Empowering philanthropy to lead the ethical and inclusive AI revolution
Artificial Intelligence (AI) is increasingly ubiquitous in our lives and societies: We use it to plan our trips, buy groceries and even to find peers. Having now reached sophisticated levels in a wide variety of domains ‒ including language processing, facial and object recognition ‒ the scale, spread and speed of change brought about by AI are unprecedented.
With its ability broadly consisting of making machines do things that would require intelligence if done by humans, AI is simultaneously driving advances in the human condition and giving rise to profound new challenges. Even though it is equipping us with sophisticated tools to combat infectious diseases, regulate traffic, and anticipate natural disasters, problems of bias, transparency of algorithms, and data privacy and security cannot be ignored.
This leads to an important and pressing debate on AI: How do we ensure that the benefits of AI’s use outweigh the costs and the associated risks? And how do we create trust that the technology is built and used according to ethical principles and this in an inclusive and human-centred way? In other words, how can we encourage and govern the development of Ethical and Inclusive AI (EIAI) for the common good? And finally, who could take a leading role in promoting EIAI?
We believe that philanthropy can and should play an important role in addressing these questions. Philanthropic organisations (POs), with their knowledge and expertise of both the private and public sectors, are uniquely positioned to lead the EIAI Revolution. This potential, however, has remained largely untapped, and POs have been loudly absent from the global and national debates on AI. This absence is further reflected by the fact that the philanthropic sector is still largely Al-illiterate. This, however, could be an advantage, as it gives philanthropy the potential to interact with AI on different fronts.
AI & Philanthropy: The Three Pillars Approach
The philanthropic sector’s relation to AI is twofold. On one hand, it should consider what AI can do for philanthropy, and on the other hand it should establish what philanthropy can do to promote ethical and inclusive AI and its governance. Specifically, POs should not only embrace the advances that AI technologies can bring to their everyday operations, but also position themselves as an ethical reference point for the promotion and governance of EIAI.
To date, very little academic knowledge is available to drive the debate around this two-fold approach. To fill this gap, we have launched at the University of Geneva, in collaboration with the Geneva Center for Philanthropy, the research project Empowering Philanthropy to Lead the Ethical and Inclusive AI Revolution. The project brings together stakeholders from the world of academia, civil society and industry in a multidisciplinary debate on the role of philanthropy for AI. The aim of this research project is to generate solid academic contributions that can provide a reference point for POs that need to expand their knowledge on the potential of AI as well as for those that want to have an active role in promoting the adoption of EIAI. To achieve these goals, this research endeavour is constructed upon three main pillars around which to develop theoretical and practical knowledge on philanthropy and AI:
1. AI for Philanthropy
AI can be a resourceful and impactful tool for POs, regardless of size and scope. This includes the use of AI to analyse, inform, and predict donor behaviour; to monitor and evaluate impact; and to automate administrative tasks to support everyday operations. The successful use of AI for similar purposes can be witnessed in the private sector. Nonetheless, an invisible wall still appears to exist between the private sector and philanthropy’s use of AI.
While some bigger POs have started to integrate AI into their systems, this continues to very much remain a far-away reality for both smaller and long-standing organisations. The reasons are multiple. First, the scarcity and cost of “AI talents”, added to the competitive market for this kind of work. Second, the lack of knowledge on the benefits and potential of AI, and the multiple uses of AI for good. Third, the reluctance to integrate what is considered to be an industry tool. This is exacerbated by the multiple challenges surrounding AI, such as bias, which have the effect of scaring off the sector. Last, and probably most important, the understanding that integrating AI is not a straightforward process, but rather a journey that will require resources and time. Digitalisation, data availability and quality, and the standardisation of data continue to remain critical barriers to overcome.
The development of a national and/or international data standard for POs appears to be a concrete solution that would not only allow for better data efficacy and collaboration, but also expose the sector to AI and address its hesitancy. Associations such as Philea and Swiss Foundations would be strategically positioned to build this bridge.
2. AI for Philanthropic Impact
Informed by the field of AI4SDGs, which promotes technological solutions to societal challenges and the realisation of the SDGs, this pillar aims to foster the use of AI for the purpose of enhanced philanthropic impact. Given the reluctance of the sector, a first critical step is the need to demonstrate the potential benefit of AI through promising use cases. This should be done through a comprehensive mapping study, in order to highlight what has already been done and by who. This hard data will allow for a comprehensive examination of not only the starting point, but also potential trajectories for the incorporation of AI and its use for impact.
The case study of the ICRC, and its use of AI for humanitarian applications, is of great relevance. But the focus should also be placed on the industry sector, which given its greater resources, has fostered important advances in the niche of AI for good. Key examples include google.org, IBM science for social good, and Microsoft AI for Earth. Importantly, these case studies will highlight the importance, and thus push for, the standardisation of AI.
3. Philanthropy, AI and Ethics
Today one of the largest gaps in the AI governance system is that there are too many principles and not enough operationalisation or actual implementation. Moreover, the forcing of untailored ethical frameworks on different sectors has been shown to be an unsustainable and at times ineffective solution. This calls for the development of a tailored ethical framework for POs, built on previous efforts yet customised to respond to the needs and characteristics of the philanthropic sector. Notably, this framework should be highly practical, scalable and measurable in terms of impact.
This framework should be developed around, as well as promote, best practices on strong and safe AI. It should take into consideration key concerns surrounding the use of AI, such as bias, security risks, data protection and privacy, and transparent algorithms. The output of this pillar should ultimately be to enable POs to understand and stimulate the development and implementation of operational EIAI guidelines and provide understanding and guidance in grantmaking in such areas as the promotion of digital rights and diversity as well as ensuring ethical and transparent algorithms.
Overall, the objective of the Empowering Philanthropy to Lead the Ethical and Inclusive AI Revolution research project is to create momentum and catalyse research and active involvement of POs to build the necessary academic body of literature and subsequent practical know-how that will lead to the endorsement of EIAI. Its goal is to empower the sector to have an authoritative voice in global debates and policy developments. For philanthropy to play a leading role in the unlocking of AI’s full potential as a force for good, cooperation and collaboration are key. We thus call on the sector to join our efforts and support the construction of an avant-garde, AI-literate philanthropic sector.