Peace Speech

Today, the extraordinary power of hate speech to mobilize destruction and violence is evident around the globe.

But what about the power of peace speech?

Peace speech is a basic structure that helps to build and sustain peacefulness between people and between groups. It may seem simplistic, but differences in the quality and tone of these simple human encounters – which occur millions or billions of times a day in communities – bubble-up to form the norms, taboos, memories and expectations that shape our lives.

Peacekeepers working in conflict zones are currently using data-science and natural language processing to track hate speech – monitoring hostile news, blogs, broadcast and social media posts in order to provide early warning predictions of increases in ethnic tensions or violence in vulnerable local communities, thereby linking linguistic indicators to actions on the ground.

But we are missing the opportunity to leverage peace speech. To date, no technique has been developed to help us understand, measure and track the power of peace speech – peaceable language for building and maintaining more robust, peaceful communities.

In response we are applying emerging techniques from data science and natural language processing to the study of highly peaceful societies in order to better understand the vital role language could play in measuring and tracking sustainable peace.

Watch this video to learn a bit more about the project:

Our initial work in this area focused on developing sustainable peace and conflict lexicons with the help of subject matter experts (SME). Using 2 months of news articles from 3 major news outlets, we used these lexicons to show that a simple lexicon-based peace metric mainly tracks with the level of peacefulness of 7 countries around the world. More recently, we tried to test the generalizability of this result. We tested these lexicons using 10 years of news reporting published by a large number of news outlets from 20 countries. The analysis on this dataset showed that, at this scale, the same peace metric does not distinguish well enough the peaceful and non-peaceful countries. We then used the same news corpus to develop frequency-based lexicons, comparing word frequencies in news reporting from peaceful and non-peaceful societies to identify lexicons for peaceful and non-peaceful societies. The peace metric based on these lexicons tracks very well the level of peacefulness of the 20 countries, over 10 years. More interestingly, these lexicons differed from the original SME-driven ones in fundamental ways. We also developed a document classifier that showed high performance in classifying news reports from peaceful countries and non-peaceful countries. Using an attention layer in the classifier we extracted the set of words that carry the most importance in the classification process, developing yet another lexicon for peaceful and non-peaceful countries.

There is significantly more to do, specifically we hope to:

  • Test the generalizability and sensitivity of these findings, across countries, time, type of reporting, etc.
  • Develop a new theoretical framework for the relationship between linguistic indicators and peaceful societies, which can offer important insights on building peace through language.
  • Test the value of these lexicons towards the development of metrics, beyond the simple peace metric used to-date, to measure and track peace.
  • Establish a live, public dashboard of peace indicators on a website that will update and track peace speech in specific geographic locations around the world.
  • Establish new metrics and measures for tracking both hate speech and peace speech in communities – providing a more robust set of tools for modeling and generating predictions.
  • Provide a baseline understanding and characterization of a general language of peace – which can be subsequently built on and supplemented by local actors with a more grounded sense of local parlance in specific communities around the world.

Of course, it would be foolish to claim that all peaceful societies are the same – their differences are clearly manifold. However, perhaps at its core, sustaining peace can be understood and modeled as a set of fundamental underlying human dynamics – like structures of language. This is the ultimate aim of the Sustaining Peace Project – to employ the best of science to learn from and celebrate the multitude of peaceful people and places that shine so bright in our world.