Predicting conflict, defusing hate speech, monitoring climate conflict: in the hands of human peacemakers, we can have an “AI For Peace”
We are happy to crosspost extracts from the introduction to AI For Peace - a new book and think tank which explores the possibilities of AI for peace-making in the world (rather than its usage in warfare), and also how a peace perspective can inflect AI’s own development. The authors are Branka Panic and Paige Arthur:
…Exploring and utilizing "Al for peace" has been limited, even though this is a growing field of work that also intersects with many other fields. We describe how citizens, data scientists, policymakers, and communities can be empowered to use Al to build and sustain peace in their countries.
But what is peace? In this book, we follow the consensus of peacebuilding practitioners that peace should be understood not as a negative term—the absence of violence—but rather as a positive one—the presence of social trust, resilience to violence, and strong civic and community institutions to manage conflicts when they (inevitably) arise.
Peace is a holistic concept that draws on many different facets of human relationships, from economic security to social trust to socio-emotional wellbeing, among others. In this book, we do not have space to cover every topic relating to peaceful societies; rather, we survey some of the most important developments for people working in this field and those hoping to contribute to it.
The use of Al for peace is new, and it has been driven by the revolutions in data and technologies that have taken place over the last few years. These revolutions have generated volumes of data (satellite, video, images, news, text, etc.) that are growing at a massive speed, scale, and frequency.
They have also resulted in highly advanced statistical modeling and the use of Al, including machine learning, neural networks, and natural language processing, to analyze that data. One example of how quickly change is happening is the advance from GPT-3.5 (ChatGPT) to GPT-4, which took less than 6 months.
Peacebuilders have already been using new technologies—not necessarily Al-driven—to do a variety of things. There is much about the digital and data revolution that has had positive benefits for peace work.
There has been an explosion of useful data to help understand how strong societies are, where their institutions to regulate conflict may be weak, and to help guide decisions about how to strengthen them further.
Peacebuilders have also seized upon a myriad of practical applications:
They have used human review of satellite images to detect evidence of war crimes in Darfur;
they have used mobile cell phone networks to warn communities of potential militia attacks in remote areas of the Democratic Republic of the Congo;
and they have used social media platforms to share ideas, connect with one another, organize against authoritarian governments, publicize human rights abuse, and generate social movements for justice.
In their daily work, peacebuilders have drawn upon productivity tools like automated translation (and now ChatGPT) to improve their ability to communicate efficiently and effectively.
These applications are relevant both to countries that are currently peaceful (to help them stay that way) and to those that are emerging from violent conflict. Indeed, Al may contribute even to sensitive peace negotiation processes between warring parties.
People involved in peace processes can already take advantage of digital tools like satellite images and geographic information systems (GIS) to monitor ceasefires and the disengagement of military forces.
During peace negotiations themselves, which have usually been the domain of a small elite, negotiators and mediators can also draw on Al-driven tools to ensure that the peace process taps into the broader public sentiment around key conflict issues-which may not be fully represented by the elite men (most often) sitting at the table.
For example, the United Nations has developed a tool for sentiment analysis and opinion mining using natural language processing of public social media sources.
Al can help mediators see patterns, anticipate stumbling blocks in the negotiation, and make sure that there is a broader buy-in to the key terms of the agreement than that of only the conflict leaders. All of this will contribute to a greater likelihood that a peace agreement, once signed, will actually endure.
In short, Al-driven tools are just one part of a larger set of human and machine "technologies" that peacebuilders can draw upon in their work, depending on their needs and the ethics of using them.
Our aim is to provide a fresh perspective on Al as a potential force for good for peace and in already conflict-affected regions.
We delve into the diverse technologies in the Al landscape, such as machine learning, natural language processing, and image processing—and demonstrate how, when ethically used, these can be harnessed to collect and process vast amounts of data, uncover patterns, and enhance the work of peacebuilders.
We also shed light on risks and emphasize the importance of embedding ethics in all stages of Al-enabled activities-design, development, and implementation. It is imperative that those dedicated to peace navigate this landscape with caution, balancing the potential benefits of Al with an awareness of its malicious use and unintended consequences.
This book is an introduction to the latest advancements in emerging technologies and encourages practitioners, especially, to take a more active role in the use of Al in their work.
At the same time, the book speaks to data scientists and Al experts, raising their awareness about the impacts of their work beyond bias and transparency and adding a peace and prevention lens to their work. It will also empower them to bring their technical expertise to the peacebuilding field.
We cannot hope to be comprehensive in addressing two such large and multifaceted fields: Al and peace. There are sadly a number of fascinating issues we have not had space to cover, such as how "smart cities" (which deploy sensors in urban areas to collect useful data for managing urban spaces) could be designed with the aims of peacebuilding, conflict prevention, and social cohesion in mind.
Or how Al-driven tools are starting to be used by UN peacekeepers in some of the world's most challenging contexts.
Our chapters therefore capture key aspects of innovation—and hopefully make readers curious enough to start asking further questions. Beyond describing Al innovations, we take care in each chapter to explain what the specific challenges are for peace for each topic, because this broader understanding helps to explain why Al tools can be relevant to solving those challenges.
In Chapter 1, we start with conflict prediction. This is because preventing violent conflict before it starts is the goal of every person in this field, given the devastation of war. We describe how risk for conflict is assessed, as well as the important Al-assisted gains that have been made in the past ten years in forecasting which countries are at highest risk for violence.
But we also show both the limits of current modeling and why predicting conflict onset—when a peaceful society will tip into war—is such a hard problem. We look at new approaches, such as Al-assisted efforts to better understand why countries are peaceful, rather than why they fall into conflict.
In Chapter 2, we look at hate speech because this is a key method by which leaders create polarization and mobilize populations for violence. Our focus is on how Al has not only led to the massive proliferation of hate speech but is also being used to detect, limit, and actively counter it.
Understanding how to fight hate speech is also important, given the parallels with the political uses of misinformation and disinformation. When citizens' arguments (and deep beliefs) are based on misinformation, deep fakes, and hateful stereotypes, the risk of tipping into violence can be greater.
While much of this chapter focuses on content that is proliferating on the big social media platforms, we also give some examples of other groups that are using Al to detect and counter hate speech.
In Chapter 3, we survey the positive uses of Al to support human rights, especially in conflict-affected countries. Respect for human rights is essential to peace, and the erosion of rights is a warning signal that a society may be tipping into violence.
We condemn the many violations of rights for which Al-enabled surveillance technologies are being used, but our focus here is on the citizens, activists, and scientists who are bending technology for their own peaceful and rights-protecting ends.
We conclude the chapter by showing how human rights frameworks are useful for curbing some of the most pernicious social effects of Al-enabled technologies.
Chapter 4 tackles an area of growing urgency: climate change. We raise awareness of the promise and perils of Al for understanding the relationship between conflict and climate change, and we explore new tools for anticipating, preventing, and responding to climate-related conflict.
Our aim is to explore the potential for building planet-centered technologies as well as foster more informed climate- and conflict-aware technologists, data scientists, designers, engineers, and technology activists.
As we navigate the intricate intersection of Al and peace, we remain mindful of the ethical considerations and challenges that accompany this rapidly evolving landscape.
In Chapter 5, the concluding chapter of this book, we grapple with the ethical dilemmas surrounding the "Al for peace" concept. We examine the risks associated with biases; privacy infringements; unintended consequences; digital colonialism; and the pressing imperative for responsible and accountable design, development, and implementation.
This book aims to raise awareness and serve as a call for action for humanity to seize the momentum in rights-respecting, ethical Al development—and steer it towards a future in which Al aids humanity in sustaining peace, rather than perpetuating wars.
….While this book illustrates the potential for Al to contribute to sustaining peace, the prevention of violent conflict, and the safeguarding of human rights, it is vital to acknowledge that AI alone is not a panacea.
Ultimately, while Al can be a valuable tool to address various challenges of conflict and peace (when used responsibly), achieving peace requires multifaceted approaches that encompass not only technological innovation.
They also require human peacebuilders—who understand peacemaking and conflict resolution practices; how to foster social and economic development; and the practice and politics of coordinating the collective efforts of governments, international organizations, and citizens themselves.
…Our hope is that this work—which is a narrow snapshot in time—will provide a common point of departure for future innovation and collaboration among peacebuilders, technologists, policymakers, and others. By harnessing the capabilities of ethical Al to advance peacebuilding efforts, we can strive towards a future where technology and humanity converge to foster lasting peace.