Architectures for Autonomy: Towards an Equitable Web of Data in the Age of AI

Abstract

Today, the Web connects over half the world's population, many of whom use it to stay connected to a multiplicity of vital digital public and private services, impacting every aspect of their lives. Access to the Web and underlying Internet is seen as essential for all—even a fundamental human right [7]. However, many contend that the power structure on large swaths of the Web has become inverted; they argue that instead of being run for and by users, it has been made to serve the platforms themselves, and the powerful actors that sponsor such platforms to run targeted advertising on their behalf. In such an ad-driven platform ecosystem, users, including their beliefs, data, and attention, have become traded commodities [13]. There is concern that the emergence of powerful data analytics and AI techniques threaten to further entrench the power of these same platforms, by putting the control of powerful and valuable new capabilities in their hands rather than the users who produce the data [10]. The fear is that it is giving rise to data and AI monopolies [2,6]. Individuals have no long-term control or agency over their personal data or many of the decisions made using it. This may be one reason we are witnessing a so called Renaissance of Ethics - a plethora of initiatives and activities that call out the range of threats to individual autonomy, self-determination and privacy, the lack of transparency and accountability, a concern around bias and fairness, equity and access in our data driven ecosystem. This keynote will argue as the remaining half of the world's population comes online, we need digital infrastructures that will promote a plurality of methods of data sovereignty and governance instead of imposing a ’single policy fits-all’ platform governance model, which has strained and undermined the ability for governments to protect and support their citizens digital rights. This is an opportunity to re-imagine and re-architect elements of the Web, data, algorithms and institutions so as to ensure a more equitable distribution of these new digital potentialities. Based on our existing research we have been developing methods and tech-nologies pertaining to the following core principles: informational self-determination and autonomy, balanced and equitable access to AI and data, accountability and redress of AI/algorithmic decisions, and new models of ethical participation and contribution. The technology that underpins the modern web has seen exponential rates of change that have continuously improved the capabilities of the processors, memory and communications upon which it depends. This has enabled huge amounts of data to be linked and stored as well as providing for increasing use of AI. A variety of projects will be described where we sought to unlock the potential of this increasingly powerful infrastructure [1, 4, 5, 9]. The lessons learnt through various efforts to develop the Seman-tic Web [8] and the insights gained through the release of open data at scale will be reviewed [11]. We will review our attempts to understand how the blending of humans, algorithms and data at scale results in social machines whose emergent properties results in behaviour and problem solving which any of the individual elements would not have been able to achieve [12]. Understanding these emergent properties of the web was one of the motivating factors behind the establishment of Web Science [3]. We will briefly review the prospects for Web Science. The importance of data as infrastructure to enable wide spread innovation, accountability and trusted reproducible science will be stressed. Recent work will be described that seeks to promote an equitable and balanced Web environment in which privacy can be upheld and better mutualities realised. Developments in technical and institutional architectures that could underpin an Ethical Web of data will be outlined.