Can #SelfiesSaveLives? Humanitarian chat-bots help disaster response in Southeast Asia
How can social media contribute to disaster relief efforts during emergency events? Banner image: Shutterstock/Herwin Bahar
This article is part of a series in which OECD experts and thought leaders — from around the world and all parts of society — address the COVID-19 crisis, discussing and developing solutions now and for the future. Aiming to foster the fruitful exchange of expertise and perspectives across fields to help us rise to this critical challenge, opinions expressed do not necessarily represent the views of the OECD.
To keep updated on all of the OECD's work supporting the fight against COVID-19, visit our Digital Content Hub.
“But first, let me take a selfie”
A flood selfie submitted by Jakarta resident to PetaBencana.id during historic flood event on January 2nd, 2020 becomes part of an alert about flood-affected neighbourhoods. Photo and image: Yayasan Peta Bencana
The lyrical refrain of the YouTube hit captures a familiar disposition for millions of Indonesians, even during disasters: whether standing knee-high in flood waters, posed against a backdrop of volcanic ash clouds, or riding motorcycles through smouldering haze, Indonesians habitually take selfies that necessarily capture the geographical characteristics of the disaster-prone region. During emergency events, social media feeds are inundated with millions of pictures within seconds, as residents of the fourth most populous country in the world share their sentiments, warnings and efforts to co-ordinate amid rapidly changing conditions. Although the “disaster selfie” has been denounced as pervasive and attention seeking—and certainly there are valid ethical and safety considerations that must be weighed—I would like to consider the implications of this habit by looking beyond simplistic assumptions and generalisations of the communication trope. Within the context of crisis response, what does the disaster selfie offer?
To consider this question, it is fundamental to look beyond these self-presentation images in isolation, and instead focus our attention on how they network together.
In a country of 260 million residents spread over 17,000 islands, centralised information collection and dissemination systems cannot feasibly or logistically capture the diversity of events and sensibilities spanning the archipelago—much less so during the sudden onset of disasters. The complexities of archipelagic governance is also reflected in the organisation of neighbourhood units (Rukun Tetangga, or RT) that act as foundational structures of micro governance in Indonesia. Besides serving administrative purposes, the social structure of the RT embody a strong sense of communal responsibility, where residents contribute to the well-being and maintenance of the neighbourhood in the spirit of gotong royong (mutual aid). The microforms of societal governance that the gotong royong spirit fosters have always been a fundamental part of residential life, enabling communities to support neighborhood maintenance and need with far more agility and precision than any centralised agency.
These decentralised forms of self-organisation, mutual aid and collective responsibility offer much to draw on for disaster preparedness and response. During disasters, information is the most important resource. Yet, official reports prepared by government agencies and NGOs are often only available hours after the original event, in large part due to the challenging processes of collecting and validating relevant data. These reports are also prepared by organisations operating at different scales, and are often not standardised in their assessment of disaster severity and impact. The lack of consistent, reliable, up-to-date information creates difficulties for agencies operating at local and national levels as they struggle to make co-ordinated decisions in a timely manner; this often results in inadequate or poorly timed response and the misallocation of critical resources.
If a picture is worth a thousand words, a contextualised selfie can be indispensable in crisis response as it captures the direct relationship of disasters to specifically affected communities. From the height of flood waters relative to a body or the accessible width of earthquake-damaged roads relative to a person in/on a vehicle, to the differentially affected damage caused to neighbourhoods by landslides, these detailed and rapidly changing conditions are crucial to understand the types of response and resources needed. Time critical decisions—such as the logistical consideration of whether to send boats, trucks or helicopters for rescue, or, as has been the case in earthquake-struck areas in Indonesia, whether to transport water via truck loads or on the backs of rugged motorcycles—all rely on an understanding of rapidly changing conditions and how they affect different socio-cultural neighbourhoods. Importantly, while these contextual details cannot be captured by satellite data, they are easily captured by the smartphones we all carry in our pockets.
Structured as an interactive micro-survey, the disaster report asks residents to confirm the location of impact, the severity of the disaster, and to add a photo and/or free-text description. Photo and image: Yayasan Peta Bencana
If the millions of disaster selfies that are routinely shared by Indonesians could be harnessed to document disaster impacts, the network of disaster selfies could be transformational in providing rapid impact at an unprecedented scale, speed and resolution, otherwise inconceivable by centralised forms of data collection. While crowdsourcing has been used to collect situational information on hazards since the mid-1800s, the notably high penetration of internet-connected mobile devices and social media usage in Southeast Asia provides an opportunity to harness collective intelligence at an unprecedented scale. However, in order to extract information from social media, current crowdsourcing methods require either the slow, manual process of data sorting or computationally demanding and expensive filtering processes. These restrictions negate the real-time nature and amplifying network capacities of social media to co-produce time-critical information for disaster-affected areas.
In 2013, a group of researchers began working on an open source software named CogniCity to address these challenges. Led by Dr Tomas Holderness and Dr Etienne Turpin, the team began to experiment with a Twitter bot that automatically responds to disaster-related tweets to prompt residents to verify their situation through participatory information sharing.
The AI-assisted chatbot—now operating on multiple social media platforms (including Facebook and Telegram) and colloquially known in Indonesia as Bencana Bot, or Disaster Bot—automatically responds to any social media post containing specific keywords such as banjir (flood), gempa (earthquake) or kebakaran hutan (forest fire) etc. The automated message asks users to confirm if they are really experiencing a disaster and share their on-the-ground observations by submitting a disaster report. Formatted as an interactive micro-survey, the disaster report asks users to give four key pieces of information: the location of the disaster, details of the disaster's severity (such as an estimate of the flood height based on a graphic of a body and vehicle), and a photo and/or free-text description. Instead of passively mining data, this framework for gathering crowd-sourced data through direct engagement via a humanitarian chatbot enables the software to collect millions of confirmed, structured reports directly from street level in real-time, and without the need for expensive and time-consuming data processing. These reports are then immediately mapped onto a live website, PetaBencana.id (Disaster Map), and are freely accessible for all residents, first responders and government agencies to make evidence-based decisions during emergencies.
“Disaster Bot” sends an automated reply to any post on social media containing disaster-related keywords. The reply includes a GIF explaining the PetaBencana.id platform and invites users to confirm their situation by replying to submit a disaster report. Image: Yayasan Peta Bencana
On New Year’s Day 2020, record-breaking rainfall inundated Jakarta with more water than its infrastructure could cope with. As rising waters blocked roads, shut down one of the city’s airports and cut off electricity, millions of residents continued to search for and share up-to-date information. As Jakarta residents turned to social media to share updates about the flood, they were greeted by Disaster Bot and promptly responded to its request to share their observations by submitting flood reports to PetaBencana.id. With these reports, residents alerted each other to broken canal walls, failed pumps, road accessibility and locations of community kitchens. In 24 hours, PetaBencana.id experienced a 24,000% increase in activity as residents actively checked the web-based map to understand the flooding situation, avoid flooded areas, and make decisions about safety and response.
According to Bambang Surya Putra, head of the Emergency Operations Center at the National Disaster Management Agency of Indonesia (BNPB), “PetaBencana has helped the government monitor real-time disaster situations reported by residents that can be followed up with quick response by BNPB and other agencies. Based on this rapid information, the assistance of resources such as equipment, logistics and evacuation teams can have a definite location that must be addressed for humanitarian support intervention".
PetaBencana.id maps disasters in real-time using crowdsourced reports gathered via social media and government agencies. This free and open-source website has been used by millions of residents, first responders and government agencies to understand and respond to emergency situations since 2013. Photos: Yayasan Peta Bencana
Flood data from PetaBencana.id was also used by NASA to calibrate satellite flood maps with crowd-sourced ground observations. “This is the first time that we were able to rapidly generate a flood proxy map from so many satellite scenes and calibrate it with ground observations”, said Dr Sang-Ho Yun, disaster response lead for NASA’s Advanced Rapid Imaging and Analysis team at the Jet Propulsion Laboratory. “This is really important for mapping flood extent in urban areas”.
Harnessing the habitual compulsion to share selfies as a prompt, this novel system for crisis mapping pairs the computational abilities of bots and the psychological dispositions of selfies, to transform social media into a life-saving information collection and dissemination system. Since 2013, PetaBencana has been used by millions of residents to view and share real-time disaster updates, giving rise to the viral hashtag #SelfiesSaveLives. Through considered design of a secure and context-specific software, coupled with extensive on-the-ground research and socialisation, disaster selfies have proven to be transformational in supporting rapid impact, assessment and delivery in Indonesia. With proven impact in Indonesia, CogniCity Open Source Software is also being adopted to power real-time disaster information sharing platforms in Vietnam, India and the Philippines.
According to the International Federation of Red Cross and Red Crescent Societies’ 2020 World Disasters Report, during the past 10 years 83% of all disasters were caused by extreme weather- and climate-related events, such as floods, storms and heatwaves—and this statistic is only expected to rise. With the undeniable trajectory of a warming planet, experts emphasise the need for climate change adaptation: we must focus on the ability to live with increasingly extreme weather. Importantly, the repeated rupture of embankments and levees placed under greater strain under intensifying weather underscores the need for adaptation strategies that go far beyond physical infrastructure. Adaptation strategies must focus on social co-ordination, enabling all residents to respond to sudden hazards.
According to the International Federation of Red Cross and Red Crescent Societies’ 2020 World Disasters Report, during the past 10 years 83% of all disasters were caused by extreme weather- and climate-related events, such as floods, storms and heatwaves—and this statistic is only expected to rise. Image: Yayasan Peta Bencana
It will be critical to involve all residents in their own disaster mitigation efforts and enable them to self-organise in the spirit of gotong royong. By leveraging collective intelligence, PetaBencana provides a method for residents to both, participate in infrastructure monitoring, and take immediate action for response and recovery. By drawing on residents’ expertise and situational knowledge, the platform enables everyone to make time-critical decisions about safety and response through an approach hinged on public engagement in humanitarian infrastructure. The approach supports a framework where the public can work on their own behalf as well as with government agencies to co-manage emergency events, without these efforts cancelling each other out.
As extreme weather events become increasingly ubiquitous, we must create systems that enable every individual to participate in disaster recovery, through non-intimidating, inclusionary, and democratic methods that invite engagement. Instead of resorting to outdated consultancy frameworks (which continue to prove to be extractive, dominating, paralysing and unsuccessful in their abilities to empower communities to act on their own behalf), we must ask how emergent socio-technological assemblages, such as ubiquitous disaster selfies, networked smartphones, and AI bots, can allow us to approach emergency governance differently. Behavioural change is a slow process, but by harnessing existing tendencies—such as the disaster selfie—we can leverage and reroute habits towards preparedness and more meaningful rituals of living and adapting together.
PetaBencana.id is a platform run by Yayasan Peta Bencana [Disaster Map Foundation] as a free, transparent platform for emergency response and disaster management for cities in South and Southeast Asia. PetaBencana.id is part of the USAID CogniCity Open Source Software for Next Generation Community Engagement in Disaster Risk Management Program.
Read more on the Forum Network: The Green and the Blue: How AI may be a force for good, by Luciano Floridi, Professor of Philosophy and Ethics of Information, University of Oxford & Anna Christina (Kia) Nobre, Head of Department of Experimental Psychology, Chair in Translational Cognitive Neuroscience, University of Oxford
Read the OECD report The impact of coronavirus (COVID-19) on forcibly displaced persons in developing countries to learn about the exposure of forcibly displaced persons to health risks and the socio-economic consequences of the pandemic, in particular in fragile contexts.
|Green Recovery||SDGs||Digital Inclusion||Tackling COVID-19|