(Note: the time of the talk is different from the regular CLIP colloquium time)
Recent years have seen an increase in the use of digital data for various decision-making purposes in the context of urban computing and smart cities. However, as such decision-making tasks are becoming more autonomous, a critical concern that arises is the extent to which such analyses are fair, inclusive and whether such technology can work when applied to real-world diversities. In this presentation, I will demonstrate the potential algorithmic biases that are surfaced due to the training data in two scenarios of the demographic detection of park visitors and obstacle detection for visual impaired navigation. I will also discuss our lab’s current research focus on designing federated learning models with keeping privacy and participation in mind.
Afra Mashhadi is an Assistant Professor of Computing and Software System in the University of Washington Bothell. She received her PhD degree in mobile computing from University College London (UCL). Previous to her current appointment she was a senior research scientist at Bell Labs in Ireland for 5 years and a consultant for UN Global Pulse Lab. Her research focus is in Ubiquitous computing where she is interested in developing mathematical and computational models that leverage the proliferation of sensors and breakthroughs in machine learning to model social dynamics of human behavior. More specifically her research focus is on sensing, modeling, understanding and predicting human behavior using the ‘digital traces’ that are generated daily in our online and offline lives. Results of her research have been published in top-tier conferences (WSDM, CHI, CSCW, Ubicomp, ICWSM) and journals, and trialled as part of multiple deployments in European projects and private entities such as WebSummit 2015. She also acts as a committee member for ACM-W, promoting and broadening participation of women in computing.