Page 51 - AI for Good - Impact Report

AI for Good

Sustainable Development Goal 10: Reduced

Inequalities

Reduce inequality within and among countries
SDG 10 currently has only 1 out of 10 targets (10.b resources
flow for development) on track. 290 Before the pandemic,
the forecasted inequality stood at -0.8% but has now risen
to 4.4%, posing challenges for countries in achieving their
SDG agenda. 291 Furthermore, discrimination based on age,
gender, religion, race, or belief affects one in six people
globally. In 2023, there was a record high of 35.8 million
refugees, and over 8,000 migrant deaths were recorded
worldwide. 292

AI and SDG 10 Impact
According to a study on the impact of AI on
The connection between AI and SDG 10 is extensively documented in vari- SDG 10, AI could act as an (positive) enabler
ous AI use case repositories from the UN: 14 use cases out of 40 in AI for for 90% of the targets and act as an inhibitor
Good: Innovate for Impact, 293 and approximately 110 use cases out of 408 (negative) for 70% of the targets. (Nature
Communications, 2020)
in the UN Activities on AI. 294 These use cases encompass a range of topics,
including the monitoring of weather events or conflicts for at-risk commu- Use case 1
nities. 295 By consolidating information from diverse sources such as official Aggregating various data points from
reports and social media, AI can assist governments and NGOs in effectively social media to assess critical situations for
tracking situations in different communities to identify risks and take appro- minorities and support in case of problems
priate action. Additionally, AI has the potential to aid refugee support by or risks.
helping to identify welcoming communities, creating chatbots to assist, and
optimizing refugee camps, among other applications. 296
While these use cases are valuable in illustrating the synergy between AI
and SDG 10, it is crucial not to overlook the associated risks for inequali-
ties. Firstly, the majority of AI solution ownership is concentrated in a few
countries (primarily in the Global North) or large companies, leading to the
centralization of benefits in specific locations 297 and unequal distribution of
ownership and value, potentially exacerbating inequalities. 298 For instance,
the training of AI models using data from individuals or small and medi-
um-sized enterprises (SMEs) often occurs without financial compensation for link
sharing their content and work. 299 Secondly, the development of AI solutions Use case 2
typically reflects the needs and values defined by the developer, potentially
resulting in significant biases, particularly as much of the technology is Using an AI platform to help refugees find
concentrated in specific geographic locations. 300 the best information in their time of need by
leveraging chatbots.
While this matter has been discussed from a gender perspective, it also
applies to country diversity and underrepresented communities. Lastly, while
AI can monitor at-risk populations for their benefit, it can also be misused
by malicious organizations to achieve the opposite outcome, endanger-
ing these populations. 301 There is a legitimate concern that AI could be
employed for surveillance of humanitarian efforts, perpetuating hate towards
minorities, and providing tools to monitor populations and restrict freedom
of movement when necessary.
link
Key Considerations for Stakeholders Use case 3
Training AI to recognize biases in historical
• Ownership sharing: To reduce the risk of monopoly on technologies, data to develop new solutions that are more
new business models should be considered, 302 where value is shared inclusive.
differently to minimize the increase in inequalities and reward all contrib-
utors.
• User-centric: Align the development of AI solutions with the “Recom-
mendation on the Ethics of Artificial Intelligence” from UNESCO to
ensure that human dignity is maintained. 303

link

41