Helping halt global deforestation with AI

To give her vision wings, MANA needed a system that could rapidly mine huge data volumes, as the body of knowledge around deforestation continues to rapidly grow. She also needed a platform that could quickly scale in line with the demand for this information.

The IBM Garage™ team in France was happy to step up and help. They understood that AI services on a cloud platform could deliver what MANA needed.

Nicolas Comete, IT Architect and IBM Garage Engineer, and Quentin Siraut, IBM Garage Software Developer, worked closely with the MANA team on the project. As avid hikers and campers, they saw it as an opportunity to use their expertise to benefit the environment.

Image above: An indigenous boy in the Amazon, September 2020 (Credit Thomas Pizer/aquaverde.org)

“It’s not always the case that developers can commit ourselves to something that corresponds to our values,” says Comete. “From my point of view, many of the problems we face today are coming from the way we behave with nature, including forests and animals. This engagement with MANA was one way I could help make the planet a little better.”

In addition to supporting MANA’s cause, Comete and Siraut had the requisite AI and cloud expertise for the project. Comete first tried his hand at software development when he was a teenager. “Since I was a child, I enjoyed playing video games. When I turned 15 years old, I began to develop video games and write algorithms,” he says.

Siraut, in turn, always wanted to innovate, which attracted him to the field of software development. Now a self-described AI, machine learning and deep learning enthusiast, he enjoys applying IBM Watson® capabilities in novel ways.

“More and more, innovation is brought to various industries with AI; I like that we can help MANA use Watson AI for good,” he says.

The drivers of worldwide deforestation vary by region but are largely attributed to human activities, including commercial and small-scale agriculture, logging, cattle ranching and mining. However, Global Canopy reports (link resides outside of ibm.com) that out of 500 companies and financial institutions whose supply chains put forests at risk, 43% do not have a policy to reduce deforestation.

Prior to engaging with IBM Garage—a transformation framework that brings together people, processes and technology—MANA had compiled thousands of trusted sources from international NGOs. These contacts, which included environmental activists in the field and residents of indigenous and other rural, often remote communities, held firsthand knowledge of ongoing issues in deforestation hot spots. “It’s these individuals who really know what corporations shouldn’t be doing and how their activities impact the region,” says Mignotte.

These sources generally rely on social media and other digital channels to alert conservation and other organizations to pending acts of environmental destruction. Yet given the competing, sometimes contradictory news and messages on social media, they often can’t connect with their intended audiences.

The MANA-Vox platform would actively listen to their voices using AI in the cloud. The process entailed scanning collective communications—including links to articles and websites—for relevant posts and then verifying the sources, cross-checking the information, and using the resulting insights to compile corporate rankings. The more people are speaking about the corporation on sources such as Twitter, the higher it will rank.

When the IBM Garage team picked up the project, MANA and its volunteers had already developed the first MANA-Vox algorithms. The IBM team in France had also helped advance the basic application design and development, including deploying some IBM Watson services. The next step was to build a stable, production-ready minimum viable product (MVP) using best practices.

“MANA has a lot of passionate, intelligent volunteers that worked on different parts of the tool, but because it wasn’t created by one person or team it lacked unity. We turned to the IBM Garage team to move the project forward and create a unique tool in a unique space. We also needed it to be easy to use,” says Zoe Berenger, Technology Consultant and Product Manager at MANA Community.

After jointly building the MVP, the IBM Garage team connected it through an API server to Twitter for testing. They decided to limit testing to information sources found on Twitter, which provided them with a deep but relatively contained data set with diverse vernacular.

“A single tweet in itself can link to an article or website with several paragraphs and hundreds of lines of copy that need to be analyzed,” explains Siraut.

Currently, the MANA-Vox solution analyzes thousands of tweets and linked content daily, and the team adds more information sources each day. Of the analyzed content, the solution labels approximately 98% as nonrelevant. MANA employees and volunteers then manually review the 2% of relevant content to verify that it’s the type of information MANA needs. The algorithms can then be fine-tuned based on the findings.

“By quickly processing Twitter feeds and pinpointing relevant content, the application’s AI saves MANA countless hours of manual work. Workers can focus their attention on evaluating a very limited set of pertinent information rather than losing time on a large quantity of data,” explains Siraut.

In the next solution phases, MANA will finish designing a public-facing dashboard and develop a back-office solution to facilitate solution administration. MANA anticipates the MANA-Vox platform will go live in early 2022. In addition, it recently reached an agreement with a major NGO that would allow it to feature MANA-Vox insights on its website. It is engaged in similar discussions with other NGOs.

Once in production, MANA will gradually scale the MANA-Vox solution to gather intelligence from multiple digital channels and thousands of credible sources. Currently, the solution follows 740 credible sources for insights on 450 corporations.

With the platform, MANA can help usher in a new era of corporate accountability. Most information about companies’ environmental footprints has traditionally come from their own annual reports, which often lack timeliness, transparency and traceability. MANA will arm conservationists, NGOs and other organizations with a near real-time list of corporations posed to do the greatest environmental damage, empowering them to work more effectively for change.

For example, they can take direct legal actions to stop deforestation, or launch consumer campaigns pressuring companies to employ environment-friendly production and sourcing practices for palm oil and other commodities. Financial advisers can also use the rankings to invest in companies that have adopted zero-deforestation policies.

“MANA fosters a democratic approach, where people from different sectors can unite and work towards the same goal,” she says.

Mignotte says that financial companies especially look forward to accessing MANA-Vox insights, which stand apart from those provided by other organizations. Because the MANA-Vox platform analyzes data from civil society and from people in the field who personally witness deforestation, it can provide deeper visibility into distributors—and not just suppliers—benefiting from raw materials produced through environmental destruction.

As MANA continues evolving the MANA-Vox platform, the IBM Garage team stands ready to support the small company however needed. “It’s really nice when we can help develop solutions for NGOs like MANA,” says Comete. “We can meet people who really make our day and have fun.”

Discover solutions to help you manage blind spots in your customer data

Based in Paris, France, MANA is a startup NGO that seeks to help protect global forest ecosystems. Once launched, the MANA-Vox platform can empower NGOs and other entities to make more evidence-based decisions by providing near real-time rankings of corporations’ environmental footprints. Founded in 2015, MANA’s core team is assisted by a large network of volunteers.