IGF 2019 WS #305
Why We Need to Move Beyond Ethics for Governance of AI

Relevance to Theme: The conversation around Artificial Intelligence in the past years has shifted from how AI will solve the problems of the world to the ‘ethics crisis’ of AI - the lack of ethical considerations in building AI systems and the harms they can cause to individuals and communities. In response to sustained criticism of harmful AI systems, almost all big technology corporations and academic institutions have started initiatives to foster ‘ethical AI’. However, while ‘ethics’ has become a buzzword, the ethical implications of any technology are not obvious, nor are they uniform across those engaged in, implicated by, or left out of technological development and use. Just calling for ‘ethics in AI’ is not enough, it is important for the ethics in AI discourse to engage with critical social science approaches that have for long studied and attempted to intervene in complex socio-cultural, economic and legal problems. It is also noteworthy that there is no consensus as such on what it might mean to build ethical AI systems, perhaps pushing us to think what values and whose values and interests are being prioritized when a general ethical framework is developed.

The joining of two broad and dynamic terms ‘AI’ and ‘ethics’ creates both ambiguity and uncertainty - and worse, suspicion that these terms are being used to duck regulation, accountability, and responsibility.The AI and ethics narrative severely waters down hard regulations on AI, painting a vision of positivity, and looking at problems that AI can cause as “concerns”. In the global south, the mainstream narrative around AI is that of a problem solver. To the extent that the National Strategy for Artificial Intelligence in India envisions India as a data marketplace and an AI test bed. As regulation tightens in the north, global south countries will become a destination to test new technologies, based on limited ethical frameworks and self regulation. Driven by corporate funding especially from the Global North, this process is aided by the diversity of a large population of the Global South and significantly reduced costs of performing trials here, giving a new sheen to the expropriation of and experimentation of the Global South through AI-enabled technologies. The use and development of AI needs to be analysed through the lense of Responsible Consumption and Production, which is one of the Sustainable Development Goals (SDG 12).There is a need to balance the SDG of Innovation (SDG 9) with that of climate action (SDG 13), well being (SDG 3), gender equality (SDG 5) and decent work practices (SDG 8).

In countries of the Global South without laws governing data and the use of the same for AI frameworks, optional or ineffective ethical frameworks might result in these being substituted for meaningful legislation and justiciable rights. This session will aim to look at how data governance frameworks could address the questions on how to bring about accountability, fairness, explainability, scrutability, and representativity into the use of data and the algorithm.The session will also discuss how inclusion, stakeholder participation, and sensitization of data-driven decision making can be developed to democratize accountability-seeking processes.


Relevance to Internet Governance: The introduction of new technologies at various points in history has been largely hailed as watershed moments. Moreover, technology’s advent in various socio-cultural settings can have different consequences for different groups of people, the environment and the labour force. The advent of technology as pervasive as AI can have the effect of shifting power and increasing vulnerabilities under the garb of ‘development’ and ‘empowerment’, especially among underserved communities. This push towards ethical AI without looking at human rights, feminist, environmental, and labour concerns that go into the design and deployment of AI technologies needs to be critically addressed. The stakeholders signaling ethical principles often fail to assess human and environmental costs associated with these technologies. The issues that will be discussed in this session go beyond fairness, transparency and accountability and look at other conditions such as scrutability, explainability, auditability and historical responsibility. The session will also seek to explore beyond mitigation of harms and look at participatory processes which includes user empowerment and autonomy. Though the conversation will be based on the global trends in AI policies, the specific questions and observations will be drawn from a Global South perspective. The session will seek to explore the role of the global south as a ‘marketplace’ of resources, data, and labour for the creation of technologies, with ethical frameworks as the only regulatory lever.