment expressed towards Taiwan (potential signaling to an offshore audience), Mandarin-language articles continued to see rising sentiment scores. This was consistent with a view that China was not preparing for physical conflict with Taiwan, based on evidence from the Russia-Ukraine conflict that narratives shift in advance of kinetic conflict.

This paper looks to further advance this study through an analysis of dyadic pairs to better understand if there are signs of a potential risk of sentiment and narrative shifts that could lead to conflict; a temperature gauge, if you will.

Source material for the analysis is contained in the Global Database of Events, Language, and Tone (GDELT), associated with Google’s Jigsaw Project. This database is analyzed to understand the total volume (V) of stories containing the search term in the country / language / publication in question, measured as a percentage of the total volume of stories (TV). Dividing the volume (V) by the total volume (TV) solves for the volume intensity (VI), or the percentage of the total stories in the specified parameters that contain the term in question. The tone (T) regarding the topic is calculated within GDELT (detailed information about dictionaries used, calculations, and contextual analysis can be found here). Multiplying the volume intensity by the tone derives the volume intensity weighted tone (VIWT) as shown in the equation below:

This approach prevents a strong tone in a very limited number of stories from dominating the results.

The full data set developed from GDELT for this paper can be found in the Harvard Dataverse here, consisting of statistics for volume intensity, average tone, and key headlines driving tonal calculations for each term from Jan. 1, 2017 through Dec. 12, 2024. This allows for replicability of all analysis and charts shown in this paper.

So what’s the temperature?

Part one of this paper, above, discusses high level conclusions and presents a theoretical and methodological overview of the use of sentiment and tonal analysis to better understand when stories might boil over to conflict. Part two will leverage these ideas and tools to provide detailed analysis of a series of country pairs to guauge whether they are trending towards or way from potential conflict. This analysis will look at publication language as a proxy for target audience to further analyze narratives told to different internal and external audiences. Dyadic pairs analyzed include Russia / Ukraine, North Korea / South Korea, China / Taiwan, China / US, and China / Japan.

James R. Sullivan, CFA ([email protected]) is an Adjunct Fellow/Non-Resident WSD-Handa Fellow at Pacific Forum and an External Associate in the Economic Competition and Conflict Research Group at King’s College London.

PacNet commentaries and responses represent the views of the respective authors. Alternative viewpoints are always welcomed and encouraged.

Photo Credit: University of Maryland, Division of Research

If you want to see more insightful analysis and impactful events in the Indo-Pacific, support us by donating. Your contribution fuels our mission for a secure and cooperative future: Donate Here