A PhD candidate launched a transdisciplinary study, with researchers from Harvard and the UN World Food Program, which looks at improving humanitarian needs assessments through a type of AI. He aims to build a new way of collecting data and a toolkit to aid an organization’s ability to respond to a crisis.
At York University, we aspire to better understand the human condition and the world around us and to employ the knowledge we gain in the service of society. One way to do this is through Artificial Intelligence (AI); the integration of AI into society is one of York’s key aspirational areas.
One intrepid grad student, Tino Kreutzer, in the School of Health Policy and Management, is doing just that – finding ways for AI to help society’s most vulnerable. He led a 13-person team whose members include York Professors Lora Appel and Aijun An as well as researchers from Harvard Medical School and the United Nations World Food Program.
The team considered how a form of AI called Natural Language Processing (NLP) could help to asses community needs in humanitarian crises. The findings were published in IBM Journal of Research & Development (2020). Kreutzer’s dissertation work is supervised by Professor James Orbinski, the inaugural director of the Dahdaleh Institute for Global Health Research at York University.
“Current humanitarian assessments fail to capture the complex and rich context in which these crises unfold,” Kreutzer stresses. “There’s an urgent need to bridge the growing gulf between the people affected by humanitarian emergencies and response professionals through improving the quality and quantity of information provided by the affected population.”
This is Kreutzer’s forte and, in fact, the subject of his PhD: Pioneering a novel system for understanding population needs in emergencies through the innovative use of new technology. Kreutzer has more than 10 years’ experience working in the response to humanitarian crises, natural disasters and the Ebola epidemic in West Africa. He now heads KoBo, Inc., which maintains the free KoBoToolbox toolkit used for collecting interview data from people affected by disaster.
Needs assessment critical first step in humanitarian crisis
Needs assessments is vital for program planning, monitoring, evaluation and accountability. Evidence has shown that it profoundly affects an organization’s ability to respond to crises. And yet, Kreutzer contends, this is still a weak link in humanitarian response.
He explains the challenges, “Current approaches often require interviewers to simplify complex, open-ended responses to questions. As a result, the amount and usefulness of information are severely limited.”
How AI could provide opportunities to gain qualitative information?
Kreutzer suspected NLP could be used to provide far-reaching new opportunities to capture qualitative data from voice responses and analyze it for relevant content to better inform humanitarian assistance decisions.
In this vein, his research venture, launched in 2018, consisted of two main components:
- Design a system using NLP to transcribe, translate and analyze large sets of qualitative responses to a humanitarian need assessment survey with a view to improving the quality and effectiveness of humanitarian assistance.
- Anticipate the ethical challenges of introducing this new technology and create a framework to reduce and mitigate these new risks.
Far more insights provided with new methodology, facilitated by AI
The table illustrates the difference that Kreutzer’s team discovered with respect to qualitative information secured by the researchers’ use of NLP, compared to current methods.
Kreutzer describes the advantages of using AI: “Using current methods, qualitative information was difficult to process, labor intensive and time consuming. NLP can provide potentially far-reaching new opportunities to rapidly analyze voice responses for relevant content to inform humanitarian assistance decisions.”
Next step – Pilot project
Kreutzer proposes a pilot phase in a humanitarian crisis for which no transcription and translation models exist. The key stages would include:
- Modifying an appropriate humanitarian assessment questionnaire to include more open-ended questions.
- Generating a transcription and translation model: Collecting speech recordings, etc.
- Creating an analysis model.
- Building a toolkit that can be applied to all humanitarian data collection contexts.
- Providing recommendations for replication and scaling up in other emergencies.
- Publishing all results.
Kreutzer emphasizes that this new tool should work in conjunction with face-to-face interviews, which offer a more personable way to interact with people who have suffered trauma and are struggling to recover.
To read the article, visit the website. To learn more about this project, visit the website.
To learn more about Research & Innovation at York, follow us at @YUResearch; watch our new video, which profiles current research strengths and areas of opportunity, such as Artificial Intelligence and Indigenous futurities; and see the snapshot infographic, a glimpse of the year’s successes.
By Megan Mueller, senior manager, Research Communications, Office of the Vice-President Research & Innovation, York University, muellerm@yorku.ca