Researchers propose methods for effective sandboxing in DPI development

A research paper has listed a number of factors that must be considered by governments for the proper conduct of sandboxes in testing how digital public infrastructure (DPI) can be secure, inclusive, and trustworthy in real-world situations.

A sandbox in the context of DPI is a restricted space where governments, regulators, and innovators can test new technologies, policies, and services before launching at scale.

According to authors of the paper published by Datasphere Initiative, sandboxes are an indispensable tool in balancing speed and safety, and support the development of DPI in a way that curbs security vulnerabilities, privacy breaches, and public distrust.

In order to conduct sandboxes the right way as part of the development of digital ID, digital payment or data exchange systems, the four authors of the paper propose that one key thing is to first establish national frameworks that facilitate the putting in place of clear legal and institutional structures to authorize and govern sandboxes.

In another point, the authors recommend an inclusive stakeholder engagement process which involves civil society organizations, local groups and other underrepresented segments of the population, in a bid to make sure the DPI being developed responds to the different needs of these groups.

Other factors to prioritize, the paper mentions, include defining clear objectives and putting the right metrics to evaluate sandbox outcomes and determine readiness for scaling, creating pathways for transitioning from a sandbox to real-world rollout while envisioning any potential challenges, and addressing limitations and recognizing the fact that real-world deployments can spring up peculiar challenges.

Critical importance of sandboxes in DPI development

The paper also examines the critical role sandboxes play in the development of DPI systems which truly secure, inclusive and interoperable, and which can effectively meet real-world needs.

Apart from mitigating risks and ensuring a safe implementation environment, the paper highlights that sandboxes are vital in building collaborative ecosystems as they “offer a structured space where government entities, civil society organizations, technologists, and private companies test interoperability, values alignment, and co design in action.”

The authors mention the example of the Maisha Namba digital ID in Kenya where failures in the ID system have been linked to a lack of collaboration at the design phase where design flaws would have been sorted out. The Maisha Namba project has faced fierce criticisms for several reasons including its reported exclusionary nature and the absence of an impact assessment.

The authors state that while sandboxing comes with its own challenges like resource constraints and the challenge of replicating the success in a real-world situation, the process remains a strategic one in speeding up DPI development while keeping the risks under checks. They hold that sandboxes should reflect a holistic strategy that prioritizes equity, adaptability, and long-term governance as countries look to expand their DPI ecosystems.

Sandboxes for “agile” AI deployment

Like DPI, AI development and deployment also requires sandboxes which can contribute to identifying and addressing ethical, cultural, and even infrastructural difficulties.

This is event more crucial for Africa where the push for AI adoption and how it can shape the continent’s digital transformation efforts is getting stronger. This explains the holding of a workshop dedicated to an AI Sandboxes Co-Creation Lab in Kigali, Rwanda, ahead of the Global AI Summit on African which took place early last month.

The move by Datasphere Initiative rallied stakeholders from several domains such as government, private sector, academia, development organizations and civil society to scrutinize how sandboxes can serve as agile governance tools for developing and implementing responsible AI systems.

During the event, which is part of a series, participants were unanimous that the three major areas where sandboxes can better contribute to “an Africa-led framework for agile and inclusive AI governance” include data governance and digital sovereignty, infrastructure and contextual solutions, as well as skills and capacity development.

Discussants emphasized the need to address challenges identified and also “defined clear goals, use cases, success indicators, and key players essential for implementing effective sandboxes.”

Acknowledging responsibility as a key aspect in designing an AI development sandbox, workshop participants came up with a number of guiding principles that can guarantee adherence to ethics and inclusivity.

Article Topics

Africa  |  digital identity  |  digital inclusion  |  digital public infrastructure  |  interoperability  |  research and development