Influencing AI Governance

Discussion with the Southeast Asia Hub of the Incubating Feminist AI Project

https://slides.com/jun-e/influencingaigovernance/

Some background

• About the researcher
• Ideas and thoughts in this presentation were generated from the writing of the report, Governance of AI in Southeast Asia published by EngageMedia in Nov 2021

Key Propositions

• We need to understand what exactly we would like to govern, and in what direction
• We need to engage with the policy processes at the level accessible to us

Breaking down the problem

• What is AI?
• What is AI governance?
• How can we influence AI governance? - a Southeast Asian perspective

What is AI?

• A definition, focusing on Machine Learning: (Internet Society, 2017)
  • Algorithms generating algorithms
• Types/Waves of AI applications (Lee, 2018)
  • Internet AI,
  • Business AI,
  • Perception AI,
  • Autonomous AI
• How AI can go wrong and violate human rights: (Raso et al., 2018)
  • Quality of training data
  • System design
  • Complex interactions/unintended consequences

What is AI governance?

• A definition of technology governance, by World Economic Forum (2019):
  • making decisions and exercising authority on the development and diffusion of technology
  • e.g. laws, regulations, incentive programs, institutional frameworks, policies and standards, etc
• Some objectives of governance:
  • Minimising AI security risks (next slide)
  • Influencing direction of AI development

Types of AI Security Risks

• From Newman (2019), Toward AI Security: Global Aspirations for a More Resilient Future
• Three main types of risk mitigation:
  • the risks or opportunity costs of not implementing AI
  • the risks of unintended consequences or unsafe outcomes of AI
  • the risks of AI being used for malicious purposes

SEA: Challenges of AI governance

1. Focus of governance is on rapid adoption and innovation, rather than checks and balances
2. Unclear use cases implicate on data governance
3. International norms need to be adapted for local governance - e.g. GDPR
4. SEA is under-represented in international standards setting
5. SEA countries do not have a strong regional voice
6. State capacity to govern AI technologies is low on average
7. Meaningful public participation in AI governance is difficult - both from top-down, bottom-up perspectives
8. Existing challenges: lack of technical capacity and digital literacy; authoritarian regimes; weak institutional frameworks; availability of quality data; fault lines along religious, racial and other cultural sensitivities

SEA: General Recommendations

1. Anchor AI governance in its societal and application contexts
2. Build constitutionality around AI and data governance
3. Consider multiple levels and sectors of policymaking

4. Enable whole-of-society participation in AI governance

5. Consider existing regulatory frameworks and processes that may be used for AI governance
6. Focus on data governance to reduce AI harms and increase AI benefits

SEA: Ways forward for civil society

1. Increase awareness and participation of civil society in AI governance
2. Build capacity to engage in AI governance (on AI, and on governance processes)
   
3. Form strategic networks and collaborations
4. Conduct more advocacy-based research and documentation on AI applications in SEA
5. Leverage on existing capacities on human rights and community work

Discussion

How do we try to influence AI governance in the context of Feminist AI in Southeast Asia?

1. What are the challenges within our context?
2. What are the possible avenues of intervention?
3. What are possible strategies?