Omdena Open Source AI for Impact Program for Civil Society and Government

Partner: UN-Habitat

The problem

Prosperous cities seek to increase their green areas for better air quality and improved quality of life for their populations. Green spaces in cities mitigate the effects of pollution and can reduce the urban heat island effect. At the same time, land use change in urban areas leads to a reduction in tree cover, contributing to the loss of biodiversity. Accordingly, it is important for cities to monitor their progress in maintaining and increasing their tree cover and green areas. The monitoring will enable city authorities to measure the environmental impacts of urban development against their mitigation measures, as well as support city policy actors in decision-making. 

Envisioned solution

The AI solution should involve the extraction of data from satellite imageries hosted on cloud-based platforms (e.g., the Earth Engine’s public data catalog), and within defined city boundaries, generate statistics on two urban indicators related to environmental sustainability. To enable comparison of city statistics, the project will utilize the urban boundaries generated through the harmonized city definition approach (JRC-UrbanCentresDatabase).

These indicators are:

a. Change in green Areas per Capita as defined in the Global Urban Monitoring Framework (UMF-47). The methodology involves estimation of a city area under vegetation cover for several time periods e.g., the year 2000, 2010, and 2020; the indicator has 2 key metrics: change in green areas over time, and change in per capita green areas over time, which factors the changes in city population.

b. Change in Tree Cover as defined in the Global Urban Monitoring Framework (UMF-48). The methodology involves estimation of the city area under tree cover for several time periods e.g., the year 2000, 2010 and 2020, and analysing the change over time.

Regional focus

The project has a global scope, but pilot regions may include countries experiencing rapid land use change/urbanization, particularly Sub-Sahara Africa and Asia.

Partner: UN-Habitat

The problem

Many frameworks on the performance of cities generate urban profiles at the city scale, providing limited or no information on the performance of different city sub-units such as districts, wards, zones, settlements, or blocks. The transformative focus of the Agenda 2030 of Leaving no one Behind aligns with local policies of many cities, their intervention focus being the reduction of spatial inequalities.

Mapping spatial inequalities within the city guides the identification of vulnerable areas, which can be expressed on a continuous scale of vulnerability. Many forms of spatial vulnerabilities such as poor access to basic services, lack of green cover, crime and insecurity, vulnerability to disaster risks, access to opportunities, and access to cultural infrastructure among others, have statistics that can be standardized for comparison and mapped – where data is available. The individual layers of vulnerability as well as the composite layer combining the layers are useful for spatially targeted intervention by city administrators and other actors.

In extension, cities may prepare profiles for their settlements based on a set of indicators to guide city residents in understanding their settlements, and service providers in setting their intervention priorities.

Envisioned solution

The AI solution should identify key characteristics of urban vulnerability and generate data layers on them from city-level data, open data sources, extraction of data from imageries, and/or complementary sources. The analysis could involve the creation of surface maps for each form of vulnerability and aggregating the layers to generate a composite ‘city vulnerability layer’. For useful results, the resolution of the data must be good to enable precise identification of vulnerable locations within the city fabric. Identification of locations with multiple vulnerabilities may guide decision makers on deteriorating locations, especially when monitoring is done at a consistent temporal scale.

Regional focus

Priority cities for this project include cities with huge intra-urban inequalities, including large cities in Sub-Sahara Africa, Asia, and Latin America.

Partner: ACET

The problem

We want to better understand the disconnect between skills and jobs in Africa, particularly digital skills.

According to the AfDB, Africa’s youth population is expected to double to over 830 million by 2050, but currently out of Africa’s population of nearly 420 million aged 15 to 35 years, one third are unemployed, another third are not secure in their jobs and only one in six is in wage employment. More youth than ever are graduating from schools and universities, but are not finding jobs.

Why? The social problem we want to help avoid is youth becoming discouraged and turning to militancy, insurgency, and risky illegal immigration. The potential impact is better alignment between skills youth are attaining and the skills that employers need, which will lead to more and better jobs, fewer people facing poverty through unemployment, and greater human wellbeing across the continent.

Envisioned solution

We envision an AI solution that can aggregate various public big data to help better assess the demand and supply of jobs and job seekers. Ultimately the outcomes can influence curriculum, learning approaches, public-private partnerships for job training, higher education policy, and industrial policy. The AI solution can compare supply data from unemployment data, graduation data, jobs data, etc., and demand data from at online job postings, numbers and types of private sector companies in specific sectors, etc. to identify gaps or skills mismatches. It can also be used to compare skills training offered online, apprentice programs and on-the-job training opportunities. Finally, the AI solution can use sentiment analysis and natural language processing to have a better understanding of how companies, schools and individuals feel about the education they receive, the jobs available, the skill-levels of job seekers, and policies that incentivize greater skills/jobs alignment.

Regional focus

As a prototype we could either focus on one country (Ghana) or we could focus on six countries – Côte d’Ivoire, Ethiopia, Ghana, Niger, Rwanda, and Uganda. These are countries where ACET has already undertaken a multi-country study on Strengthening Education and Learning Systems to Deliver a 4IR-Ready Workforce, and therefore provides a unique opportunity to leverage current research with new AI solutions. The project will also allow experts to better understand what public and private data are available across countries from different sub-regions and languages, varying types of educational systems, and disparate focus on digital skills in particular.

Partner: Save The Children

The problem

The project is designed to reduce Online Sexual Exploitation and Abuse of Children (OSEAC). With a 15,000% rise in online Child Sexual Abuse Materials (CSAM) online from 2005 to 2020, it is clear that online child violence is growing exponentially. In 2021, the National Center for Missing and Exploited Children’s CyberTipline received 29.3 million reports of CSAM, making 2021 the worst year on record for online child sexual abuse.

 A primary way that adults with a sexual interest in children or those who wish to harm them in other ways is through online grooming. As described by Sørensen, 2015; Greijer et al., 2016,

“Grooming is a multidimensional phenomenon in which an adult aims to solicit a child into a seemingly voluntary interaction with the intention of sexually abusing that child.” In a study Save the Children published last year, Grooming in the Eyes of a Child (Juusola et al., 2021), we found that children who are the object of grooming often do not realize what is happening so they do not recognize they are in danger until they are being extorted into providing increasingly harmful imagery or even to meeting an online predator in person.

The solutions

Our goal is to stop online violence against children by deploying an accurate classifier to identify grooming behavior in online chats with children. Once suspicion of grooming reaches a threshold based on its similarity to the training data, it will trigger an action, which may differ depending on the platform it is deployed on and the objectives of the intervention.  As examples, we may warn the child through the chatbot without alerting the groomer, call a moderator, or shut down the chat entirely.

In 2020, Save the Children US collaborated with Omdena to address online violence (https://omdena.com/projects/children-violence/). Of the various products that were generated from the sprint, the most promising was a classifier algorithm using Natural Language Programming to identify online grooming combined with a chatbot that can warn the children that they may be chatting with a groomer. Since then, a team of three engineers associated with the original project has continued to refine the technology. The core team now wants to expand on the work to build an industry usable solution at scale.

From the original challenge, we have a large dataset of more than 800,000 lines taken from the Perverted Justice project, a project from 2003 to 2019 that used online volunteers as decoys to entrap predators that sought to contact minors to obtain sexual images or videos from them or to meet them in person. During the challenge and afterward, we tagged much of the training data with labels, such as male or female, predator or victim, and level of risk of the conversation, but the data still requires extensive processing, and in particular, we need to improve and systematize the way judge and annotate the level of risk. In addition to the data we already have, we are actively attempting to obtain additional databases of online grooming chats from a variety of sources, such as law enforcement agencies.

The project deliverables/milestones will include:

1. Build on existing data and further annotate additional sentences. Target is to achieve 100,000 annotated sentences with risk levels (non-risky, potentially risky, or risky). If new data is made available by law enforcement, annotate that data.

2. Look for and scrape user data from online resources.

3. Create a language model [Classification] to detect grooming behavior by labeling it as non-risky, potentially risky, or risky.

4. Test the data on various models and provide ablation studies.

5. Deploy the system as an API.

6. Make the API a stand-alone chrome extension that predicts labels in an impromptu manner [The Grammarly execution process is the best example to relate with the final deliverable]

Partner: Equitech Futures / Omdena Nairobi Kenya Chapter

The problem

In order to address potential biases in this self-reporting mechanism, the contribution of independent observation systems is being increasingly sought (IPCC, 2019). Our focus here is on the direct observation of carbon dioxide (CO2) emissions from space and on the quantification of CO2 emissions from this observation independently.

NASA’s second Orbiting Carbon Observatory (OCO-2) polar satellite (Eldering et al., 2017) is one of the best existing instruments for the retrieval of column-averaged dry-air mole fraction of CO2 (XCO2). It observes the clear-sky and sun-lit part of the Earth with the footprints of a few km2 (1.29 km × 2.25 km) gathered in a ~10 km wide swath for each orbit, 35 particularly suitable for informing natural CO2 budget at the continental scales

We intend to use NASA’s OCO2 and OCO3 Satellite data and publicly available data on critical CO2 emitting sectors e.g. power plants, steel mills , cement plants, atmospheric “spillover” from agricultural and forest fires, traffic emissions, demographic and economic variables etc to build an AI model that predicts the contribution from each sector.

This CO2 emissions model could be used to track progress for large and small cities, sub-regions or project areas within Kenya which can be used to support decarbonization efforts

The goal

• Develop a model that can be used to track progress for large and small cities, sub-regions or project areas within Kenya.

• Provide Insights on CO2 emissions in Kenya over time.

• Provide prediction on project/sectoral CO2 emissions.

• Create dashboards to visualize XCO2 measurements by region/overtime.