AI’s Double-Edged Sword: Navigating the Risks and Rewards for Humanitarianism
By Rob Tyrie
This week I participated in the panel session that was called “A Conversation about AI”, for the International federation of Red Cross and Red Crescent Societies. It gave me a chance to work and think along with experts in the field and leaders in artificial intelligence and its practical applications. It was thrilling.
“The International Federation of Red Cross and Red Crescent Societies (IFRC) is the world’s largest humanitarian organization, providing assistance without discrimination as to nationality, race, religious beliefs, class, or political opinions. It was founded in 1919 and is part of the International Red Cross and Red Crescent Movement, along with the International Committee of the Red Cross (ICRC) and 191 National Red Cross and Red Crescent Societies" .
Heather Leson and Dr Mahendra Samarawikrama designed an intelligent conversation focused on the concerns and needs of the members of the societies. Nirendika Wanigasekara, PhD and I provided thoughts and commentary on the implications of AI for both humans and humanitarians along with answering thoughtful questions from the professionals of the IRC. But, let’s back up a second. In my research I wanted to know more about the IFRC and what it was doing in technology to support its mission and the resiliency of humanity. So I used the tools that I was talking about and found out the following:
What does the IFRC do with AI?
The International Federation of Red Cross and Red Crescent Societies (IFRC) is actively exploring and utilizing artificial intelligence (AI) to enhance its humanitarian work. Here are some key initiatives and applications of AI within the IFRC:
1. Generative AI: The IFRC has organized workshops and training sessions to explore the potential of generative AI in humanitarian work. This includes using AI to analyze large amounts of data, generate reports, and support decision-making processes[1][2].
2. Data Science: The IFRC has a dedicated data science community that leverages AI and machine learning to improve humanitarian services. This includes projects such as predicting disaster risks, analyzing financial data, and enhancing supply chain management[8].
3. Digital Transformation: The IFRC has launched a Digital Transformation Impact Platform to strengthen the digital capabilities of its National Societies. This includes investing in AI, data analytics, and digital literacy to improve the efficiency and effectiveness of humanitarian services[5][7].
4. Anticipatory Action: The IFRC uses AI to support early warning and early action systems, enabling proactive measures to protect people before disasters strike. This includes forecast-based financing and risk-informed early action partnerships[14].
5. Innovation and Partnerships: The IFRC collaborates with various organizations, such as Nesta and the UK Humanitarian Innovation Hub, to research and develop AI solutions for crisis response. This includes projects like collective crisis intelligence and participatory AI for local humanitarian action[11].
6. Ethics and Governance: The IFRC emphasizes the need for ethical and responsible AI development, ensuring that AI systems align with humanitarian principles and respect human rights[4][10].
7. Capacity Building: The IFRC provides training and capacity-building programs for its staff and volunteers to enhance their understanding and use of AI in humanitarian contexts[6].
These initiatives demonstrate the IFRC's commitment to harnessing the potential of AI to improve humanitarian outcomes, while also addressing the ethical and governance challenges associated with AI adoption.
AI is here. We made it for humans
In the rapidly evolving landscape of artificial intelligence (AI) and robotic process automation (RPA), debates about their societal implications rage across sectors. For international organizations like the Red Cross engaged in humanitarian relief efforts, these powerful technologies offer tantalizing opportunities as well as profound ethical and practical challenges. As we grapple with this technological transformation, key issues surrounding data privacy, algorithmic bias, workforce impacts, and ethical deployment demand rigorous examination.
The Data Dilemma: Privacy, Security and Public Trust
Humanitarian work depends on handling highly sensitive data - from personal details of displaced populations to medical records in crisis zones. AI systems could drastically improve data security through advanced encryption, biometric authentication, and anomaly detection. For instance, schemes like AI-enabled homomorphic encryption may allow analytics on encrypted data without decryption, preventing data leaks. Well clearly doesn’t work is leaving did in the clear or transmitting it in the clear. It risks the safety and privacy of human life. There needs to be a program I policy wrapped around identity and encryption that will become our human right hopefully sometime in the future.
Powerful data-crunching capacity also magnifies risks. A major breach, Like we’ve just seen in 2024 with hundreds of companies using Snowflake, this could enable exploitation of vulnerable groups via identity theft, targeted violence, or discrimination. The controversial Clearview AI facial recognition system illustrated how unchecked use of personal data can violate civil liberties. Such failures jeopardize public trust so vital for humanitarian organizations.
Data protection requires a delicate balancing act - leveraging AI/RPA capabilities while adhering to robust frameworks like GDPR and sector-specific guidelines. Efforts like the Harvard Humanitarian Data Ethics Principles advocate transparency, deliberative decision-making processes, and human oversight. But continual auditing of AI systems is crucial to detect unintended consequences or biases infiltrating models trained on messy real-world data. The same tools that can save people’s lives can also provide a web of autocratic surveillance to a level that’s no different from incarceration.
Confronting the Ethical Minefield
Beyond privacy concerns, the humanitarian deployment of AI systems raises thorny ethical quandaries. These range from perpetuating societal biases through tainted training data, to the moral hazards of delegating life-impacting decisions to algorithms.
A notorious example is the COMPAS recidivism prediction tool used in criminal sentencing, which exhibited racial bias by over-predicting Black defendants as future re-offenders. In humanitarian crises triaging limited resources like food, water or medical care based on biased predictions could prove disastrous and even life-threatening. We must ask - is it ethical to automate such high-stakes decisions without direct human oversight? My easy answer to that is that it’s not possible. We must have counsels, courts, investigators and of course, international laws, that is beyond boundaries and borders. We need them today.
Another issue is AI’s "opaque and inscrutable" nature - complex deep learning models can be impossible for humans to interpret fully. We know now that AI today is more than a stochastic parrot. It can reason and it can plan in swarms. But what kind of reasoning? An AI may recommend withholding aid based on legitimate data patterns, but cannot provide satisfactory reasoning if queried. This lack of explainability in humanitarian contexts raises accountability and legal liability concerns.
Ultimately, the AI ethics challenge will need ongoing dialogue between affected communities, humanitarian workers, technologists and policymakers. Multistakeholder partnerships like the UN PULSE initiative are promoting responsible AI deployment guided by human rights frameworks. And concepts like "human-centered AI" advocate for AI assistants but not replacements for humanitarian experts with on-ground context.
The Shifting Job Frontier
The rise of AI/RPA stokes pervasive fears of mass job displacement. While legitimate, these concerns in the aid sector must be viewed holistically. Some roles may indeed diminish, like data entry or manual supply chain operations eclipsed by automation software. Administrative work and moving data around, transforming and summarizing it is work that will no longer be done by humans.. It will be designed by humans or invented by humans but it’ll be done by robots. One of the areas that will definitely change this and scale it will be the adoption of using robots to collect observable data without humans involved into categories and classify it because that is the core and each of the technology. Without understanding it can collect information classify it summarize it and distribute it without human touch. Although we’re spending a huge amount of effort understanding the results of the harvesting of language... The harvesting of direct information from the environment will create something that aren’t large language models but well but large world models developed by robots. These may be possible to understand and only are able to be compared to human models. Like we study the world, scientists and analysts will have to study these newly-developed models that are probably underway in certain labs around the world. These are the labs where robots through sensors are being directly connected to the physical world and are designed to process that information digitally with a level of dimensions that are not possible to be understood by probably almost all of humans.
However, such disruptive technologies also generate new job categories. These new jobs will be the salvation of many countries. AI could augment and evolve - not eliminate - roles involving interpersonal skills, complex reasoning or on-ground humanitarian expertise. Machine learning engineers, data scientists, ethical AI reviewers emerge as critical future positions. I will keep repeating this most likely for the next 5 years. AI does not replace human jobs it replaces tasks. That will be the first step. However the second step in large organizations, as they reorganize around this new technology... There will be humans that were replaced as tasks are automated and humans are freed up to do other work, or released from employment because they are no longer needed in that new organization. It will take decades to move from this hybrid stance to new organizations. The organizations that will be the bellwether will be greenfields organizations just beginning now or very rich organizations that are unable to change while also operating. Think of the likes of Microsoft, Amazon or Google, this is how they work today. They reorganize based on innovation, new technology, new methods and new materials.
Furthermore, judicious AI deployment can empower aid workers by relieving burdens of tedious tasks. If menial paperwork is automated via RPA, staff can refocus on higher-impact duties like community outreach or disaster reconnaissance. With AI assistants filtering and visualizing data, decision-makers can act with greater agility.
But this workforce transition demands robust upskilling investments - and humanitarian agencies face budget constraints and fundraising pressures. Innovative approaches like partnerships with tech companies and universities, or establishing reskilling "humanitarian AI institutes", may accelerate closing the AI skills gap.
AI pioneer Kai-Fu Lee argues "For every old job that is displaced, we’ll get many new jobs in AI’s wake." Achieving this vision requires proactive workforce development plans - continuous learning paths, cross-training, and collaboration between technologists and mission-critical staff. There’s a reason it’s so much education about artificial intelligence and its impacts and its nature are being distributed in the world today. Leaders are recognizing that there’s a gap and there’s a shift that’s greater than any other technological wave in the history of the world. I believe it’s a righteous response to change. To educate is to survive. To educate is resilience. We must demand more education from every level of government in the world. We must pay for their donations from the rich and taxation of those who can afford it. There has to be a transfer of wealth to support the transition in this next technology wave or we will drown.
The Integration Imperative
Truly harnessing AI/RPA in humanitarian scenarios necessitates seamless integration - both technical and operational. On the technology front, interoperability remains a stumbling block. Diverse partner NGOs often utilize incompatible data platforms, tools and processes, hindering centralized AI deployment.
Recent efforts aim to tackle this challenge. The World Bank, UNICEF, and others collaborated on the Data Integration System for Development (InsightDIS) to harness AI models, data visualization, and shared repositories across agencies. Still, widespread adoption requires standardization both between and within institutions.
Beyond just technological concerns, embedding AI demands a collaborative cultural shift. Incorporating advanced technologies within humanitarian workflows raises apprehensions about automation encroaching on human decision-making. Aid workers may distrust opaque AI recommendations without explanations rooted in contextual nuances.
This underscores the importance of human-centered AI approaches co-designed by field staff, AI engineers, and community representatives. Cross-disciplinary teams can collectively identify high-impact use cases, address operationalized AI concerns, and demystify models through interpretable design or interactive interfaces.
Real-time crisis response illustrates the potential of collaborative human-AI networks. ALMA, an AI assistant from Singarea, rapidly processes satellite imagery, social media data, and field reports to generate maps and visualizations. This allows humanitarian coordinating bodies like the United Nations Office for Coordination of Humanitarian Affairs (UN-OCHA) to synthesize disparate data streams - augmenting situational awareness and resource allocation for disaster relief.
From Promise to Impact
When responsibly implemented, AI/RPA present immense opportunities to amplify humanitarian efforts' scale and efficacy. These transformative use cases span the entire crisis lifecycle:
Preparedness:
- Predictive analytics mining social media, climate and conflict data to anticipate emerging high-risk areas and preposition emergency resources.
- Population mapping through computer vision and satellite imagery to assess vulnerable communities.
Crisis Response:
- Damage assessment using drone imagery and machine learning models to triage impacted areas and expedite search/rescue.
- Supply chain optimization to maximize allocation of food, medicine and supplies through forecasting algorithms.
Recovery and Resilience:
- Refugee text analysis and needs assessment via natural language processing to identify urgent personal requirements.
- Climate pattern modeling with reinforcement learning to identify sustainable farming practices post-disaster.
Such AI-driven systems are already reality. Efforts like the UN Global Pulse employed natural language processing of Twitter/X data to track displacement during the Haiti earthquake. Machine learning models developed by Harvard and Direct Relief predicted cholera outbreaks in Yemen by analyzing environmental factors.
Yet realizing AI’s full humanitarian potential requires overcoming significant funding and capacity gaps. NGOs spend less than 5% on digital transformation compared to 33% by private companies. Public-private partnerships, opening technology transfers, and developing "AI4Good" curricula may propel adoption.
Ultimately, technology remains just a tool - our moral and ethical compasses must guide the journey. Concerns around algorithmic bias, transparency and accountability will continue to demand scrutiny. But a future where AI/RPA empower rather than replace humanitarian actors offers profound promise for alleviating suffering and building a more equitable world.
Notes
"Homomorphic Encryption" by Brenda Gail Buckowitz (W.W. Norton & Company, 2022)
"Humanitarian Data Ethics Principles": https://data.humdata.org/faq-dataethics
"Weapons of Math Destruction" by Cathy O’Neil (Broadway Books, 2016)
"AI 2041" by Kai-Fu Lee (Bantam, 2021)
"Innovation for the World’s Most Challenging Problems": https://www.unglobalpulse.org/
"Prioritizing Humanitarian Aid through Machine Learning": https://sitn.hms.harvard.edu/flash/2022/prioritizing-humanitarian-aid-machine-learning/
"Data Integration System for Development": https://datacatalog.worldbank.org/insights-aid
"Digital Transformation: Embracing the Data Revolution": https://www.mercy-corps.ai/
Perplexity Citations:
[1] AI is changing our humanitarian work — IFRC | Solferino Academy https://solferinoacademy.com/ai-is-changing-our-humanitarian-work/
[2] Generative AI in the IFRC Network - Solferino Academy https://solferinoacademy.com/generative-ai-in-the-ifrc-network/
[3] Bringing the Artificial Intelligence / Machine Learning Community ... https://preparecenter.org/story/bringing-the-artificial-intelligence-machine-learning-community-and-humanitarian-assistance-and-disaster-response-hadr-together/
[4] artificial intelligence and anticipatory humanitarian action https://blogs.icrc.org/law-and-policy/2021/08/19/artificial-intelligence-anticipatory-humanitarian/
[5] [PDF] DIGITAL TRANSFORMATION IMPACT PLATFORM - IFRC https://www.ifrc.org/sites/default/files/2023-12/DTIP-brochure-design-pages-v2.pdf
[6] Humanitarian AId?: Considerations for the Future of AI-use in ... https://reliefweb.int/report/world/humanitarian-aid-considerations-future-ai-use-humanitarian-action
[7] [PDF] IFRC Digital Transformation Strategy https://digital.ifrc.org/sites/default/files/media/document/2023-04/ifrc_digital_transformation_strategy_v2_102220_english.pdf
[8] IFRCGo/rcrc-data-science - GitHub https://github.com/IFRCGo/rcrc-data-science
[9] Discover Your AI Pioneer Finalists — IFRC - Solferino Academy https://solferinoacademy.com/red-cross-and-red-crescent-humanitarian-ai-pioneers-awards/discover-your-ai-pioneer-finalists/
[10] [PDF] AI for humanitarian action: Human rights and ethics https://international-review.icrc.org/sites/default/files/reviews-pdf/2021-03/ai-humanitarian-action-human-rights-ethics-913.pdf
[11] Localising AI for Crisis Response - UK Humanitarian Innovation Hub https://www.ukhih.org/news/localising-ai-for-crisis-response/
[12] Innovation at The Red Cross https://www.redcross.org/about-us/who-we-are/innovation.html
[13] Harnessing the potential of artificial intelligence for humanitarian ... http://international-review.icrc.org/articles/harnessing-the-potential-of-artificial-intelligence-for-humanitarian-action-919
[14] Early warning, early action | IFRC https://www.ifrc.org/our-work/disasters-climate-and-crises/climate-smart-disaster-risk-reduction/early-warning-early
[15] AI is changing our humanitarian work — IFRC | Solferino Academy https://www.linkedin.com/posts/ifrc-solferino-academy_ai-is-changing-our-humanitarian-work-ifrc-activity-7140718423184928768-vp3o
