Library Element Report

Notes from the AI frontier | Applying artificial intelligence for social good

Uploaded by RRI Tools on 04 December 2018

Notes from the AI frontier | Applying artificial intelligence for social good. By Michael Chui, Martin Harrysson, James Manyika, Roger Roberts, Rita Chung, Pieter Nel, and Ashley van Heteren. Discussion Paper McKinsey Global Institute November 2018


Artificial intelligence, while not a silver bullet, could contribute to the multi-pronged efforts to tackle some of the world’s most challenging social problems. AI is already being leveraged in research to tackle societal “moon shot” challenges such as curing cancer and climate science. The focus of this paper is on other social benefit uses of AI that do not require scientific breakthroughs but that add to existing efforts to help individuals or groups in both advanced and developing economies who are experiencing challenges or crises and who often live beyond the reach of traditional or commercial solutions. We assess the AI capabilities that are currently most applicable for such challenges and identify domains where their deployment would be most powerful. We also identify limiting factors and risks to be addressed and mitigated if the social impact potential is to be realized.

Through an analysis of about 160 AI social impact use cases, we have identified and characterized ten domains where adding AI to the solution mix could have large-scale social impact. These range across all 17 of the United Nations Sustainable Development Goals and could potentially help hundreds of millions of people worldwide. Real-life examples show AI already being applied to some degree in about one-third of these use cases, ranging from helping blind people navigate their surroundings to aiding disaster relief efforts.

Several AI capabilities, primarily in the categories of computer vision and natural language processing, are especially applicable to a wide range of societal challenges. As in the commercial sector, these capabilities are good at recognizing patterns from the types of data they use, particularly unstructured data rich in information, such as images, video, and text, and they are particularly effective at completing classification and prediction tasks. Structured deep learning, which applies deep learning techniques to traditional tabular data, is a third AI capability that has broad potential uses for social good. Deep learning applied to structured data can provide advantages over other analytical techniques because it can automate basic feature engineering and can be applied despite lower levels of domain expertise.

These AI capabilities are especially pertinent in four large domains—health and hunger, education, security and justice, and equality and inclusion—where the potential usage frequency is high and where typically a large target population would be affected. In health, for example, AI-enabled wearable devices, which can already detect potential early signs of diabetes through heart rate sensor data with 85 percent accuracy, could potentially contribute to helping more than 400 million people afflicted by the disease worldwide if made sufficiently affordable. In education, more than 1.5 billion students could benefit from application of adaptive learning technology, which tailors content to students based on their abilities. 

Scaling up AI usage for social good will require overcoming some significant bottlenecks, especially around data accessibility and talent. In many cases, sensitive or monetizable data that could have societal applications are privately owned, or only available in commercial contexts where they have business value and must be purchased, and are not readily accessible to social or nongovernmental organizations. In other cases, bureaucratic inertia keeps data that could be used to enable solutions locked up, for example in government agencies. In most cases, the needed data have not been collected. Talent with high-level AI expertise able to improve upon AI capabilities and develop models is in short supply, at a time when competition for it from the for-profit sector is fierce. Deployment also often faces “last mile” implementation challenges even where data and technology maturity challenges are solved. While some of these challenges are nontechnical and common to most social good endeavors, others are tech-related: NGOs may lack the data scientists and translators needed to address the problem and interpret results and output from AI models accurately. 

Large-scale use of AI for social good entails risks that will need to be mitigated, and some tradeoffs to be made, to avoid hurting the very individuals the AI application was intended to help. AI’s tools and techniques can be misused by authorities and others with access to them, and principles for their use will need to be established. Bias may be embedded in AI models or data sets that could amplify existing inequalities. Data privacy will need to be protected to prevent sensitive personal information from being made public and to comply with the law, and AI applications will need to be safe for human use. The continuing difficulty of making some AI-produced decisions transparent and explainable could also hamper its acceptance and use, especially for sensitive topics such as criminal justice. Solutions being developed to improve accuracy, including model validation techniques and “human in the loop” quality checks, could address some of these risks and concerns.

Stakeholders from both the private and public sector have essential roles to play in ensuring that AI can achieve its potential for social good. Collectors and generators of data, whether governments or companies, could grant greater access to NGOs and others seeking to use the data for public service and could potentially be mandated to do so in certain cases. To resolve implementation issues will require many more data scientists or those with AI experience to help deploy AI solutions at scale. Capability building, including that funded through philanthropy, can help: talent shortages at this level can be overcome with a focus on accessible education opportunities such as online courses and freely available guides, as well as contributions of time by organizations such as technology companies that employ highly skilled AI talent. Indeed, finding solutions that apply AI to specific societal goals could be accelerated if technology players dedicated some of their resources and encouraged their AI experts to take on projects that benefit the common good. 

The application of AI for societal benefit is an emerging topic and many research questions and issues remain unanswered. Our library of use cases is evolving and not comprehensive; while we expect to build on it, data about technological innovations and their potential applications are incomplete. Our hope is that this paper sparks further discussion about where AI  capabilities can be applied for social good, and scaled up, so that their full societal potential can be realized


  1. Mapping AI use cases to domains of social good
  2. How AI capabilities can be used for societal benefit
  3. Six illustrative use cases
  4. Bottlenecks to overcome
  5. Risks to be managed
  6. Scaling up the use of AI for social good

Related Resources