Human-Centered Privacy Research in the Age of Large Language Models (2402.01994v1)
Abstract: The emergence of LLMs, and their increased use in user-facing systems, has led to substantial privacy concerns. To date, research on these privacy concerns has been model-centered: exploring how LLMs lead to privacy risks like memorization, or can be used to infer personal characteristics about people from their content. We argue that there is a need for more research focusing on the human aspect of these privacy issues: e.g., research on how design paradigms for LLMs affect users' disclosure behaviors, users' mental models and preferences for privacy controls, and the design of tools, systems, and artifacts that empower end-users to reclaim ownership over their personal data. To build usable, efficient, and privacy-friendly systems powered by these models with imperfect privacy properties, our goal is to initiate discussions to outline an agenda for conducting human-centered research on privacy issues in LLM-powered systems. This Special Interest Group (SIG) aims to bring together researchers with backgrounds in usable security and privacy, human-AI collaboration, NLP, or any other related domains to share their perspectives and experiences on this problem, to help our community establish a collective understanding of the challenges, research opportunities, research methods, and strategies to collaborate with researchers outside of HCI.
- Is GitHub’s Copilot as bad as humans at introducing vulnerabilities in code? Empirical Software Engineering 28, 6 (September 2023). https://doi.org/10.1007/s10664-023-10380-1
- Quantifying memorization across neural language models. arXiv preprint arXiv:2202.07646 (2022).
- Extracting Training Data from Large Language Models.. In USENIX Security Symposium, Vol. 6.
- Can Language Models be Instructed to Protect Personal Information? arXiv preprint arXiv:2310.02224 (2023).
- Sheryl Estrada. 2023. A startup CFO used ChatGPT to build an FP&A tool—here’s how it went. https://fortune.com/2023/03/01/startup-cfo-chatgpt-finance-tool/ Accessed: 09/11/2023.
- Pedro Ferreira. 2023. Can ChatGPT Improve Technical Analysis and Trading Techniques? https://www.financemagnates.com/trending/can-chatgpt-improve-technical-analysis-and-trading-techniques/ Accessed: 09/11/2023.
- Thomas Germain. 2023. A Mental Health App Tested ChatGPT on Its Users. The Founder Said Backlash Was Just a Misunderstanding. https://gizmodo.com/mental-health-therapy-app-ai-koko-chatgpt-rob-morris-1849965534/ Accessed: 09/11/2023.
- Privacy Concerns in Chatbot Interactions. Springer International Publishing, 34–48. https://doi.org/10.1007/978-3-030-39540-7_3
- Youjeong Kim and S. Shyam Sundar. 2012. Anthropomorphism of computers: Is it mindful or mindless? Computers in Human Behavior 28, 1 (January 2012), 241–250. https://doi.org/10.1016/j.chb.2011.09.006
- Daniel Kimmel. 2023. ChatGPT Therapy Is Good, But It Misses What Makes Us Human. https://www.columbiapsychiatry.org/news/chatgpt-therapy-is-good-but-it-misses-what-makes-us-human. Accessed: 09/11/2023.
- Andrew Leonard. 2023. ‘Dr. Google’ meets its match: Dr. ChatGPT. https://www.latimes.com/science/story/2023-09-08/dr-google-meets-its-match-dr-chatgpt Accessed: 09/11/2023.
- Coconut: An IDE Plugin for Developing Privacy-Friendly Apps: An IDE Plugin for Developing Privacy-Friendly Apps. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 2, 4 (December 2018), 1–35. https://doi.org/10.1145/3287056
- Can LLMs Keep a Secret? Testing Privacy Implications of Language Models via Contextual Integrity Theory. arXiv preprint arXiv:2310.17884 (2023).
- Scalable extraction of training data from (production) language models. arXiv preprint arXiv:2311.17035 (2023).
- Asleep at the Keyboard? Assessing the Security of GitHub Copilot’s Code Contributions. In 2022 IEEE Symposium on Security and Privacy (SP). IEEE. https://doi.org/10.1109/sp46214.2022.9833571
- Privacy in the Time of Language Models. In Proceedings of the Sixteenth ACM International Conference on Web Search and Data Mining (WSDM ’23). ACM. https://doi.org/10.1145/3539597.3575792
- Beyond memorization: Violating privacy via inference with large language models. arXiv preprint arXiv:2310.07298 (2023).
- Mikhail Taver. 2023. ChatGPT is Coming to Finance, So Let’s Talk About the Risks and Rewards. https://www.unite.ai/chatgpt-is-coming-to-finance-so-lets-talk-about-the-risks-and-rewards/. Accessed: 09/11/2023.
- ReactGenie: An Object-Oriented State Abstraction for Complex Multimodal Interactions Using Large Language Models. arXiv preprint arXiv:2306.09649 (2023). https://doi.org/10.48550/arXiv.2306.09649
- Counterfactual memorization in neural language models. arXiv preprint arXiv:2112.12938 (2021).
- ” It’s a Fair Game”, or Is It? Examining How Users Navigate Disclosure Risks and Benefits When Using LLM-Based Conversational Agents. arXiv preprint arXiv:2309.11653 (2023). https://doi.org/10.48550/arXiv.2309.11653
- Tianshi Li (22 papers)
- Sauvik Das (13 papers)
- Hao-Ping Lee (3 papers)
- Dakuo Wang (87 papers)
- Bingsheng Yao (49 papers)
- Zhiping Zhang (9 papers)