Amin Maghsoudi

Project Researcher at the HUMLOG Institute

Phase 1: 20 years back, a sudden deadliest Bam earthquake occurred in the cold December  2003, just 100 miles aways from my hometown that vibrated our house for couple of seconds in the early morning. The tragedy event has the moment magnitude of 6.6 with a dead toll of at least 34000 people and injuring up to 200,000. The majority of houses (traditionally made by mud) were collapsed. The international humanitarian teams were immediately deployed in coordination with the local actors. Rescue dogs were used to search for survivals and if anyone could survive under the collapsed buildings. There were no automated detective or robots implemented at that time at that context to locate survivors quickly, no social media to track the messages of survival, no satellite data to predict and estimate the damage, and even no unified message to the early waring system reached to people in advance (controversial messages spread from different sources such as TV đź“ş). Only few survivors were stayed in the car in the event of Earthquake during the cold winter in the mid of desert 🏜.  

Fig 1. A generated photo by open AI DALL.E2 lab, the text description as “2003 Bam earthquake collapsed houses made by mud” (Note: this is a fake picture created by AI tech, for more real picture on the events, you can google and there are plenty images).  

Phase 2: Few days back, a magnitude of 5.9 earthquake happened in north-western Iran has killed three and injured hundreds on 29th January 2023. The two events are not comparable in terms of infrastructure, strength and scale of earthquake, density of population, geography, and etc, but the question is what has been changed in terms of technology advancements to help reducing the number of casualties, damages, costs, etc. To date, there are so my many studies on the use of technology in humanitarian response and also in humanitarian logistics (e.g., the use of drones for last medical delivery, or blockchain for the smart contracts to track the aid supplies). But also, a lot of data on social media are shared via different platforms such as Telegram, Twitter and Instagram, and that could be of help to track the need of affected populations. Yet the question of misinformation and reliability of such data and how to filter such massive unstructured data and track them in the real time is the key. Lets also not forget, every second in a disaster reponse is vital and the need of affected communities changes over time very quickly. Today, they may need a shelter, tomorrow need to rebuild their house and day tomorrow they may need to get back to their business, for example.  

Figure 2. A real damaged house after the earthquake struck city of Khoy in Iran’s West Azerbaijan province (Source: Soheil Faraji/ISNA/AFP)

The two examples are among thousands of disaster events annually occurring across the glob. Each has their unique characteristics and thus needs for a different approach to respond. At the end of the day, this is a human’s (or Human-Robots) decision to decide how and on what way to respond and thus an effective coordination between all decision makers is the key for a successful reponse.  

Phase 3: Gear Up.2022 was very informative for all logisticians as participants and facilitators, which gathered over 142 humanitarians from 37 organizations. It was my honour to participate to such insightful simulation training as HUMLOG representative. The training was led by World Food Program on behalf of the Logistics Cluster and Emergency Telecommunication Cluster in Germany. For the 12 days of intensive simulation work (days and nights), the main point raised into my attention was the management of data and information and how to communicate/coordinate with each other during the response. On daily basis, over 6000 emails has been sent and received by each actors, and they needed to go through and search among unstructured data on needs for example, and transfer them to a very structured data and information to be used for the decision making. Other that than that, there was a desk room including young fellow specialized in information system to communicate and manage data and inputs between facilitators and participants. Excel was used massively as means of integrating inputs to store data who do what, where and in what capacity. The picture below shows setting up satellite antennas in camps. This was the key concern at the beginning, as all participants begged to connect to the network to access to data and information.  

Figure 3. Members of aid organizations setting up antennas and satellite dishes in camps (Source: THW/Daniel Schrieck)

Phase 4. ChatGPT: What about it?   

I have inspired by a conference paper most recently, addressing how responsible artificial intelligence is going to be in crisis management by Lee et al. (2022). In fact, according to this paper, there are a lot that AI can offer to the field of crisis management.  As there are huge datasets including social media data, satellite imagery data (e.g., CCTV camera data), digital trace data (traffic data), mobile phone activity data, and sensor data. There are a lot that AI can do to process, analyse and transfer them into fruitful information to be used for decision making in disaster responses. AI is a field of compture science and engineering that aims to create machines that can perform normally require human intelligence, such as visual perception, speech recognition, and even decision making. One of the cool and mostly discussed issues around AI in social media, and also among academics and practice, is about the open AI platform called ChatGPT (Chat Generative Pre-Trained Transformer). It eventually, give you a response based on what you write in the platform. This is another mean that can help people to get some input. But, its on human to check their validity and reliability of provided information. Even, open AI has another application called “classifier” can indicate whether the generated text is written by human or AI. Yet, the classifier is not fully also reliable as has some error. However, I played around the platform, and I think it can be useful when searching for something , and the robot can give you a summary and overview of possible options. For example, I asked and wrote “how AI has been or can be used for logistics of aid during emergency responses and disasters”? Look whats the AI reponse to my question:  

“Artificial intelligence (AI) can be used in a variety of ways to improve logistics during emergency responses and disasters. Here are a few examples: 

  1. Route optimization: AI can be used to optimize the routes for delivery of aid, supplies and personnel during a disaster. By analyzing real-time data on traffic, weather, and road conditions, the AI system can determine the best routes for delivery and reduce delays, which can be critical in emergency situations. 
  2. Inventory management: AI can be used to help manage the inventory of supplies and aid. By analyzing data on demand, supply, and distribution, the AI system can predict what resources will be needed and when, and help to ensure that the right supplies are delivered to the right places at the right time. 
  3. Predictive maintenance: AI can be used to predict when vehicles and equipment used in disaster response and logistics will need maintenance, so that they can be serviced before they break down. This can be critical in emergency situations, where delays can have serious consequences. 
  4. Resource allocation: AI can be used to help optimize the allocation of resources, such as aid, supplies, and personnel, during a disaster. By analyzing data on demand, supply, and distribution, the AI system can determine where resources are needed the most and how best to allocate them. 
  5. Decision support: AI can be used to provide decision support for logistics managers and emergency responders, by providing them with insights and recommendations based on real-time data, this can include traffic, weather, and other information that can help them make better-informed decisions.

It’s worth noting that AI can be a powerful tool for improving logistics during emergency responses and disasters, but it’s important to use it responsibly and in conjunction with human judgement. Additionally, for AI to be effective, it is important to have reliable and accurate data that can be used to train models to make predictions and decision.” 

Source: openAI, ChatGPT 

The summary is updated only until 2021, the reliability of information needs to be checked, but in overall all five examples of application of AI in humanitarian logistics sounds relevant. I like the last point, the human judgment is needed for the decision making. Overall, the technology itself, particularly AI and its features, is another tool to be used and applied in disaster responses and humanitarian logistics to help decision makers in order to reduce their human error, cost, time of delivery of aid. However, it needs to be understood, studied, and people needs to be trained how to use it efficiently in advance before its implementation at any event. All is about the data, data analytics, data management and coordinating them through engaged actors , which matters highly in disaster responses.  

The viewpoint is my own.  

#BamEarthquake, #KhoyEarthquake, “HumanitarianLogistics, #AI , “ChatGPT, #DisasterResponse