School of Computer Science PGR Seminar Rhayssa Silva De Oliveira and Bocheng Su

Rhayssa Silva De Oliveira will present Before the Scroll: Detecting Dark Patterns in the Design Phase

Abstract: Dark patterns, also known as deceptive patterns, are design choices that steer, pressure, or mislead users into actions they would not otherwise take, whether that means sharing more personal data than intended, staying on a platform far longer than planned, or agreeing to terms they never quite meant to accept. First named by Harry Brignull in 2010, they have since been studied extensively across different academic fields and have begun to attract serious regulatory attention. Yet they remain woven into the everyday software most people rely on, with social media as perhaps their most fluent and most profitable medium. Most existing work studies dark patterns after release, in interfaces that have already reached millions of users. This seminar asks a different question. Could dark patterns be caught earlier, during the design phase of software development?

Bio: Rhayssa Oliveira is a first-year PhD student supervised by Dr Dharini Balasubramaniam. Her research investigates how dark patterns might be identified earlier in the development process, before they reach the public. Before starting the PhD, she worked in social media management for small businesses and creators, an experience that sparked many of the questions her research is now trying to answer.

Bocheng Su will present Local Geometry of k-Nearest-Neighbor Graphs in High-Dimensional Space

Abstract: Graph-based approximate nearest-neighbor search (ANNS) is widely used in industrial systems because it offers a strong trade-off between recall and latency. In this work, we take a first step toward a theoretical analysis of graph-based ANNS on k-nearest-neighbor graphs under high-dimensional Gaussian data. More precisely, using large deviation techniques, we derive bounds for the population neighbor-distance threshold and show that conditional local neighbors concentrate around a deterministic center.

Bio: Bocheng Su is a first-year PhD student in the School of Computer Science, supervised by Peter Macgregor. His research focuses on the theoretical analysis of approximate nearest-neighbor search on high-dimensional data. His current interests include graph geometry and the development of new graph-based ANNS algorithms.