By: Aakanksha Joshi & Becky Duane, IBM


Introduction

It was an honor to have the opportunity to meet young leaders from 76 countries coming together at the 27th Session of the Youth Assembly in New York City to engage in actions towards the United Nations Sustainable Development Goals (SDGs). Technology and Inclusion was one of the four core areas where changemakers focused their energy and we had a chance to lead a workshop on Trustworthy Artificial Intelligence (AI) and Social Impact, where we provided an overview of IBM’s 7 pillars of Trustworthy AI: fairness, performance, explainability, confidence, transparency, privacy and adversarial robustness. The image below shows how we defined each of these pillars: 

pastedGraphic.png

In this workshop, participants formed groups to focus on a specific SDG. They rolled up their sleeves to work together and identify how AI could be applied to advance that particular SDG. Teams were asked to consider which Trustworthy AI pillars were key to keep in mind when pursuing their projects. Check out a few projects highlighted below.

Trustworthy AI Project Highlights

How might we use Trustworthy Artificial Intelligence (AI) to achieve SDG 11: Sustainable Cities and Communities?

Traffic congestion makes cities unsustainable in several ways, from increasing air pollution/CO2 emissions to diminishing regional economic and personal health as individuals spend non-productive time stuck in traffic, away from both work and their loved ones.

Our team envisions employing AI to determine work schedules at factories with the goal of minimizing impact to the environment while maximizing manufacturing production output.

The AI would conduct a combined analysis of live traffic data from Google Maps alongside daily production targets of factories to determine individualized shift times for workers, as well as map out the safest, most sustainable route for individuals heading to and from their worksites. For example, certain workers might be asked to depart for work at 7:30am by electric carpool using the freeway, others at 10:47am by public transportation, and others at 11:52am using local bike lanes.

An important challenge would be overcoming privacy concerns, which could be addressed by ensuring residential address data is anonymized, or by only using aggregate data to suggest shift times for individuals city-wide (rather than just an individual factory), in cooperation with both private sector and government. We hope that over time, our AI would improve its recommendations and seek further ways to minimize waste, increase quality of life for citizens overall, and achieve SDG 11.

Team Members: Fareeha Salahuddin, Jennifer Duran, Jonte Pietsch, Mariana Reyes, Tarnish Bhardwaj, Zeyad Tolba

How might we use Trustworthy Artificial Intelligence (AI) to achieve SDG Goal 3: Good health and well-being?

I am currently working at Trust for Youth and Child Leadership India and we are working towards building a Youth Helpline Mobile Application where we are hoping to use AI.

Trust for Youth and Child Leadership, India is working towards a goal of Zero Suicide youth suicide rates in Puducherry by 2030. Currently, Puducherry has the 4th highest suicide rates in the country and we have received 700+ calls on the helpline regarding psychological counselling for youth, education loans, entrepreneurship and job support, natural disaster response, and career guidance. Currently, the helpline is active between 9am to 5pm and we, at TYCL, have been grappling with how to provide this support to youth 24/7. We aspire to build an AI chatbot to respond to youth reaching out on the helpline after work hours and if it is a high risk situation, there could be a triage system to re-route the call to a counselor. However, we need to be mindful of the privacy of the youth and fairness as the biases that could build into the system as this is a sensitive area to use AI. The data that would be needed to build this AI system would be around the existing callers on the helpline if they give consent to use the data without revealing their identity. This application and AI system would be built locally by tech experts from the local community in Puducherry.

Select Team Members: Sakshi Nikhil Shah, Selena Lee

How might we use Trustworthy Artificial Intelligence (AI) to achieve SDGs?

BirdBox is a legal tech startup that aims to democratize the production of reliable audiovisual evidence, through blockchain, for concrete or imminent situations of emergency.

We came up with the idea inspired by George Floyd’s story. As notably known, George Floyd was an African-American man who was murdered by a police officer in Minneapolis, Minnesota. With this tragic case of racism, BirdBox team identified that George Floyd was not an isolated case, but actually, he was the result of an oppressive system that daily dehumanized black people through violence and marginalization. Through bibliographic review, data analysis, and empiric research, we identified that there are a lot of many other George Floyds around the world. However, just a small percentage of them are reported to both media and public institutions.

Birdbox is an App for video/audio recording and location tracking in case of emergencies. Our technology allows users to produce reliable evidence on their smartphones and use it to prove crimes committed against them. On our App, the integrity of the recorded content is ensured by blockchain technology to guarantee the immutability and indestructibility of the recordings. Therefore, we protect the victim against the seizure of evidence and against coercion to delete the recorded content.

Select Team Members: Maurício Pimenta Ferreira

Conclusion 

As we can see in the examples above, there is a strong connection between social impact through sustainable development goals and trustworthy AI. These projects are widening the horizons when it comes to applications of AI but are cognizant of the risks that AI can pose and are moving forward with the concepts of trustworthy AI in mind.