Introduction

As AI progress proceeds at breakneck speed, companies, governments and international bodies are recognising that new norms and more inclusive and equitable approaches are needed to measure the impact of these technologies, mitigate risks of harm, and ensure their responsible development and use. Building on multidisciplinary research on AI governance, Chatham House, with the Office of the UN Secretary-General's Envoy on Youth (OSGEY) and Google, hosted a panel discussion to foster an inclusive and informed public debate, and policy engagement, on how collectives - countries, communities and companies - can frame and guide the responsible development of AI technologies.

The discussion focused on several questions, including: 

• What is the role of ‘powerful’ actors, such as governments and the private sector, in the governance of AI development and use? 

• What community-led efforts are in place to govern the responsible development of AI? 

• How should communities be engaged and what incentives are there for their participation in AI governance? 

• What are some of the main considerations communities ought to take into account when devising and implementing governance approaches and solutions in the AI space?

Establishing a common understanding around key themes, questions and risks, and ensuring diverse and systematic input regarding responsible AI development through this session, will ultimately contribute to global efforts at ensuring that these technologies are built for all, by all, and empower all.

AI Governance: A multi-faceted approach

The discussions included perspectives from the private sector, international organisations, and civil society on what ongoing efforts are in place to engage communities in the responsible development and deployment of AI technologies and their subsequent governance. 

Key initiatives from across sectors and geographies have emerged, including the OSGEY’s work in engaging with the youth on AI governance issues: the youth are, indeed, arguably the largest and most connected generation. They are bound to inherit current policies, decision-making, and systemic issues; hence there is a critical need - and desire, from the youth - to ensure their representation in decision-making processes and, ultimately, a sense of control over their digital future. Greater inclusivity is a necessary response to the ongoing lack of trust and assurances with regard to the development and deployment of these technologies. The successful inclusion of the youth in AI governance will require intergenerational support and multistakeholder allyship. 

A couple of notable industry-led initiatives have also emerged from the discussions:. Google Seach’s Quality Raters isa select group of individuals across the globe trained under Google’s set of guidelines to help stress-test and refine the quality of search queries. This programme is a key example through which technology companies can, by proxy, engage communities and the value of established processes in the roll-out and testing of products and subsequent changes. 

In addition, research and development in voice recognition being predominantly in English, ‘non-priority’ languages therefore tend to neither be reflected in products, nor data to even develop and train these programmes in the first place. In response, Mozilla’s Common Voice initiative seeks to overcome this limitation through extensive building and engagement with communities, one notable example being with Kiswahili-speaking groups. These engagement opportunities take many forms, including: competitions for students; partnership with grassroot groups across Kenya; the ‘community champions’; as well as collaboration with linguists to capture the many dialects and variants in Kiswahili. 

Key considerations for the way ahead

It is clear that in order to leverage and maximise the benefits of AI technologies, governance solutions ought to think of their implications both for individuals and, most of all, collectives. As initiatives to engage with communities in the responsible development and deployment of AI technologies emerge, key considerations have emerged to inform future efforts: 

Effective community engagement requires enabling environments conducive to meaningful solutions. There is a particular emphasis for those on the ‘powerful’ end in the new social contract (i.e., governments and big technology companies) to facilitate such environments through, for example, capacity building, established processes, incentivisation and creating opportunities to address other, pressing issues such as climate change, healthcare access, and disability exclusion. 

The definition of communities is fluid in time and space. Individuals can belong to more than one community at the same time; and communities span across geographies, interests, languages and shared history, among others. As such, there is a need to reconsider and re-evaluate the social contract in light of AI development and the role and place of communities.

AI must not be mistaken as a monolithic technology and is highly contextual. The nature of the technology’s impact - and the outcome of subsequent policy responses - is highly contextual and will change depending on what is at stake. For example, the human rights implications of AI technologies will be different depending on the technological application, as well as the wider societal context in which it is being deployed. As such, governance solutions must be sustainable, agile, and look into both existing and long-term risks, and strive to foster both horizontal and vertical opportunities. 

There is no easy solution, safe experimentation is key. Concrete implementation of governance measures requires extensive research and experimentation on defining what works, and ensuring that solutions are agile, trustworthy, and meet the needs of communities. These experimentations must, however, be done in a safe environment: many lessons can be drawn from user research practices. 

Trust is multi-faceted. There are two particular aspects to establishing trust. Trust by communities; and trust in the product and its outputs. Establishing trust by communities can be done through greater ownership over the products’ development and the many benefits they bring, as well as through meaningful engagement and implementation of policy measures and solutions. Establishing trust in the tool requires, first, reflections on a number of elements and how to adapt technical solutions accordingly. One notable example pertains to users’ perception and how labels affect a product’s trustworthiness in the eyes of the users. The question of trust is particularly relevant as AI technologies are at risk of deployment and use for disinformation campaigns, an element of increasing importance given upcoming election cycles, increasing polarisation, conflict, and risks of harm against vulnerable communities.