CompAcctSys

Focus areas

Research Themes

  • 01

    Algorithmic 'reviewability' and 'contestability'

    Contrasting with the general focus on 'explanation', we consider that which is necessary to facilitate the review, understanding, scrutiny and challenge of algorithmic and automated decision-making (ML) systems.

  • 02

    Meaningful transparency

    Exploring requirements and mechanisms for facilitating the meaningful inspection and interrogation of socio-technical systems and their behaviour(s). One focus is on auditable augmented/mixed/virtual reality systems, as well as the Internet of Things, and the use of machine learning in various contexts.

  • 03

    Algorithmic supply chains

    Investigating how responsibility, control and opacity issues occur across the components of modern AI systems — which consists of multiple actors and multiple components.

  • 04

    Data protection and privacy enhancing technologies (PETs)

    Considering issues, critiques, management and interventions regarding personal and confidential data.

  • 05

    Algorithmic bias / fairness

    Focusing on trade-offs in various context, and methods and tooling for supporting practitioners.

  • 06

    Platforms and online harms

    Exploring the design, use and abuse of platforms in perpetuating harms. The current focus is on social media (recommender systems), virtual user-spaces (games/XR), and cloud services, including 'AI as a Service'.

  • 07

    Decision provenance

    Considering how tracing the flow of data can be leveraged to assist accountability in complex, automated and ML-driven environments.

  • 08

    Compliance and rights engineering

    How systems can be better built to be (demonstrably) compliant with legal obligations, and to account for rights – fundamental, group and individual.

  • 09

    Centralised v. decentralised infrastructures

    Considering the potential of data management and compute infrastructures, and their legal, regulatory and policy implications. Currently looking at personal data stores and data trusts.