AI Governance

How Public Participation Can Improve AI Governance: vTaiwan’s Initiatives

The civic tech project vTaiwan has gained international attention for its design of effective public participation mechanisms, which has led to the facilitation of over 20 legislative reviews in Taiwan. Most notably, in 2019, the platform hosted an online deliberation leading to regulatory adjustments for the ride-hailing platform Uber in Taiwan. In 2023, vTaiwan was relaunched as a community-based project. Now, they use their proven participatory approach to tackle the challenge of inclusive AI governance.

Why Is Public Participation Necessary for AI Governance

Artificial intelligence (AI), like other technologies, offers both transformative benefits but also poses risks. To manage the risks, governments across the world are developing norms and regulations, such as the Artificial Intelligence Risk Management Framework by the National Institute of Standards and Technology (NIST) in 2023. NIST emphasizes that governance is essential to integrate technological development with organizational values and principles. For example, the importance of balancing innovation with human rights protection for governments, and optimizing model performance while managing harm for businesses.

Public participation plays a crucial role in this process, ensuring that AI models meet diverse needs and that developers understand broader community concerns. It also helps the public understand AI technology, reducing distrust and enhancing transparency. For governments, effective public participation mechanisms are key to evaluating the tools needed for AI governance and addressing challenges like privacy violations and rights infringements.

vTaiwans initiatives address the necessity of public participation in AI development, showing that mechanisms built by citizens can effectively and meaningfully collect public opinions for developers' reference.

vTaiwan: A Successful Digital Democracy Platform

vTaiwan emerged after the Sunflower Movement as an initiative from the g0v (gov-zero) community with the aim to establish effective public participation mechanisms through collaboration with civil society. Despite its achievements with successful legislative reviews, vTaiwan faced challenges after 2019, such as member attrition, the government opting for other public participation mechanisms, and the impact of the pandemic, leading to a halt in activities and the necessity of a community transformation. Fortunately, due to the project's open-source nature, willing community contributors revived it.

vTaiwan advocates for human-centered public participation, leveraging various digital tools to connect different stakeholders. Its mechanism consists of four stages: Proposal, Opinion Collection, Deliberation, and Realization. During the Proposal stage, the community focuses on the issue to be deliberated and identifies stakeholders of that issue. In the Opinion Collection stage, vTaiwan will openly call for and gather public opinion, and  identify the most agreed-upon and contentious perspectives among citizens' opinion. During vTaiwan’s previous process of deliberating laws and regulation, it would also invite relevant authorities to respond and systematically compile and prepare information for the Deliberation stage. During Deliberation, stakeholders will be invited to discuss the most contentious points to form a consensus. Finally, in the Realization stage, the discussions are compiled and the preliminary consensus is implemented in the form of draft regulations or reports.

The vTaiwan mechanism's flexibility in connecting various digital tools makes it a platform to experiment with different digital tools and discuss and evaluate the effectiveness of these digital tools when it comes to participation and deliberation. Recently, vTaiwan community members experimented with using generative AI for transcript recording, opinion clustering, and analysis during deliberation and collection activities, and they are also applying these new methods and the strength of such flexibility to integrate public participation into the development of artificial intelligence.

vTaiwan 2023 owards: Challenges and Findings the Deliberation of AI Risk

In May 2023, OpenAI announced the "Democratic Input to AI" initiative, selecting ten teams to develop mechanisms for making AI more democratic, each receiving a $100,000 grant. vTaiwan, in collaboration with the UK-based Chatham House, was selected with their "Bridging the Recursive Public" project. The project aims to combine vTaiwan's opinion collection and discussion mechanisms to create a database of AI policies and risk management guidelines for government and corporate reference. Considering the rapid technological advancements, vTaiwan aims for enabling this database to recursively update and compare AI policies and guidelines over time.

However, vTaiwan faced three main challenges when initiating deliberation on AI:

1. Tackling Knowledge Gap on AI among Stakeholders

A major challenge is ensuring that participants have sufficient knowledge of AI to contribute meaningfully to discussions. Given the diverse backgrounds of stakeholders, many lack the technical expertise. To address this, preparatory materials must be accessible for everyone and informative, covering key aspects of AI development. This often requires the involvement of experts who can break down complex issues for a broad audience to understand.

2. Balancing Depth and Breadth of Input

The second challenge is navigating the depth and breadth of opinions. Collecting opinions from participants first often results in uneven content in depth and proficiency, while consulting experts first could raise participation barriers. Balancing this in the process is critical for meaningful AI public participation.

3. Moderating Complex Discussions

AI-related issues are often interconnected, such as systemic bias resulting from training data bias, which relates to privacy in data use and acquisition. It’s crucial to discover how to systematically divide issues during deliberations in order to allow focused and in depth discussions on specific topics without overlooking their interconnections with other topics at the same time and build consensus.

To facilitate the deliberation, the vTaiwan community employed the pol.is tool to collect public opinions, identifying the most divisive topics for further, in-person discussions, and Talk to The City to process and analyze transcription of the discussions.

Despite the challenges, the process yielded several consensus points. Among them:

1. AI systems should have higher cultural sensitivity from development to deployment.
2. Open-source AI systems should be encouraged, ensuring the coexistence of diverse AI models.
3. Ethical concerns surrounding AI, such as the source code and training topics, should be publicly accessible for scrutiny.

In physical deliberations of this topic, participants tackled contentious topics like personal data protection and the influence of major tech companies. They concluded that Taiwan's current Personal Data Protection Act is insufficient to address AI's impact and that the content of AI models should not be monopolized by tech giants.

The demand to take part in a public discourse around AI governance was evident. Participants saw the deliberations as a rare opportunity to influence AI developers like OpenAI but voiced concerns about whether their input would be taken seriously by those in charge of development.

Reflecting on the process, the experience highlighted the importance of having expertise in organizing deliberative activities. Yet, the discussions provided a platform where people from all walks of life—engineers, legislators, gender groups, and indigenous communities—could engage in shaping the future of AI. This inclusive approach offers a model for governments and organizations seeking to involve the public in AI policymaking.

vTaiwan's Ongoing Efforts to Democratize AI

Looking ahead, vTaiwan hopes governments will harness public participation to build trust in technology and achieve effective risk management when implementing emerging technologies like AI. Despite ongoing challenges in rebuilding connection with the government, the project has successfully reorganized its community. In the future, vTaiwan aims to forge stronger connections with civil society groups and projects focused on digital rights, law, participation, and deliberation. The goal is to expand the application of its AI governance work while fostering collaboration with new partners. Current initiatives also include collaborating with the Open Parliament Project to gather public opinion on parliamentary reform and conducting global surveys on digital participation tools and best practices. Through these efforts, vTaiwan is exploring innovative ways to utilize technology to enhance public participation on AI governance and to contribute to democratic engagement.

*This article is written by Mr. Peter Jia Wei Cui, Contributor of vTaiwan community.