ICTworks

Scholars and practitioners have contributed much to our understanding of the broader macro drivers of violence in fragile settings. They identify how shocks interact with grievances to potentially trigger violence, and how these interactions demand consideration of the multidimensional nature of risk and the potentially self-reinforcing nature of multiple risks that entrap fragile societies.

Much work has been done to identify what constitutes hostile vernacular in political systems prone to violence, however, it has not considered the language of specific influential actors. In study, The relationship between influential actors’ language and violence: A Kenyan case study using artificial intelligence, natural language processing was used to identify the sentiment associated with leaders’ language.

Natural Language Processing (NLP) software reads and understands words’ meaning in multiple languages to allocate sentiment scores using the latest artificial intelligence advances, including deep learning, transforms
unstructured textual data (i.e. a tweet or blog post) into structured data (i.e. a number) to gauge the
authors’ changing emotional tone over time.

The model predicts both increases and decreases in average fatalities for look ahead periods between 50 and 150 days, with overall accuracy approaching 85%.

This demonstrates the utility of local political and sociological theoretical knowledge for calibrating algorithmic analysis. This approach may enable identification of specific speech configurations associated with an increased or decreased risk of violence.

This finding is significant and suggests a significant role for influential actors in determining the positive or negative pathway of a society towards or away from violence. This finding is based solely on the sentiment of the language used by leaders. It is a first indicative step that serves to illuminate a significant unexplored sphere of social science, recently accessible as a consequence of emerging technologies.

The analysis and findings in this paper indicates a statistically significant relationship between language and future political violence, demanding further inquiry. Larger training data and further testing of other machine learning models, including deep neural networks, could produce even more robust results.

The study also collected sentiment data about influential actors in Kenya. However, limited resources did not allow for sufficient exploration of its impact on capacity to predict violence.

Further, this approach only addressed the cumulative body of sentiment among the selected 30 influential actors, rather than the relationship of specific individuals’ language to violence.

A lightly edited synopsis of The relationship between influential actors’ language and violence: A Kenyan case study using artificial intelligence