Ethical Frontiers: Steering AI Towards a Principled Future
Combined Session
Wednesday, June 05, 2024 17:30—18:30
Location: B 05
Wednesday, June 05, 2024 17:30—18:30
Location: B 05
In the last 10 years machine learning has become ubiquitous and touches all lives in ways that was unimaginable before. The machines can make decisions that required considerable human effort at a much faster speed and reduced cost with a little human oversight. As a result, machines don’t just have a higher than before influence in shaping our lives but are also under increased scrutiny by both regulators as well as user rights advocates.
The adage “with great power comes great responsibility” has long been used – from French revolution to superhero comics. It has never been truer as the great power that machine learning wields is now in the hands of almost anyone making a software product. It ranges from giving people access to the funds that can alter their lifepath, medical diagnosis that can increase their life expectancy or reduce it dramatically to their social media feed that cannot just provide them the content that keeps them engaged, but also polarise their beliefs by feeding them information that reinforces their existing notions.
Given this, it’s important to look at AI from lens of Responsible and Ethical AI where the talk will go through Fairness (discrimination-free algorithms), explainability (from data, model, business, risk and counterfactuals), Privacy (design and algorithms), accountability, sustainability. The talk will give real life examples from different domains and suggestions on implementation methods of RAI. And extend the concept of RAI to much talked about Gen AI (ChatGPT, BARD etc) as well.
The rapid evolution of AI has grabbed attention, including a focus on its impact on human needs and aspirations, and on potential harms. AI, is, however, not the only recent digital information technology that affects the human condition. The focus on AI offers a prompt for examination of the human condition generally and its relationship to the social interaction and information environment, and the potential for impact of digital information technologies – beyond AI systems - on human rights.
We’ve all heard of “Secure-by-Design,” “Privacy-by-Design,” “User-centered Design,” “Human-centric Design” and other similar invitations to consider human-related elements at the inception of information system initiatives. Building off of the “Human-Centric Digital Identity” whitepaper recently published by 13 non-profits, Elizabeth and Henk argue that many well-intentioned identity solutions – including many designed with “user-centric” or “human-centric” principles at their core - have had unanticipated and unintended negative human rights consequences.
While such consequences may be understandable in the context of such fast-moving technologies like biometrics, AI, and growing server capacities to deliver change at pace and at scale, the potential for rapid erosion of human rights, which reflects the human condition generally, is also accelerating. In these high-pace, high-uncertainty environments, it is imperative that Human Rights and human values underpin design (and implementation) decisions. Helpfully, the core of human rights aspirations are shared globally, so the attention to such broadly shared rights has the potential to help us identify corresponding global standards for “duties” to which parties across the species will voluntarily self-bind, and by so doing breathe life into aspirational “rights” to make them realities.
This session will discuss some of the tensions that practitioners grapple regarding project and product life-cycles, and suggest approaches to help manage the paradoxes. The discussion will explore the human rights impacts of quickly-evolving technologies and the challenges of predicting and managing future outcomes of complex digital information network systems, the organization and operation of which, reflect myriad business, operating, legal, technical, and social goals and aspirations.
In these high-pace, high-uncertainty environments, it is imperative that Human Rights and human values underpin design (and implementation) decisions. Building on their talks at EIC 2023, the panelists delve deeper into how we can manage the many variables and support human rights goals and expectations in digital systems.. They will highlight various new industries to consider emerging best practices and discuss how Human Rights can be most effectively operationalized by governments and private entities.
