Introduction
In a world of accelerating technological changes, the ways that we help and protect people are currently being redefined and adapted to new circumstances. Amidst the potential for technological advancements in genetic engineering, robotics, and healthcare to widen the socioeconomic divide even further, now more than ever, it is essential to define how technological innovation can be used as a force for good. In today’s contentious environment we have a responsibility to reckon with the questions of how civil society and public policy can harness the power of machine learning to initiate a shift from the public good to the common good. Within this discussion, there has been an increase in collaboration between the public and private sector, thereby leveraging the nexus of interactions between public affairs, civil society and technology. The United Nations is a leading player in driving increased collaboration between the public and private sector, in the hopes of spurring innovations in the realm of humanitarian aid. Their Sustainable Development Goals are galvanizing worldwide participation from large tech companies, social entrepreneurs, startups, international organizations (IOs) and non-governmental organizations (NGOs). Through these partnerships formed, we are currently witnessing a revolution in humanitarian aid, with increasingly technologically and digitally driven forms of prediction, deliverance and recovery in humanitarian contexts. The term Digital Humanitarians has been coined to define a movement of individuals guided by a desire to protect vulnerable individuals through innovative solutions to Big Data. Digital Humanitarians mobilize data and online resources in collaborations with NGOs to support relief efforts around the world.
The humanitarian community is currently faced with record numbers of people afflicted by population displacement, political wars, environmental crises and food and water insecurity. These challenges are only going to increase with the impacts of climate change and population growth. Perhaps the most promising advances in humanitarian aid relate to the possibility for AI to engage with the vast amounts of data and information flow. By relying on increasingly data-driven and technology driven solutions, aid and disaster response will be dramatically improved. Developments in prediction, delivery and recovery missions will foster faster and more focused responses and allow for sustainable innovations to help beneficiaries. In general, the ability to analyze and aggregate data in unprecedented ways, allow people to predict, monitor and respond in more coordinated and efficient ways (Bellevue, 2016).
This paper will analyze how insights from Artificial Intelligence (AI) can be harnessed as a positive force, in terms of addressing many of the world’s greatest challenges, including responding to natural disasters, famine, conflict and governance issues. Humanitarian aid is defined in this paper as providing material or logistical help to people, with the aim of alleviating suffering and saving lives. The three main areas in which AI will drive the future of humanitarian aid will be in prediction of human conflict or environmental disaster, deliverance of aid, and recovery from devastation. AI has the potential to transform humanitarian response and help the lives of millions of people, currently afflicted by trauma and devastation. However, when developing algorithms to help those in the most vulnerable situations around the world, it is essential to prioritize the security and safety of their data. Furthermore, in the context of Big Data and AI we must reckon with concerns relating to data bias, dissemination of false news and data security.
Background
AI, at its core, is a machine used to recognize patterns in data to simulate traditional responses consistent with humans (West & Allen, 2018). The algorithms use real-time data collected from a variety of sources to analyze and react to data instantly (West & Allen, 2018). Today, AI is predominantly associated with driverless cars, personal assistants and military drones. However, within the humanitarian community, there is a growing movement to incorporate the use of AI to help in humanitarian efforts in the realm of image analytics, drones and crowdsourcing. AI can be used in this area as a way of processing large quantities of information to help in analysis and in conversion of unstructured data into actionable knowledge.
There are currently many challenges and failings in the world of humanitarian aid, within which AI could provide some form of a solution. Currently, the key issues in the humanitarian community lie in the lack of organization, the lack of coordination and the lack of resources. A serious issue discussed by relief workers is that emergency aid does not adequately address the needs of the population, it often arrives late and is largely determined by the media profile or the political criteria, rather than humanitarian need (Fleshman, 2006). Furthermore, often media and press coverage are the primary means for pushing governments into action. Nonetheless, there is no way to guarantee that the media will focus on all given crisis situations or that these efforts will be sufficient to stimulate aid from governments and relief organizations. Often, by the time the cameras arrive and relief action is in place, it is only in time to record the dying, not to prevent the looming disaster (Fleshman, 2006). Another significant failing is in the structure of the humanitarian enterprise (Kent, Bennett, Donini & Maxwell, 2016). Essentially, the powers that govern humanitarian aid form an oligopoly of western donors, UN agencies and large NGOs. Within this system, however, there are large inefficiencies and much of the services provided are dictated by political drive, rather than by adequate needs assessment (Kent, Bennett, Donini & Maxwell, 2016). Finally, a fundamental problem lies in the physical deliverance of aid, due to the lack of information and the difficulty of providing aid in perilous zones.
Clearly, AI cannot solve all the current failings in the humanitarian system. However, it holds the power to provide faster and more extensive analyses of information, using large quantities of data and satellite imagery and to enhance the deliverance of aid to people in danger zones through the use of drones or self-driving trucks. Furthermore, with the waning influence of international organizations and the declining amounts of government funding to development programs, now more than ever, it is essential to invest and foster partnerships between private companies, international organizations, non-governmental organizations (NGOs), experts in AI and data scientists to help solve some of the world’s greatest challenges.
Prediction
In order to deliver effective humanitarian aid, it is imperative to truly understand what the people in need want. In humanitarian contexts, the information available is often distributed across multiple groups, making it difficult to synthesize and to guide effective response. Furthermore, international organization, NGOs and government agencies often lack access to information on conflict resolution, cultural insight and other key knowledge to guide interventions. This makes it exceptionally hard to intervene in time sensitive missions. One key area where AI and machine learning can be of benefit in terms of humanitarian aid are in prediction. AI can be used to create predictive models for disaster relief, allowing responders to analyze large-scale population behavior and movement by using data gathered from a variety of sources, including social media posts, news sources and live imagery. Using this data, responders can create targeted programs to distribute supplies in a more orderly and efficient way.
Recently, the United Nations, World Bank, International Committee of the Red Cross, Microsoft Corp., Google and Amazon Web Services announced a partnership dedicated to using AI to predict famine to help millions of people affected by food insecurity (World Bank, 2018). This initiative is called Famine Action Mechanism (FAM) and uses a combination of satellite data of rainfall, crop health, social media and news reports on uprisings, violence or surges in food prices (World Bank, 2018). Currently, most humanitarian action to relive famine occurs after the conflict or disaster, meaning that people receive material assistance too late. By using data to predict famine crises in advance, interventions can happen earlier and more efficiently to save more lives and reduce humanitarian costs by 30% (World Bank, 2018). Essentially, FAM’s aim is to shift global action from crisis response to crisis prevention by building capacities from large amounts of data to predict future crises. The technology used in this scenario includes a set of analytical models called Artemis, which uses AI and Machine Learning to estimate and forecast worsening food security crises in real-time (World Bank, 2018). By predicting famine earlier, organizations will hopefully be able to promote interventions, which address the early signs of emerging food crises, including safety nets and coping mechanisms (World Bank, 2018).
Currently, with the rise of environmental disasters there are increasing risks of water wars or ‘hydro-political’ issues. Competition over limited water resources will be a key concern in future decades, whereby the effects of climate change alongside population growth will exacerbate competition for water, increasing regional instability and social unrest (Whigham, 2018). By using large amounts of data of daily temperature, rainfall estimates, emissions forecasts, crop failures, droughts, among other factors it is possible to create algorithms used to predict the likelihood of conflict in various areas (One Concern, 2018). A company called One Concern has created an AI program called Seismic Concern that accurately predicts seismic events and is also working on solutions for floods, wildfires and hurricanes (One Concern, 2018). Using AI, One Concern is able to predict and react to impacts of natural disasters, “conduct multi-hazard analysis of its critical infrastructure”, “run realistic training scenarios and build collaborative plans to better prepare for emergencies” and analyze the “complex interdependencies of our environments, like healthcare, power, food, shelter and more” (One Concern, 2018), in order to allow organizations to help ease recovery. In terms of seismic predictions, their platform analyzes hyperlocal, near-real time insights on live earthquake to provide quickly developed situational awareness allowing people to prioritize resources to save more lives (One Concern, 2018). The system, therefore, relies on a combination of human intelligence alongside AI to help accomplish missions in critical situations. Their technology essentially “assigns a unique, verifiable “digital fingerprint” to every natural or manmade element from the smallest rock to complete structures to mega cities” (One Concern, 2018). Overall, by providing insights from hyperlocal data, One Concern is helping to build resilient communities to allow for faster relief action amidst the rise of deadly natural disasters (One Concern, 2018).
Another area where AI can be of benefit in terms of prediction is in the agricultural sector. In regions afflicted by chronic hunger, AI can help to transform the agricultural industry. Smallholder famers occupy 60% of the world’s poor and hungry (Opp, 2018). Using AI to help farmers increase their yields could, subsequently, help to eradicate hunger around the world. The data collected from low-cost sensors, UAV’s, satellite imagery, weather data, soil conditions and crop status, can be analyzed with AI to help farmers understand how to maximize yield, when to fertilize their produce and how to improve market-value (Opp, 2018). This idea of “smart agriculture” is already being practiced in the developed countries, so it could be easily expanded to developing countries, where the need for the maximization of crop yield is significantly greater (Opp, 2018).
Deliverance
AI driven drones are frequently discussed in the context of military intervention. People fear advances in drone technology could lead to an apocalyptic world of autonomous military weapons. However, drones hold the potential to help humanitarian efforts in terms of delivering medical and food supplies to remote areas, mapping terrain and predicting population movements.
Currently, the UN’s World Food Program (WFP) is looking at how AI can be used to support disaster relief efforts. Unmanned Aerial Vehicles (UAV) can use AI to deliver necessary medical and food supplies to remote or conflict afflicted regions unreachable by manned vehicles during times of environmental disaster or sociopolitical conflict(WFP, 2018). Aid agencies are facing more difficulty than ever before to deliver food assistance to people in need due to air strikes, attacks on humanitarian convoys and blocked access to besieged areas (WFP, 2018). Autonomous self-driving trucks provide a promising solution in these cases in terms of providing assistance in inaccessible or perilous environments. While the use of self-driving trucks in this context is yet to be done, WFP is collaborating with the German Aerospace Center (DLR) to build a blueprint for remotely piloted and self-driving technology in contexts of emergency response in dangerous areas (WFP, 2018). WFP and DLR have developed the first technical and operational concepts to equip and operate resilient, remote-controlled trucks delivering goods (WFP, 2018). Currently, they are looking at “how to set up operation centers, from where specialists can control the trucks, and global connectivity to prepare for an envisaged pilot” (WFP, 2018).
Some companies in Africa are using AI to improve deliverance and access to medical care. In Rwanda, for instance, a company called Zipline is using drones to deliver medical supplies and blood to hospitals, which are difficult to access by manned vehicles (Giles, 2018). This effort has significantly helped those living in remote parts of the country where immediate access to medical help is rare (Giles, 2018). Furthermore, the drone system has reduced waste of blood by 95% (Giles, 2018).
The number of refugees and displaced people in the world are at a record high (Smith, 2018). AI and machine learning have the “potential to improve the lives of approximately 68 million displaced people in the world, 28 million of whom are refugees” by better understanding their needs (Smith, 2018). The World Food Program (WFP), Norweign Refugee Council and Microsoft are working on using AI to improve communication with people in situations of protracted crises, such as in Syria and Yemen. The chatbots use language understanding, machine translation and speech recognition to intelligently assist individuals and connect them with resources, allowing aid workers to better understand the immediate needs and circumstancesof refugees (Smith, 2018). The chatbot communicates in 20 different languages at very low costs, enabling targeted assistance (Opp, 2018). Alongside that, Microsoft’s AI for Goodprogram is looking at how to protect the human rights of individuals in protracted living situations. Their work on deep learning focuses on how to better predict, analyze and respond to serious human rights abuses (Smith, 2018). By utilizing “AI-powered speech translation, people can connect with pro bono lawyers who are protecting people’s human rights” (Smith, 2018).
WeRobotics is another company involved with using AI driven drones for disaster relief. Their drones are commonly used for humanitarian response through delivering supplies, medicines and through search and rescue efforts.
Recovery
Humanitarian organizations are dependent upon credible and timely information to respond quickly and efficiently to natural disasters and emergencies. Today, organizations are using alternative sources of information to gain increased insight into the situations of those in need, specifically data from social media. However, it is challenging to process social media data in real-time. To account for these difficulties, AI is being increasingly employed in these contexts to help with the retrieval and classification of data and with predictive analytics. Essentially, using AI with these vast amounts of data can provide more targeted disaster response, since it allows government and relief organizations to parse through large quantities of fragmented data in a more effective way.
AIDR (Artificial Intelligence for Disaster Response) is an open source platform that uses AI to filter and classify social media posts related to environmental emergencies, humanitarian crises and state conflict. The platform sifts through messages and twitter posts by specifying keywords or hashtags, based on supervised machine learning, thereby relying on both human and machine computing, allowing human users to train algorithms to automatically classify tweets and determine whether or not they are relevant to a particular disaster (Imran, 2014).AIDR is specifically “designed to scale up the abilities of human workers by intelligently removing noise from the data e.g., in the form of duplicate and irrelevant messages” (Imran, 2014). Alongside that, automated surveillance is becoming increasingly common as researchers are looking at how to use machine learning to analyze live video footage and verify its authenticity (Imran, 2014).
A key part of recovery initiatives is through donations and funding of money to the country or local organizations. Corruption is a significant issue in delivering funds, particularly in Africa. The World Bank reported $245 million lost through fraud or corruption between 2007 and 2012 (Kenny, 2017). However, people are largely unaware of who is taking the money and how much is being stolen. The question of whether AI can stop corruption in its tracks is discussed amongst humanitarian actors. As systems and procedures become increasingly digitized the possibility to leverage available data to track corruption and other fraudulent risks is more apparent. AI could be utilized to sift through the large quantities of data in the hopes of picking up a red flag, which might point to where the corruption is happening (Sharma, 2018). The World Bank is currently looking at this use case, specifically how AI can promote transparency in all areas of government administration by sifting through “diverse datasets to detect patterns that hint at the possibility of corrupt behavior… to see links in bidding patterns of the winning and losing bidders to numeric patterns under “Benford’s Law,” along with beneficial ownership information from around the globe” (Sharma, 2018). The possibility to use machine learning to analyze available data on World Bank- financed procurement, alongside datasets gathered from other international organizations and government data, would allow us to gain increased insight into how to make safer and better decisions on public spending and how to mitigate certain risks of financial corruption (Sharma, 2018).
Concerns
AI led missions in the context of humanitarian intervention are only likely to increase as collaborations between the public and private sector continue to grow. We are seeing more and more innovation and collaboration between governments, tech companies, international organizations and NGOs. Alongside that larger tech companies are increasingly engraving social impact ethos’s into their missions. With this in mind, it is essential to have a number of safeguards in place to protect individual’s data and right to privacy.
When discussing AI more broadly the key concerns fall under three main categories; technological unemployment, lethal autonomous weapons and data manipulation/ biased systems. The biggest concerns in the context of humanitarian intervention are in the areas of security, privacy and data mismanagement. The threat to privacy is a key issue, particularly when discussing such a vulnerable population. Therefore, when large tech companies are analyzing and controlling the data of populations at risk their need to be safeguards in place to protect their information being used for unintended purposes. Furthermore, if smart systems, such as a drone delivering system or a self-driving truck, become compromised or hacked the consequences could become a matter of life or death.
It is critical to have ethical principles in place to govern how artificial intelligence programs are developed. Moreover, privacy protection mechanisms need to be inscribed into the data frameworks to ensure that algorithms are processed fairly and accurately. AI essentially automates critical human decisions in real-time. Although in the examples discussed above of cases of humanitarian intervention there are rarely examples when ethical or moral issues arise in decision making, these problems can always occur. Whether it might be an issue of a drone or self-driving truck crashing and hurting civilians or an algorithm incorrectly analyzing civilians needs assessment, the logic behind the systems choices must be well defined and understood.
Another significant risk is in creating fake media and misinformation, which AI could potentially make worse. When reporting on humanitarian crises it is critical that the information provided is accurate. However, with the rise of fake images, news and videos, it is becoming increasingly hard to distinguish fact from reality. We have already witnessed instances of fake images circulating around war situations, such as in Syria, so there lies a serious danger in relying on AI to report on humanitarian aid when the program is unable to distinguish between a real or fake image, tweet or Facebook post. Currently, social media accounts on Twitter and Facebook are spreading misinformation throughout the internet to manipulate readers and sabotage politicians and organizations. If humanitarian interventions, therefore, become increasingly reliant on AI for predicting war or environmental disasters, we risk facing the same problems that arose in the United States Presidential Election of 2016.
In terms of specific limitations in the humanitarian use cases discussed above, there are certain drawbacks with the tools reliant on AI. NetHope’s ICT4D Webinar Series discussed the use of UAV’s and machine learning in the context of aerial imagery to detect buildings during the Vanuatu Cyclone Pam with the aim of creating a multi-risk index of the area. In the webinar, speakers discussed how humanitarian drone missions take a lot of time and how there are challenges with aerial data, regarding how disaster and damage assessment is defined. Essentially, AI does not diminish all problems related to quality control, since the system still requires human assessment and analysis of imagery. Humanitarians, therefore, need to inscribe their own frameworks for damage and disaster assessment, regardless of the use of AI and machine learning systems.
Conclusion
We are living through a unique time where collaboration between private companies and public organizations is chipping away at the traditional structure of power held by the government. AI and algorithms are reshaping many aspects of society, including the economy, healthcare, employment and education, and we need to realize how these changes can positively impact society. Emerging technologies have the ability to reduce inequalities in society, improve education and foster development around the world. While the humanitarian community is often a late adopter of new technologies due to fears of infringement on individual security and privacy, now more than ever it is important for the humanitarian community to participate and contribute to these changes. The current humanitarian community is comprised of an exciting assortment of individuals in NGOs, startups, private companies and international organizations. The new tech driven solutions of satellite monitoring, biometric scanning, digital identity and mobile money allow humanitarian actors to act faster, to respond to people’s direct needs and to inspire progress through a bottom up approach. We are living through a revolution in humanitarian aid, which is focused on equipping individuals with the tools necessary to improve their circumstances, rather than providing the traditional forms of material aid, which render the beneficiary dependent on a continuous source of assistance.
The effectiveness of humanitarian action is dependent on coordination between humanitarian practitioners and those in need. Clearly, technology alone cannot solve the problems currently affecting millions of people and we must be wary of the risk that an increasing reliance on data and technology might widen the gap between those helping and those being helped. It is critical to not desensitize those from their humanitarian missions, but rather allow them to embed core humanitarian principles into the systems (Bellevue, 2016). Furthermore, we need to imagine a future of humanitarian aid that is augmented by leveraging technology, but is not solely reliant on it. With the use of AI, it is important to envisage an algorithm that keeps society in the loop, thereby embedding a socially good ethos into an algorithm where both the aid responders and the affected civilians understand and oversee algorithmic decision. It is clear today that progress is extremely dependent on the actions of those in power, who are increasingly tech companies, to ensure that the rights and freedoms of the vulnerable populations are protected.
Overall, the use of AI in humanitarian aid contexts holds promise for mitigating some of the current flaws in the humanitarian system, many of which are outlined in this paper. Innovation within the humanitarian community is currently flourishing through the partnerships of public and private sectors, allowing for the sharing of skills and knowledge from entrepreneurs, data scientists, aid workers and public policy experts. Clearly, the use of AI or other technology driven tools in humanitarian contexts cannot solve all the problems of humanitarian aid. However, more than anything the advancements we are witnessing today generate increased access and communication. Right now the success of these innovative technologies in humanitarian contexts hinges upon practitioners ensuring that the technological tools are deployed in a way that aligns with the core humanitarian principles and enhances the connection between those providing aid and those receiving it.
References
Belliveau, J. (2016, June 9). Humanitarian access and technology: opportunities and applications. Retrieved from. https://ac.els-cdn.com/S187770581632330X/1-s2.0-S187770581632330X-main.pdf?_tid=696a8d95-ba68-4bae-9257-9f741ae2d79d&acdnat=1545241980_853922c33ae01aa305542ef527dfae2
Collins, K. (2017, October 04). How AI, Twitter and digital volunteers are transforming humanitarian disaster response. Retrieved from https://www.wired.co.uk/article/digital-humanitarianism
Digital Humanitarians. (n.d.). Retrieved from http://www.digital-humanitarians.com/
Fixing the humanitarian aid system | Africa Renewal Online. (n.d.). Retrieved from https://www.un.org/africarenewal/magazine/january-2006/fixing-humanitarian-aid-system
Giles, M. (2018, April 05). Zipline launches the world's fastest commercial delivery drone. Retrieved from https://www.technologyreview.com/s/610735/zipline-launches-the-worlds-fastest-commercial-delivery-drone/
How Much Aid is Really Lost to Corruption? (n.d.). Retrieved from https://www.cgdev.org/blog/how-much-aid-really-lost-corruption
Humanitarian health computing using artificial intelligence and social media: A narrative literature review. (2018, January 31). Retrieved from https://www.sciencedirect.com/science/article/pii/S1386505618300212
Imran, M. (April, 2014) AIDR: Artificial Intelligence for Disaster Response. Retrieved from https://mimran.me/papers/enabling_rapid_disaster_response_using_artificial_intelligence_and_social_media.pdf
(n.d.). Retrieved from https://www.oneconcern.com/product
New Artificial Intelligence Tools for Deep Conflict Resolution and Humanitarian Response. (2015, July 07). Retrieved from https://www.sciencedirect.com/science/article/pii/S187770581501036X
(Opp, R) (2018, November 15). The Future Of Humanitarian Aid?. Retrieved from https://aibusiness.com/xprize-ai-future-humanitarian/
Planning from the Future: Failings of the Humanitarian System. (n.d.). Retrieved from http://fic.tufts.edu/publication-item/failings-of-the-humanitarian-system/
WFP. (2018, April 27). Self-Driving Trucks. Retrieved from https://innovation.wfp.org/project/self-driving-trucks
Sharma, V. (2018, November 16). Can artificial intelligence stop corruption in its tracks? Retrieved from http://blogs.worldbank.org/governance/can-artificial-intelligence-stop-corruption-its-tracks
Smith, B. (2018, September 24). Using AI to help save lives. Retrieved from https://blogs.microsoft.com/on-the-issues/2018/09/24/using-ai-to-help-save-lives/
United Nations, World Bank, and Humanitarian Organizations Launch Innovative Partnership to End Famine. (n.d.). Retrieved from https://www.worldbank.org/en/news/press-release/2018/09/23/united-nations-world-bank-humanitarian-organizations-launch-innovative-partnership-to-end-famine
West, D. M., & Allen, J. R. (2018, May 09). How artificial intelligence is transforming the world. Retrieved from https://www.brookings.edu/research/how-artificial-intelligence-is-transforming-the-world/
Whigham, N. (2018, October 18). Researchers use AI to predict outbreak of water wars in the future. Retrieved from. https://www.news.com.au/technology/environment/researchers-use-ai-to-predict-outbreak-of-water-wars-in-the-future/news-story/f1d91a6ba462aba9b94886b9a1112dfb