Every time someone trains an AI system, the quality of that system depends on the data labels—the annotations that tell the algorithm what it is looking at—yet no one has standardised how those annotations are made, recorded, or checked for bias. This Fellowship tackles a blind spot in AI development. Researchers and companies pour resources into collecting massive datasets, but the annotations themselves remain an afterthought. Nobody knows, for example, whether a set of image labels was produced by experts in a controlled setting or crowdsourced from tired workers on a Friday afternoon. The project also addresses newer forms of annotation—such as the rewards used in reinforcement learning or the causal links in graph-based models—that do not fit the old static-label model. If successful, the work will create transparent protocols that let data curators document how annotations were produced, let practitioners automatically assess annotation quality, and let end users give feedback that actually improves the system. The deeper goal is to reframe annotations as a two-way interface between humans and AI—a channel through which people can express their preferences directly into a learning system’s objectives. That shift could make AI systems more reliable and trustworthy, particularly in sensitive applications where hidden biases in training data currently erode confidence.
View original technical description
With the prevalence of data-hungry deep learning approaches in Artificial Intelligent (AI) as the de facto standard, now more than ever there is a need for labelled data. However, while there have been interesting recent discussions on the definition of readiness levels of data, the same type of scrutiny on annotations is still missing in general: we do not know how or when the annotations were collected or what their inherent biases are. Additionally, there are now forms of annotation beyond standard static sets of labels that call for a formalisation and redefinition of the annotation concept (e.g., rewards in reinforcement learning or directed links in causality). During this Fellowship we will design and establish the protocols for transparent annotations that empowers the data curator to report on the process, the practitioner to automatically evaluate the value of annotations and the users to provide the most informative and actionable feedback. This Fellowship will address all these through a holistic human-centric research agenda, bridging gaps in fundamental research and public engagement with AI. The Fellowship aims to lay the foundations for a two-way approach to annotations, where the paradigm is shifted from annotations simply being a resource to them becoming a means for AI systems and humans to interact. The bigger picture is that, with annotations seen as an interface between both entities, we will be in a much better position to guide the relation of trust in between learning systems and users, where users translate their preferences into the learning systems' objective functions. This approach will help produce a much needed transformation in how potentially sensitive aspects of AI become a step closer to being reliable and trustworthy.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know