Researchers analyze peace with computer science

Paul Mayne // Western NewsWestern Modern Languages and Literatures professor Juan Luis Suárez, right, and graduate student Yadira Lizama-Mué are using Natural Language Processing (NLP) – a computer science technique – to scan, read, and analyze thousands of pages of drafts, documents and media releases about the peace agreement to answer this question.

Words can play a critical role in turning dreams of peace into reality.

Researchers at Western have found this is particularly true for victims of the Colombian conflict, which ended in 2016 when the government and the country’s largest insurgent group, the Revolutionary Armed Forces of Colombia (known by its Spanish acronym, FARC), signed a peace agreement ending nearly six decades of violence and human rights abuses.

During this period, approximately 220,000 people, including 45,000 children, lost their lives. The Unique Registry of Victims in Colombia provides an official registry of victims of the armed conflict.

The final peace agreement stated redress for victims were at the core of the agreement.

But do the words of the peace agreement reflect both negotiating parties’ intent to focus on victims?

Western Modern Languages and Literatures professor Juan Luis Suárez and graduate student Yadira Lizama-Mué are using Natural Language Processing (NLP) – a computer science technique – to scan, read and analyze thousands of pages of drafts, documents and media releases about the peace agreement to answer this question.

“If victims are not in the centre, it is easy to exploit their rights and it leaves them vulnerable,” Suárez said.

A victim-centred approach focuses on empowering victims as engaged participants in the peace and justice process, and prioritizes their wishes, safety and well-being.

“Our research recognizes the significance of the Final Peace Agreement. Throughout the process, the parties showed a genuine interest in including the victims in the dialogue,” Suárez said. However, he warned the language used during the negotiations in Havana doesn’t reflect this intention as deep as the victims need.

Suárez and Lizama-Mué programmed NLP to count the frequency of certain words –“peace,” “conflict,” “victims,” and “justice,” as examples – and look for sentence patterns and themes around them. “We wanted to focus on the language around the ‘conflict’ and its inseparable terms, ‘victims’ and ‘justice’,” Suárez explained.

They found the agreement focuses little on victims, dominated instead by discussions of governance, national identity and legal processes.

“If complex and technical documents like peace agreements are not ironclad, every step of the future is a minefield,” Suárez said. “It becomes very easy for vested interests to destroy the agreement.”

The end result is victims themselves don’t reap any of the agreement’s benefits.

Suárez and Lizama-Mué’s work comes at a critical time in Colombia’s history. The country is counting down to presidential elections in May 2018, a mere 15 months after it officially ended the continent’s largest armed conflict.

“Alliances will define the future of the nation, and consequently, the future of peace implementation,” Lizama-Mué said.

Suárez and Lizama-Mué are the first researchers in the world to use computer science and artificial intelligence tools to study the language of peace and reconciliation in peace agreements and treaties.

Their work will help governments and international agencies like the United Nations and the International Criminal Court review peace agreements and treaties to ensure they focus on conflict victims and help those who need it most. In Canada, this is particularly relevant to documents produced by Canada’s Truth and Reconciliation Commission for Indigenous communities.

Given the complexity and duration of negotiations, peace agreements and treaties go through multiple drafts and revisions. The final agreement is often a complicated political and legal document.

“NLP is an economical and efficient way to automatically read these documents,” Lizama-Mué said.

Suárez wants future negotiating parties to test the content of documents they publish, to “make sure their public declarations and the principles that inspire the negotiations are really reflected in the language choices they make.”

Suárez and Lizama-Mué will forward these recommendations to the Colombian government with the results of their research.

They also combined NLP and Twitter data to track discussions about the peace process by analyzing three million tweets and hashtags by former FARC leaders, politicians, journalists and other Colombian citizens during the year-long negotiations in 2016.

“The discussions in 2016 were very polarized, but it is better to have war through words over social media than through arms in the battlefield,” Suárez said, pointing out the importance of having victims actively participate in the debates and discussions, which makes the peace process more participatory and effective.

He is cautious, warning: “When you sign a peace agreement, that’s when conflict ends but that’s not when peace starts.”