Participatory Approaches in AI Development and Governance Paper released by IIT-Madras

Posted 11 Nov 2024

2 min read

The Paper identifies how participatory approach in AI can improve the outcomes of the AI algorithm and enhance its fairness.

About Participative AI (PAI)

It refers to the involvement of a wider range of stakeholders than just technology developers in the creation of AI systems.
- Core tenets of PAI are derived from participatory governance. (see box)
Need: Progress in AI and its deployment by public and private actors, like Facial Recognition Technology in Law enforcement, etc.

Counter unilateral, top-down decision making in AI deployment: Addressing potential contentious breakdowns in implementation.
Inclusion and Fairness: Mitigate risks like bias, discriminatory output, etc., to communities where an AI system is deployed.
Feedback loops: In flagging technical glitches and post deployment impact assessment.
Enhance trustworthiness of AI Systems: Ensuring minimal false positives and false negatives causing a more enthusiastic adoption of such systems.

Co-optation: Domination by select dominant actors to serve their vested interests.
Limited participation of non-experts: Existing AI governance models primarily see the participation of experts like industry representatives, bureaucrats, select civil society, etc.
Participatory washing and tokenism: Stakeholder participation merely done for formal compliance.
Transparency Paradox: Information shared about algorithms can be misused by malicious actors.

Land Acquisition Act, 2013: Social impact assessment study in collaboration with affected families.
Forest Rights Act, 2006: Gram Sabha as the statutory institution due to its participatory and democratic nature.
Nagoya Protocol on Access and Benefits Sharing: Benefits arising from the use of traditional knowledge be shared with indigenous communities.