Voice AI technologies such as voice authentication and voice cloning are increasingly integrated and easily accessible, raising concerns around security, privacy, and fairness. In this demo paper, we describe our methodology of reflective design in a real-time Voice AI experience with the focus on reflection via estrangement, personalization, friction and reflection-in-action. We argue that designing such experiences can support critical reflection on Voice AI systems and lead to increased awareness of security risks, data collection practices, and potential identity loss.

4238

5488

As voice-enabled technologies are becoming increasingly more prevalent, voice-enabled permission requests become a crucial topic of investigation. It is yet unclear how to appropriately inform users in voice user interfaces (VUIs) about data processing practices. To understand how modality (text vs. voice) and the speech rate of the voice can influence users’ perceptions and decisions to grant permission, we conducted two preregistered studies (N = 343 and N = 594) and one pre-study, including two listening tasks to design potentially deceptive voice patterns. We found that users can distinguish between different levels of intrusiveness in the voice modality. However, they are less likely to accept voice-based permissions, pointing to cognitive problems associated with them. Moreover, we found that speech rate manipulations of action verbs “Accept” and “Decline” shifted users’ decisions towards acceptance, making the effect less controllable than predicted. This work highlights implications and design considerations for future voice-enabled permission requests.

In an age of voice-enabled technology, voice anonymization offers a solution to protect people’s privacy, provided these systems work equally well across subgroups. This study investigates bias in voice anonymization systems within the context of the Voice Privacy Challenge. We curate a novel benchmark dataset to assess performance disparities among speaker subgroups based on sex and dialect. We analyze the impact of three anonymization systems and attack models on speaker subgroup bias and reveal significant performance variations. Notably, subgroup bias intensifies with advanced attacker capabilities, emphasizing the challenge of achieving equal performance across all subgroups. Our study highlights the need for inclusive benchmark datasets and comprehensive evaluation strategies that address subgroup bias in voice anonymization.

Speaker recognition is a widely used voice-based biometric technology with applications in various industries, including banking, education, recruitment, immigration, law enforcement, healthcare, and well-being. However, while dataset evaluations and audits have improved data practices in face recognition and other computer vision tasks, the data practices in speaker recognition have gone largely unquestioned. Our research aims to address this gap by exploring how dataset usage has evolved over time and what implications this has on bias, fairness and privacy in speaker recognition systems. Previous studies have demonstrated the presence of historical, representation, and measurement biases in popular speaker recognition benchmarks. In this paper, we present a longitudinal study of speaker recognition datasets used for training and evaluation from 2012 to 2021. We survey close to 700 papers to investigate community adoption of datasets and changes in usage over a crucial time period where speaker recognition approaches transitioned to the widespread adoption of deep neural networks. Our study identifies the most commonly used datasets in the field, examines their usage patterns, and assesses their attributes that affect bias, fairness, and other ethical concerns. Our findings suggest areas for further research on the ethics and fairness of speaker recognition technology.

Defence is held on 26.11.2021 12:00 – 15:00 Zoom: https://aalto.zoom.us/j/61255513284