Blog Phil-algorithms—the end of human-led philanthropy?

The possibilities offered by Artificial Intelligence (AI) are gaining the attention of a growing number of industries and sectors. One of them is philanthropy. The ability to harness machine intelligence and software to analyze masses of data can help many human endeavors operate more efficiently and successfully. That possibility is proving very appealing to philanthropy service leaders, who want to harness more of AI’s innovative capabilities to support social good.

AI has many applications that can help address some of the world’s most pressing issues. For example, it can improve our capacity to collect, predict, or automatize intelligence at scale. This can offer benefits in many sustainability-linked areas, such as the following:

Healthcare: AI is already being used to develop more precise and/or lower cost diagnostics. In Kenya, Neural Labs Africa are using AI to accelerate the detection of respiratory diseases. The technology uses deep learning and computer vision to screen medical images for radiologists and hospitals in real time to identify diseases such as pneumonia and tuberculosis.¹

Education: The Om3ga program in Serbia provides a speech to text solution, integrated with generative AI and a chatbot builder in Slavic languages,² to help children with disabilities communicate more effectively.

Environment and climate: Restoring forests plays a vital part in tackling climate change. The RESTOR program founded by the Crowther Lab at ETH Zurich and powered by Google allows anyone to analyze the restoration potential of any place on Earth, thereby helping guide naturalists to plant the right species in the right places. It uses high-resolution satellite imagery and machine learning models to predict where trees could naturally grow, providing people with scientific data about the ecosystem and enabling a restoration project to be more effective and progress to be tracked.

These are a few current examples. AI could in the future be used to run virtual experimentation to predict the feasibility and efficacy of a program intervention and mitigate any risks before it is implemented.

The technology has also revolutionized the approach towards fundraising. Today, organizations are using voice interface or chatbots to guide and engage with potential donors when visiting their webpage by answering their questions in a human-like conversation with tailored responses—and even to make donating easier. One example is 'Say It Now', which uses Natural Language Processing to enable people hear radio ads on their Alexa-enabled smart speaker to simply respond verbally using voice commands to donate directly to the charity of their choice.³ AI can also be used by fundraisers to define when to solicit a donor, tailoring the topics so as to increase the chance of receiving a donation—in a similar manner to how internet companies tailor news feeds and advertising based how individuals consult homepages. By identifying these patterns, AI can act as a key tool to pinpoint potential recurring donors, what triggers their donation, and how to best engage them.

Giving is simple, but giving effectively is not. The most effective forms of philanthropy require the use of rigorous research, based on a rational use of data and evidence. Yet most donations today are based on emotions or tax incentives.

AI has the potential to make giving more effective by applying machine learning onto existing grantmaking data to help identify patterns and predict what programs have more chance to be successful in the long term, which can therefore inform future funding.

While AI has the potential to make acts of giving more effective, it is not a cure-all. One risk is that AI can be manipulated in ways that humans wouldn't be. It could be programed in a way that it does not incorporate certain aspects, or its analysis could be skewed to benefit the preferences of particular people. Ultimately, AI is only as effective as the information it has, and the way in which it is programmed. It requires huge volumes of accurate data and sophisticated impact standards if it is to make accurate predictions.

It can be easy for AI to fail to receive sufficiently rigorous levels of understanding, leading to poor outcomes. One well-known example is an algorithm used in law enforcement that had the tendency to predict recidivism disproportionately more in people of color;⁴  another is when some algorithms behind image databases associate women with domestic chores and men with sports.⁵

In fact, studies have shown that AI technologies do not just replicate such biases—at times, they amplify them. In the philanthropy sector, the danger of such poor programming would be that ineffective or harmful philanthropic programs are recommended at scale to potential donors.

Hence, it is imperative that the underlying data is as unbiased as possible, to build a strong foundation for the machine learning models and ensure data accuracy and authenticity. This type of programming requires a great deal of research and a strong methodology so that it can serve as an effective starting point for the machine to learn and provide guidance to the donors.

Another challenge is that AI is not universally helpful for sustainable causes. One study showed that the technology can enable the accomplishment of 134 targets across all the Sustainable Development Goals, but its usage may simultaneously inhibit 59 other targets.⁶ Additionally, applying AI consumes a lot of energy, thus presenting a sustainability challenge.

While AI is not a silver bullet, it could prove to be a game changer, for good or ill.

The technology’s enormous potential could, if well-implemented, help donors to be more impactful, but it will only achieve this if it can gain vast amounts of accurate and impartially gathered data, along with a well-thought-out evaluation framework. That will not be easy.

However, the potential of AI is great enough that organizations may be incentivized to help create the tools needed to make it successful. Ultimately, humans will be responsible for the investments they make. AI is just a tool to help them make those decisions. Given both its possibilities and the growing desire to implement AI across all fields, it is important that the non-profit sector gets a grip on it, takes its seat at the table, and explores how this technology could enhance its impact.

¹ Source: https://unicefventurefund.org/story/neural-labs-using-ai-accelerate-medical-imaging-diagnosis-respiratory-diseases (last consulted 28.11.2023)
² Source: https://www.unicef.org/innovation/venturefund/ai-ds-learning-health-cohort (last consulted: 28.11.2023)
³ Source: https://sayitnow.ai/case-studies/donate-now (last consulted 28.11.2023)
⁴ Source: https://www.bu.edu/articles/2023/do-algorithms-reduce-bias-in-criminal-justice/ (last consulted 12.12.2023)
⁵ Source: https://www.ey.com/en_lv/workforce/are-we-coding-gender-bias-into-our-future (last consulted 12.12.2023)
⁶ Vinuesa, R., Azizpour, H., Leite, I. et al. The role of artificial intelligence in achieving the Sustainable Development Goals. Nat Commun 11, 233 (2020). https://doi.org/10.1038/s41467-019-14108-y

The author is grateful for feedback from: Jackie Bauer, Lisa Michel, Richard Morrow, Richard Mylles, William Nicolle, Mike Ryan, Gayatri Suri, Nalinia Tarakeswar.

For further information, please visit www.ubs.com/institute-disclaimer