Much is known about social engineering strategies (SE) during the attack phase, but little is known about the post-attack period. To address this gap, we conducted 17 narrative interviews with victims of cyber fraud. We found that while it was seen to be important for victims to act immediately and to take countermeasures against attack, they often did not do so. In this paper, we describe this "delay" in victims' responses as entailing a period of doubt and trust in good faith. The delay in victim response is a direct consequence of various SE techniques, such as exploiting prosocial behavior with subsequent negative effects on emotional state and interpersonal relationships. Our findings contribute to shaping digital resistance by helping people identify and overcome delay techniques to combat their inaction and paralysis.
Recently, there is a growing trend of using generative AI systems and tools for fostering and protecting online collaborative communities. Yet, existing AI tools may introduce new risks and even harm to diverse communities’ online safety. How to better maximize the novel opportunities of AI and mitigate its emerging risks and harm for our future online safety is a critically needed discussion for the HCI community. Featuring experts from both industry and academia, the goal for this panel is to promote interdisciplinary, community-wide discussions and collective reflections on important questions and considerations at the unique intersection of AI and online communities, including but not limited to: how the design of AI systems may discourage existing online harm but also invite new online harm in various online spaces; how different populations, cultures, and communities may perceive and experience AI’s new roles for their online safety; and what new strategies, principles, and directions can be envisioned and identified to better design future AI technologies to protect rather than harm various online communities.
Automated decision-making algorithms are increasingly prevalent in consumer-facing industries, particularly in insurance risk assessments. The traceability of these decisions is crucial for trust, acceptance, and individual autonomy. While the General Data Protection Regulation (GDPR) grants individuals the right to information about such decisions, the implementation of this right remains under-researched from a usable privacy perspective. This study employs a qualitative exploratory approach with 12 participants exercising their right to be informed about automated decision-making with German household insurers. Through interviews and observations, we investigate consumer requirements and prevailing implementation practices. Our findings unveil actual process design practices that may undermine the usability and efficacy of this data subject right. By identifying these concerns and correlating them to existing deceptive patterns, our research contributes to usable security by alerting process designers, data protection authorities, and enterprises to the significance of user-centric implementations. Furthermore, this study advances research on GDPR data subject rights, emphasizing the need for secure and usable interfaces in the context of automated decision-making systems. Our work highlights the practical challenges of safeguarding usable implementation of regulatory compliance in the realm of data protection.
Advanced Driver Assistance Systems, including Driver State Monitoring like Drowsiness Detection Warning (DDW) and Driver Health Warning (DHW), are increasingly integrated into vehicles to enhance road safety. However, factors of user acceptance of such assistance remains understudied. This paper presents an exploratory user survey on the acceptance of both warning systems using the Driver Monitoring Acceptability Scale (DMAS) questionnaire. Results show high perceived ease of use but varying general acceptability and concerns, with no significant differences between drowsiness and health warning. Willingness to pay is low, and responses among experienced users are polarized. On a conceptual level we contribute by providing a German translation of a shortened questionnaire as a candidate instrument for future surveys. Our study advances the empirical understanding of drowsiness detection and health warning acceptance, offers insights into DMAS applicability, and outlines future research for improving automotive health system design.
Deepfakes - synthetically generated or manipulated audiovisual content realistically depicting people - pose growing risks to public trust, security, and information integrity. As generation techniques rapidly advance, automated detection has improved yet still generalizes poorly to real-world conditions, and humans struggle to reliably distinguish real from manipulated content. While human-AI collaboration has shown potential for more robust verification, current detection models remain opaque, typically offering only confidence scores or post-hoc attribution visualizations. No systematic characterization exists of which intermediate outputs - including feature maps, embeddings, latent states, attention weights, and other structural by-products of model inference - could be surfaced to support human-AI joint decision-making.This paper provides an exploratory analysis of intermediate outputs across contemporary deepfake detection approaches. We analyzed 28 open-source codebases and executed all models to identify and characterize intermediate outputs generated during inference. Our findings reveal a diverse landscape of internal representations that could serve as design material for human-centered explainable AI in deepfake detection. The study is descriptive and exploratory in nature and does not evaluate the practical effectiveness of such outputs for human decision-making.We contribute a mapping of analyzed deepfake detection pipelines, their intermediate outputs and a conceptual characterization of their evidential role. Our findings highlight that many detectors already generate rich internal signals during inference, yet these remain largely undocumented, non-standardized, and not readily accessible for human-centered verification workflows. By identifying and categorizing these intermediate outputs, this work aims to inform the design of more transparent and collaborative deepfake detection systems.