Thursday, January 16, 2020 - 12:30pm to 1:30pm

Location:

Speaker:

For More Information, Contact:

Social media has established a new era of news manipulation. Deceptive news — misleading, falsified and fabricated content — is routinely originated and spread on social platforms with the intent to create confusion and widen political and social divides. In this talk I will review computational approaches to detect deception online, explain model predictions and access model robustness, measure the immediate spread of deceptive content and quantify user reactions to it. I will start by presenting linguistically-infused neural network models for deception detection and an in-depth linguistic analysis across broad deceptive news categories. These categories are grouped based on an intent to deceive, ranging from disinformation to misinformation. I will then show how neural models can be extended to make predictions in multilingual and multimodal settings, and present an interactive tool to explain multimodal predictions, highlighting challenges and model limitations. I will present key findings on evaluating neural model robustness in cross-domain and cross-lingual setting, and report model susceptibility to adversarial inputs. Next, I will attribute and characterize user behavior while engaging with deceptive news. I will present the insights about the immediate spread of deceptive news by characterizing the audience and measuring speed and scale of spread, to uncover who shares deceptive content, how quickly, how much and how evenly. I will flesh out user reactions to deceptive news, distinguishing the reactions of users identified as bots versus humans. Finally, I flesh out a few ideas to take advantage of deep translation and generation models to enable better defense against digital deception.

About Dr. Volkova

Dr. Svitlana Volkova is a Senior Scientist at the Data Sciences and Analytics group, National Security Directorate, Pacific Northwest National Laboratory. She develops AI models to explain and predict human social systems and behaviors as they relate to national security challenges in the human domain. Her scientific contributions cover a range of topics on natural language processing, applied machine learning, deep learning, and computational social science. Approaches developed by Dr. Volkova advance understanding and effective reasoning about extreme volumes of dynamic, multilingual, and diverse real-world social data. 

Prior to joining PNNL in October 2015, Svitlana interned at Microsoft Research at the Natural Language Processing and Machine Learning and Perception teams. She was awarded the Google Anita Borg Memorial Scholarship in 2010 and the Fulbright Scholarship in 2008. She received her PhD in Computer Science in 2015 from Johns Hopkins University where she was affiliated with the Center for Language and Speech Processing and the Human Language Technology Center of Excellence. For more information about Dr Volkva’s research, visit her website at https://www.cs.jhu.edu/~svitlana/