English
- Natural disasters are on the rise due to climate change.
- Artificial intelligence can improve disaster response, from reducing the time to assess damage to monitoring social media to more quickly and effectively deliver aid.
- It's important to be cautious about the limitations of AI, and more collaboration is key to maximize its benefits.
More than 160 million people a year are threatened by floods, hurricanes, fires and other natural disasters. And the situation will likely get worse.
Already, natural disasters occur four times as often as they did in 1970. According to estimates, such events could grow in frequency and ferocity with the effects of climate change.
Artificial intelligence has potential to alleviate the damage by marshalling relief resources more efficiently and effectively. It can accelerate the delivery of aid and sharpen the decisions of relief workers on the front lines.
Have you read?
Disaster resilience efforts may look very different tomorrow from how they appear today. Once an advancing cyclone or hurricane is identified, for example, geo-spatial, weather and previous disaster data could be used to predict how many people will be displaced from their homes and where they will likely move. Such insights could help emergency personnel identify how much aid (water, food, medical care) will be needed and where to send it. AI algorithms could instantaneously assess flooding, building and road damage based on satellite images and weather forecasts, allowing rescuers to distribute emergency aid more effectively and identify those still in danger and isolated from escape routes.
McKinsey’s Noble Intelligence is just one example of an initiative trying to harness AI’s potential to support humanitarian causes. For instance, the team is developing an algorithm that will reduce the time it takes to assess damage to buildings such as schools from weeks to minutes, using a combination of satellite, geo-spatial, weather and other data. This information can then be used to identify the best places to set up temporary school tents and where to prioritize reconstruction efforts.
As another example, other organizations are using AI techniques to interpret social media feeds following disasters. This type of analysis could provide vital on-the-scene information about infrastructure damage and aid being provided to victims by flagging images from shelters where people are without blankets or waiting outside in the streets.
Yet even as many public sector organizations and private sector data players such as Mastercard, Microsoft and Google contribute to the improvement of disaster relief, the impact of the efforts is still held back by several challenges.
One is limited scope. Many private-sector initiatives involve one or a few government or NGO partners, and focus on specific use cases, often in relative isolation from the larger disaster-relief community and without integration into established disaster relief protocols. This leads to fragmentation of efforts and may result in AI-derived insights and algorithmic tools being given to organizations that cannot maintain them or incorporate them effectively into their decision processes.
Second, while much data exists that could benefit disaster relief – satellite, geo-spatial, telecom, social media, financial – it’s not always accessible when it’s needed. What’s more, datasets are rarely combined in ways that could unlock additional insight, both with other big datasets but also with data from experienced operatives on the ground. This ground view can be even more valuable than insights from big data but is often not captured and analyzed in a systematic way.
Finally, in disaster situations where, by definition, human lives are at stake, it’s important to be cautious about AI’s limitations. Data analysis doesn’t always deliver what proponents claim, making it challenging to assess such claims without an established process to rigorously review algorithm methods and assumptions. For example, AI models designed to assess residential damage have been used on commercial buildings even through these buildings rely on different materials, construction methods and regulations. In a world where the ethics of AI are increasingly scrutinized, there are no standards to which developers and users have agreed to adhere.
How has the World Economic Forum helped initiate a more effective response to natural disasters and humanitarian crises?
In 2005, the World Economic Forum helped to establish the Logistics Emergency Teams (LET), a network of representatives from four of the world’s largest logistics and transport companies (Agility, DP World, Maersk and UPS) who work together in partnership with the World Food Programme-led Global Logistics Cluster to deliver free humanitarian assistance.
To date, the LET has responded to more than 20 large-scale natural disasters and humanitarian crises, providing critical logistical support for hurricane victims in Haiti, Rohingya refugees in Bangladesh, tsunami victims in Indonesia, civilians in war-ravaged Yemen and many more.
In 2018, 1,943 employees of LET member companies were trained in humanitarian logistics, contingency operations and disaster response to ensure that they were better prepared for future crises.
Read more about how the LET initiative continues to be an exemplary model for public-private partnerships.
Contact us if you're interested in getting involved in impactful initiatives as a member or partner of the World Economic Forum.
How can we maximize the benefits of AI in natural disaster scenarios? There are three opportunities:
- First, enhance collaboration between current initiatives, focused on specific use cases between a few partners, into a more impact-focused network of AI-driven disaster support. The attention currently devoted to developing algorithms should be balanced with at least as much energy and resources to make sure these tools are widely available and used on the front line of disaster relief. In many cases, that means more capability building. We also see duplication of efforts, with the data science community working on similar use cases, which could be streamlined. One option might be to establish a domain-specific partnership or coalition across which industry and global agencies would coordinate focused development teams, as just one model.
- Second, in the near term, develop more basic data capture and coordination tools across different agencies on the ground, rather than focusing the majority of investment on highly advanced AI. This could provide the information “fuel” for new lifesaving algorithms in future. Therefore, it would be beneficial to spend an equal amount of development effort on these foundational tools while more sophisticated algorithms are also being developed.
- Finally, there is an urgent need for more domain-specific agreements on ethical AI principles. Many initiatives have been started by global agencies, including the United Nations and the European Union, to develop principles to guide beneficial uses of AI generally. But given the broad scope, this is likely to take time. In the interim, it would be useful to align stakeholders more narrowly in specific domains, such as disaster response. This might include setting an algorithm review process to ensure AI solutions meet specified standards before they are widely released.
The opportunity for AI to help in the disaster resilience arena is vast – guiding relief efforts, ensuring better evacuations, distributing aid that could help tens if not hundreds of millions of people per year. While there are challenges to overcome, with the right level of coordination and partnership, this brighter future could be a bit more within reach.
