Search for: All records

The coronavirus pandemic has fostered an explosion of misinformation about the disease, including the risk and effectiveness of vaccination. AI tools for automatic Scientific Claim Verification (SCV) can be crucial to defeat misinformation campaigns spreading through social media channels. However, over the past years, many concerns have been raised about the robustness of AI to adversarial attacks, and the field of automatic scientific claim verification is not exempt. The risk is that such SCV tools may reinforce and legitimize the spread of fake scientific claims rather than refute them. This paper investigates the problem of generating adversarial attacks for SCV tools and shows that it is far more difficult than the generic NLP adversarial attack problem. The current NLP adversarial attack generators, when applied to SCV, often generate modified claims with entirely different meaning from the original. Even when the meaning is preserved, the modification of the generated claim is too simplistic (only a single word is changed), leaving many weaknesses of the SCV tools undiscovered. We propose T5-ParEvo, an iterative evolutionary attack generator, that is able to generate more complex and creative attacks while better preserving the semantics of the original claim. Using detailed quantitative and qualitative analysis, we demonstrate the efficacy of T5-ParEvo in comparison with existing attack generators.

Behaviours across terrorist groups differ based on a variety of factors, such as groups’ resources or objectives. We here show that organizations can also be distinguished by network representations of their operations. We provide evidence in this direction in the frame of a computational methodology organized in two steps, exploiting data on attacks plotted by Al Shabaab, Boko Haram, the Islamic State and the Taliban in the 2013–2018 period. First, we present $\textsf{LabeledSparseStruct}$, a graph embedding approach, to predict the group associated with each operational meta-graph. Second, we introduce $\textsf{SparseStructExplanation}$, an algorithmic explainer based on $\textsf{LabeledSparseStruct}$, that disentangles characterizing features for each organization, enhancing interpretability at the dyadic level. We demonstrate that groups can be discriminated according to the structure and topology of their operational meta-graphs, and that each organization is characterized by the recurrence of specific dyadic interactions among event features.