pith. sign in

arxiv: 2606.03248 · v1 · pith:2KB7SPX7new · submitted 2026-06-02 · 💻 cs.HC

Investigating Novice Researchers' Perceptions of Research Privacy Within LLM-Assisted Workflows

Shuning Zhang , Changxi Wen , Eve He , Ying Ma , Robert Xiao , Xin Yi , Hewu Li This is my paper

Pith reviewed 2026-06-28 08:44 UTC · model grok-4.3

classification 💻 cs.HC
keywords LLM privacynovice researchersidea leakageresearch workflowspublication pressureprivacy mitigationsLLM-assisted research
0
0 comments X

The pith

Novice researchers' fear of idea leakage from LLMs paradoxically increases their reliance on these tools to speed up publication.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper examines how early-career researchers handle privacy risks when using public large language models in their scholarly work. Interviews show that worries about ideas being stolen do not slow people down; instead they drive faster use of LLMs to meet publication deadlines. Participants also believed their ideas were too ordinary to attract attacks and that any inputs would be lost in huge training sets. They tried partial mitigations such as breaking up prompts but viewed them as ineffective. The work points to needs for stronger institutional safeguards and clearer privacy training.

Core claim

Semi-structured interviews with 44 novice researchers across disciplines found that fear of idea leakage accelerates rather than deters reliance on public LLMs, because researchers turn to them to expedite publication; participants held misconceptions that their ideas lacked unique value for targeted attacks and that inputs would be safely diluted in massive datasets; five types of mitigations were identified but largely seen as ineffective.

What carries the argument

Semi-structured interviews that surface perceptions of privacy-publication trade-offs and list mitigations such as input fragmentation and adversarial probing.

If this is right

  • Privacy concerns lead researchers to increase LLM use to meet publication pressure.
  • Misconceptions about idea value and data dilution reduce perceived personal risk.
  • Common mitigations such as input fragmentation are viewed as ineffective by users.
  • Institution-level sandboxed isolation would address the identified risks.
  • Scenario-based privacy pedagogy could improve decision-making.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the pattern holds, reliance on public LLMs may grow even as privacy awareness rises.
  • The findings could extend to questions about how publication incentives interact with emerging AI tools in other high-pressure fields.
  • A follow-up study could test whether providing verifiable deletion audits changes reported behavior.

Load-bearing premise

That the self-reported perceptions from a convenience sample of 44 novice researchers accurately reflect the beliefs and decision processes of the broader population.

What would settle it

A larger random-sample survey in which most novice researchers state that privacy fears reduce their use of public LLMs would falsify the central claim.

read the original abstract

Large Language Model (LLMs)-assisted scholarly workflows introduce critical privacy and intellectual property risks. As a uniquely vulnerable cohort driven by publication pressure and a lack of institutional support, novice researchers rely heavily on public LLMs, compelling them to navigate high-stakes privacy-publication trade-offs. To investigate these concerns, we conducted semi-structured interviews with 44 researchers across diverse disciplines. Our findings reveal that the fear of idea leakage paradoxically accelerates, rather than deters, reliance on LLMs, as researchers utilize them to expedite publication. They also held misconceptions that their ideas lacked the unique value to attract targeted attacks, and that their inputs would be safely diluted within massive datasets, preventing reconstruction. From interviews, we identified five types of mitigations including input fragmentation and adversarial probing, though we found that participants largely perceived these measures as ineffective. We outline implications including implementing institution-level sandboxed isolation, scenario-based privacy pedagogy, and verifiable data-deletion audits for transparency.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript reports a qualitative study using semi-structured interviews with 44 novice researchers across disciplines to examine perceptions of privacy and IP risks when using public LLMs in scholarly work. It claims that fear of idea leakage paradoxically increases rather than decreases LLM reliance (to accelerate publication), that participants hold misconceptions about their ideas lacking unique value and about safe dilution in large datasets, identifies five mitigation types (e.g., input fragmentation, adversarial probing) that participants viewed as ineffective, and derives implications for institution-level sandboxed LLMs, scenario-based privacy training, and verifiable deletion audits.

Significance. If the thematic findings hold for the sampled cohort, the work supplies concrete, user-grounded evidence of counterintuitive privacy-publication dynamics and specific misconceptions among early-career researchers, which could directly inform HCI research on AI tool design, institutional policy, and ethics training. The interview-based approach usefully surfaces mitigation strategies and perceived ineffectiveness that quantitative surveys might miss.

major comments (3)
  1. [Abstract and Methods] Abstract and Methods: The central claims (paradoxical acceleration of LLM use; dilution and low-value misconceptions) are derived from thematic analysis of a convenience sample of 44 participants, yet the abstract and methods description supply no information on recruitment channels, inclusion/exclusion criteria, response rates, or steps taken to mitigate selection bias. This is load-bearing because the paper generalizes perceptions to 'novice researchers' as a cohort under publication pressure.
  2. [Methods] Methods: No details are provided on the coding process, codebook development, or inter-rater reliability for the thematic analysis that produces the five mitigation types and the paradoxical finding. Without these, the mapping from raw interview data to the reported misconceptions cannot be evaluated.
  3. [Findings] Findings: The claim that fear of leakage 'paradoxically accelerates' reliance rests solely on self-reported perceptions without triangulation (e.g., behavioral logs, follow-up probes, or cross-validation with usage data), leaving open the possibility that reported acceleration reflects post-hoc rationalization rather than a stable decision-making pattern.
minor comments (2)
  1. [Abstract] Abstract: The five mitigation types are referenced but not enumerated, reducing the abstract's utility for readers deciding whether to read the full findings section.
  2. [Discussion] Discussion: Consider adding an explicit limitations subsection that directly addresses the convenience sample, self-report biases, and scope of generalizability rather than leaving these implicit.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback. We address each major comment below and outline the revisions we will make to improve transparency and rigor.

read point-by-point responses

  1. Referee: [Abstract and Methods] Abstract and Methods: The central claims (paradoxical acceleration of LLM use; dilution and low-value misconceptions) are derived from thematic analysis of a convenience sample of 44 participants, yet the abstract and methods description supply no information on recruitment channels, inclusion/exclusion criteria, response rates, or steps taken to mitigate selection bias. This is load-bearing because the paper generalizes perceptions to 'novice researchers' as a cohort under publication pressure.

    Authors: We agree that recruitment details are essential for evaluating selection bias and generalizability. The submitted manuscript's Methods section describes the 44-participant sample but omits explicit channels and criteria. In revision we will add: recruitment via university mailing lists, discipline-specific forums, and social media; inclusion criteria (researchers with <5 years experience actively publishing); exclusion criteria (prior professional LLM training or non-novice status); and bias-mitigation steps (purposive sampling across fields and disciplines). We will also note the convenience-sampling limitation. The abstract will be updated if space allows. These additions will be incorporated. revision: yes

  2. Referee: [Methods] Methods: No details are provided on the coding process, codebook development, or inter-rater reliability for the thematic analysis that produces the five mitigation types and the paradoxical finding. Without these, the mapping from raw interview data to the reported misconceptions cannot be evaluated.

    Authors: We will revise the Methods section to detail the thematic analysis: iterative codebook development beginning with open coding of a subset of transcripts, collaborative refinement of codes into themes by the research team, and consensus-building through regular discussion meetings. We will report any formal inter-rater reliability statistics if they were calculated or, alternatively, describe the multi-coder process used to ensure consistency. This expanded description will make the derivation of the five mitigation types and the paradoxical finding traceable to the raw data. revision: yes

  3. Referee: [Findings] Findings: The claim that fear of leakage 'paradoxically accelerates' reliance rests solely on self-reported perceptions without triangulation (e.g., behavioral logs, follow-up probes, or cross-validation with usage data), leaving open the possibility that reported acceleration reflects post-hoc rationalization rather than a stable decision-making pattern.

    Authors: As a qualitative interview study centered on perceptions, self-reported accounts constitute the primary data source; the paradoxical acceleration emerged as a consistent cross-participant theme supported by verbatim quotes. Follow-up probes were used during interviews to elicit reasoning about privacy-publication trade-offs. Behavioral triangulation was outside the scope of this perception-focused design. We will add an explicit limitations paragraph in the Discussion acknowledging the possibility of post-hoc rationalization and the absence of usage logs, while noting that self-report methods are standard and appropriate for surfacing nuanced decision-making patterns in HCI privacy research. revision: partial

Circularity Check

0 steps flagged

No significant circularity; qualitative findings are direct summaries of interview data

full rationale

The paper's central claims derive from thematic analysis of 44 semi-structured interviews conducted by the authors. There are no equations, fitted parameters, statistical predictions, or self-citation chains that reduce any result to its own inputs by construction. The reported perceptions, misconceptions, and mitigation strategies are presented as outputs of the data collection process itself, with no reduction to prior fitted quantities or external uniqueness theorems. This is a standard self-contained qualitative report.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard qualitative assumptions about interview validity and participant self-reporting; no free parameters or invented entities are introduced.

axioms (1)
  • domain assumption Self-reported perceptions in semi-structured interviews reflect participants' actual beliefs and decision processes.
    Core validity assumption for any interview-based perception study.

pith-pipeline@v0.9.1-grok · 5708 in / 1061 out tokens · 22459 ms · 2026-06-28T08:44:13.839715+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

152 extracted references · 17 canonical work pages · 2 internal anchors

  1. [1]

    Yasemin Acar, Sascha Fahl, and Michelle L Mazurek. 2016. You are not your developer, either: A research agenda for usable security and privacy research beyond end users.2016 IEEE Cybersecurity Development (SecDev)(2016), 3–8

    2016

  2. [2]

    Atif Ahmad, Rachelle Bosua, and Rens Scheepers. 2014. Protecting organiza- tional competitive advantage: A knowledge leakage perspective.Computers & Security42 (2014), 27–39

    2014

  3. [3]

    Mutahar Ali, Arjun Arunasalam, and Habiba Farrukh. 2025. Understanding users’ security and privacy concerns and attitudes towards conversational ai platforms. In2025 IEEE Symposium on Security and Privacy (SP). IEEE, 298–316

    2025

  4. [4]

    American Geophysical Union. 2026. Honors Career Stages and Other Defi- nitions. https://www.agu.org/Honor-and-Recognize/Honors/Nomination- resources/Career-Stages [Accessed: 2026-04-29]

    2026

  5. [5]

    Apple Security Engineering and Architecture (SEAR). 2024. Private Cloud Compute: A new frontier for AI privacy in the cloud. Apple Security Research Blog. https://security.apple.com/blog/private-cloud-compute/ [Accessed: 2026- 04-29]

    2024

  6. [6]

    Jinheon Baek, Sujay Kumar Jauhar, Silviu Cucerzan, and Sung Ju Hwang. 2025. Researchagent: Iterative research idea generation over scientific literature with large language models. InProceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers...

    2025

  7. [7]

    Michael Bailey, David Dittrich, Erin Kenneally, and Doug Maughan. 2012. The menlo report.IEEE Security & Privacy10, 2 (2012), 71–75

    2012

  8. [8]

    David G Balash, Dongkun Kim, Darika Shaibekova, Rahel A Fainchtein, Micah Sherr, and Adam J Aviv. 2021. Examining the examiners: Students’ privacy and security perceptions of online proctoring services. InSeventeenth symposium on usable privacy and security (SOUPS 2021). 633–652

    2021

  9. [9]

    Leonard Bauersfeld, Angel Romero, Manasi Muglikar, and Davide Scaramuzza

  10. [10]

    Plos one18, 6 (2023), e0287611

    Cracking double-blind review: authorship attribution with deep learning. Plos one18, 6 (2023), e0287611

    2023

  11. [11]

    Tom L Beauchamp et al . 2008. The belmont report.The Oxford textbook of clinical research ethics(2008), 149–155

    2008

  12. [12]

    Sonja Bjelobaba, Lorna Waddington, Mike Perkins, Tomáš Folt`ynek, Sabuj Bhat- tacharyya, and Debora Weber-Wulff. 2025. Maintaining research integrity in the age of GenAI: an analysis of ethical challenges and recommendations to researchers.International Journal for Educational Integrity21, 1 (2025), 18

    2025

  13. [13]

    Ramazan Bulut. 2026. Mapping teachers’ awareness of artificial intelligence in the changing education paradigm: insights from a mixed methods inquiry. Frontiers in Psychology17 (2026), 1687155

    2026

  14. [14]

    Nicholas Carlini, Florian Tramer, Eric Wallace, Matthew Jagielski, Ariel Herbert- Voss, Katherine Lee, Adam Roberts, Tom Brown, Dawn Song, Ulfar Erlingsson, et al. 2021. Extracting training data from large language models. In30th USENIX security symposium (USENIX Security 21). 2633–2650

    2021

  15. [15]

    Jake Chanenson, Brandon Sloane, Navaneeth Rajan, Amy Morril, Jason Chee, Danny Yuxing Huang, and Marshini Chetty. 2023. Uncovering privacy and security challenges in K-12 schools. InProceedings of the 2023 CHI conference on human factors in computing systems. 1–28

    2023

  16. [16]

    Kang Chen, Xiuze Zhou, Yuanguo Lin, Shibo Feng, Li Shen, and Pengcheng Wu. 2025. A survey on privacy risks and protection in large language models. Journal of King Saud University Computer and Information Sciences37, 7 (2025), 163

    2025

  17. [17]

    Zhaoyang Chu, Yao Wan, Zhikun Zhang, Di Wang, Zhou Yang, Hongyu Zhang, Pan Zhou, Xuanhua Shi, Hai Jin, and David Lo. 2025. Scrub It Out! Erasing Sensitive Memorization in Code Language Models via Machine Unlearning. arXiv preprint arXiv:2509.13755(2025)

    work page arXiv 2025

  18. [18]

    Victoria Clarke and Virginia Braun. 2017. Thematic analysis.The journal of positive psychology12, 3 (2017), 297–298

    2017

  19. [19]

    Shaanan Cohney, Ross Teixeira, Anne Kohlbrenner, Arvind Narayanan, Mihir Kshirsagar, Yan Shvartzshnaider, and Madelyn Sanfilippo. 2021. Virtual class- rooms and real harms: Remote learning at {US}. universities. InSeventeenth Symposium on Usable Privacy and Security (SOUPS 2021). 653–674

    2021

  20. [20]

    A Feder Cooper, Aaron Gokaslan, Ahmed M Ahmed, Amy B Cyphert, Christo- pher De Sa, Mark Lemley, Daniel E Ho, and Percy Liang. 2025. Extracting memorized pieces of (copyrighted) books from open-weight language models. InICML 2025 Workshop on Reliable and Responsible Foundation Models

    2025

  21. [21]

    John W Creswell and Timothy C Guetterman. 2019. Educational Research: Planning, Conducting, and Evaluating Quantitative and Qualitative Research. Pearson(2019)

    2019

  22. [22]

    Department of History, National University of Singapore. 2025. Guiding Principles on Generative AI Use AY2025–2026. https://fass.nus.edu.sg/hist/wp- content/uploads/sites/7/2025/08/HY_Guide-to-Use-of-AI-Tools_Aug2025.pdf. Accessed: 2026-04-29

    2025

  23. [23]

    Yi Dong, Ronghui Mu, Yanghao Zhang, Siqi Sun, Tianle Zhang, Changshun Wu, Gaojie Jin, Yi Qi, Jinwei Hu, Jie Meng, et al. 2025. Safeguarding large language models: A survey.Artificial intelligence review58, 12 (2025), 382

    2025

  24. [24]

    André Vicente Duarte, Xuandong Zhao, Arlindo L Oliveira, and Lei Li. 2024. DE-COP: Detecting Copyrighted Content in Language Models Training Data. InInternational Conference on Machine Learning. PMLR, 11940–11956

    2024

  25. [25]

    Timothy J Ellis and Yair Levy. 2009. Towards a Guide for Novice Researchers on Research Methodology: Review and Proposed Methods.Issues in Informing Science & Information Technology6 (2009)

    2009

  26. [26]

    Thomais Gkrimpizi, Vassilios Peristeras, and Ioannis Magnisalis. 2023. Classi- fication of barriers to digital transformation in higher education institutions: Systematic literature review.Education sciences13, 7 (2023), 746

    2023

  27. [27]

    Yichen Gong, Delong Ran, Xinlei He, Tianshuo Cong, Anyu Wang, and Xiaoyun Wang. 2025. Safety Misalignment Against Large Language Models.. InNDSS

    2025

  28. [28]

    Igor Grossmann, Matthew Feinberg, Dawn C Parker, Nicholas A Christakis, Philip E Tetlock, and William A Cunningham. 2023. AI and the transformation of social science research.Science380, 6650 (2023), 1108–1109

    2023

  29. [29]

    Tom Gunter, Zirui Wang, Chong Wang, Ruoming Pang, Andy Narayanan, Aonan Zhang, Bowen Zhang, Chen Chen, Chung-Cheng Chiu, David Qiu, et al. 2024. Apple intelligence foundation language models.arXiv preprint arXiv:2407.21075 (2024)

    work page arXiv 2024

  30. [30]

    Qianyue Hao, Fengli Xu, Yong Li, and James Evans. 2026. Artificial intelligence tools expand scientists’ impact but contract science’s focus.Nature(2026), 1–7

    2026

  31. [31]

    Harvard Catalyst Regulatory Foundations, Ethics, and Law Program

  32. [32]

    https://catalyst.harvard.edu/regulatory/ai-human-participant- research/educational-resources/

    Artificial Intelligence and Human Participant Research: Educational Resources. https://catalyst.harvard.edu/regulatory/ai-human-participant- research/educational-resources/. [Accessed: 2026-04-29]

    2026

  33. [33]

    Harvard Catalyst Regulatory Foundations, Ethics, and Law Program. n.d.. Emerg- ing Technologies, Ethics, and Research Data. https://catalyst.harvard.edu/ regulatory/emerging-technologies/. [Accessed: 2026-04-29]

    2026

  34. [34]

    Harvard Gazette. 2023. Harvard designs AI sandbox that enables exploration, interaction without compromising security. https: //news.harvard.edu/gazette/story/newsplus/harvard-designs-ai-sandbox- that-enables-exploration-interaction-without-compromising-security/. Accessed: 2026-04-29

    2023

  35. [35]

    Harvard Medical School Information Technology. n.d.. Generative AI. https: //it.hms.harvard.edu/about/policies-and-guidelines/generative-ai. [Accessed: 2026-04-29]

    2026

  36. [36]

    Don’t Forget the Teachers

    Emma Harvey, Allison Koenecke, and Rene F Kizilcec. 2025. " Don’t Forget the Teachers": Towards an Educator-Centered Understanding of Harms from Large Language Models in Education. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems. 1–19

    2025

  37. [37]

    Aslinda Hassan, Mas Nida Md Khambari, Najwan Khambari, Nurul Azma Za- karia, Taqwan Thamrin, and Wan Mohd Yaakob Wan Bejuri. 2025. Awareness and Misconceptions of AI Among Educators.International Journal of Research and Innovation in Social Science (IJRISS)9, 11 (2025)

    2025

  38. [38]

    Amanda Heidt. 2024. Intellectual property and data privacy: the hidden risks of AI.Nature(2024)

    2024

  39. [39]

    Bo Hui, Haolin Yuan, Neil Gong, Philippe Burlina, and Yinzhi Cao. 2024. Pleak: Prompt leaking attacks against large language model applications. InProceed- ings of the 2024 on ACM SIGSAC Conference on Computer and Communications Conference acronym ’XX, June 03–05, 2018, Woodstock, NY Trovato and Tobin, et al. Security. 3600–3614

    2024

  40. [40]

    Sangeeta Jadhav, Jayshree Lavhare, Ritik Kumar, Soumyadip Roy, Ankit Kumar Singh, and Kavya Chauhan. 2025. Edge-AI Multimodal RAG Chat bot with Personalized Knowledge Graph for Offline Privacy-Preserving Information Retrieval. In2025 5th Asian Conference on Innovation in Technology (ASIANCON). IEEE, 1–4

    2025

  41. [41]

    Hyejun Jeong, Mohammadreza Teymoorianfard, Abhinav Kumar, Amir Houmansadr, and Eugene Bagdasarian. 2025. Network-Level Prompt and Trait Leakage in Local Research Agents.arXiv preprint arXiv:2508.20282(2025)

    work page arXiv 2025

  42. [42]

    Antonia Karamolegkou, Jiaang Li, Li Zhou, and Anders Søgaard. 2023. Copyright violations and large language models. InProceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. 7403–7412

    2023

  43. [43]

    Easton Kelso, Ananta Soneji, Syed Zami-Ul-Haque Navid, Yan Shoshitaishvili, Sazzadur Rahaman, and Rakibul Hasan. 2025. Investigating the Security & Privacy Risks from Unsanctioned Technology Use by Educators. InProceedings of the Extended Abstracts of the CHI Conference on Human Factors in Computing Systems. 1–6

    2025

  44. [44]

    Easton Kelso, Ananta Soneji, Sazzadur Rahaman, Yan Shoshitaishvili, and Rak- ibul Hasan. 2024. Trust, Because You Can’t Verify: Privacy and Security Hurdles in Education Technology Acquisition Practices. InProceedings of the 2024 on ACM SIGSAC Conference on Computer and Communications Security. 1656–1670

    2024

  45. [45]

    Sushmita Khan, Mehtab Iqbal, Oluwafemi Osho, Khushbu Singh, Kyra Derrick, Philip Nelson, Lingyuan Li, Emily Sidnam-Mauch, Nicole Bannister, Kelly Caine, et al. 2024. Teaching middle schoolers about the privacy threats of tracking and pervasive personalization: A classroom intervention using design-based re- search. InProceedings of the 2024 CHI Conference...

    2024

  46. [46]

    Diana Kramer, Lambert Rosique, Ajay Narotam, Elie Bursztein, Patrick Gage Kelley, Kurt Thomas, and Allison Woodruff. 2025. Integrating large language models into security incident response. InTwenty-First Symposium on Usable Privacy and Security (SOUPS 2025). 133–148

    2025

  47. [47]

    Monika Blue Kwapisz, Avanya Kohli, and Prashanth Rajivan. 2024. Privacy concerns of student data shared with instructors in an online learning manage- ment system. InProceedings of the 2024 CHI Conference on Human Factors in Computing Systems. 1–16

    2024

  48. [48]

    Hao-Ping Lee, Yu-Ju Yang, Matthew Bilik, Isadora Krsek, Thomas Serban von Davier, Kyzyl Monteiro, Jason Lin, Shivani Agarwal, Jodi Forlizzi, and Sauvik Das. 2026. Privy: Envisioning and Mitigating Privacy Risks for Consumer-facing AI Product Concepts. InProceedings of the 2026 CHI Conference on Human Factors in Computing Systems. 1–30

    2026

  49. [49]

    Hao-Ping Hank Lee, Lan Gao, Stephanie Yang, Jodi Forlizzi, and Sauvik Das

  50. [50]

    I Don’t Know If We’re Doing Good. I Don’t Know If We’re Doing Bad

    " I Don’t Know If We’re Doing Good. I Don’t Know If We’re Doing Bad": Investigating How Practitioners Scope, Motivate, and Conduct Privacy Work When Developing {AI} Products. In33rd USENIX Security Symposium (USENIX Security 24). 4873–4890

  51. [51]

    Yoonjoo Lee, Hyeonsu B Kang, Matt Latzke, Juho Kim, Jonathan Bragg, Joseph Chee Chang, and Pao Siangliulue. 2024. Paperweaver: Enriching topical paper alerts by contextualizing recommended papers with user-collected pa- pers. InProceedings of the 2024 CHI Conference on Human Factors in Computing Systems. 1–19

    2024

  52. [52]

    Junyi Li, Jie Chen, Ruiyang Ren, Xiaoxue Cheng, Wayne Xin Zhao, Jian-Yun Nie, and Ji-Rong Wen. 2024. The dawn after the dark: An empirical study on factuality hallucination in large language models. InProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 10879–10899

    2024

  53. [53]

    Qinbin Li, Junyuan Hong, Chulin Xie, Jeffrey Tan, Rachel Xin, Junyi Hou, Xavier Yin, Zhun Wang, Dan Hendrycks, Zhangyang Wang, et al . 2024. LLM-PBE: Assessing Data Privacy in Large Language Models.Proceedings of the VLDB Endowment17, 11 (2024), 3201–3214

    2024

  54. [54]

    Tianshi Li, Elizabeth Louie, Laura Dabbish, and Jason I Hong. 2021. How developers talk about personal data and what it means for user privacy: A case study of a developer forum on reddit.Proceedings of the ACM on Human- Computer Interaction4, CSCW3 (2021), 1–28

    2021

  55. [55]

    Zhuoyang Li, Yanlai Wu, Yao Li, Xinning Gui, and Yuhan Luo. 2026. Privacy Control in Conversational LLM Platforms: A Walkthrough Study. InProceedings of the 2026 CHI Conference on Human Factors in Computing Systems. 1–26

    2026

  56. [56]

    Daogao Liu, Edith Cohen, Badih Ghazi, Peter Kairouz, Pritish Kamath, Alexander Knop, Ravi Kumar, Pasin Manurangsi, Adam Sealfon, Da Yu, et al. 2025. URANIA: Differentially Private Insights into AI Use. InSecond Conference on Language Modeling

    2025

  57. [57]

    Lanjing Liu, Xinran Adeline Li, Allen Yilun Lin, and Yaxing Yao. 2026. Designing Privacy Choice in Generative AI Chatbot Ecosystems. InProceedings of the 2026 CHI Conference on Human Factors in Computing Systems. 1–20

    2026

  58. [58]

    Xiao-Yang Liu, Rongyi Zhu, Daochen Zha, Jiechao Gao, Shan Zhong, Matt White, and Meikang Qiu. 2025. Differentially private low-rank adaptation of large language model using federated learning.ACM Transactions on Management Information Systems16, 2 (2025), 1–24

    2025

  59. [59]

    Yupei Liu, Yuqi Jia, Runpeng Geng, Jinyuan Jia, and Neil Zhenqiang Gong. 2024. Formalizing and benchmarking prompt injection attacks and defenses. In33rd USENIX Security Symposium (USENIX Security 24). 1831–1847

    2024

  60. [60]

    Rongjun Ma, Caterina Maidhof, Juan Carlos Carrillo, Janne Lindqvist, and Jose Such. 2025. Privacy perceptions of custom gpts by users and creators. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems. 1–18

    2025

  61. [61]

    Sebastian Porsdam Mann, Mateo Aboy, Joel Jiehao Seah, Zhicheng Lin, Xufei Luo, Daniel Rodger, Hazem Zohny, Timo Minssen, Julian Savulescu, and Brian D Earp

  62. [62]

    AI and the Future of Academic Peer Review.arXiv preprint arXiv:2509.14189 (2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  63. [63]

    Abigail Marsh and Lauren R Milne. 2025. I Don’t Want to Sound Rude, but It’s None of Their Business: Exploring Security and Privacy Concerns around As- sistive Technology Use in Educational Settings.ACM Transactions on Accessible Computing17, 2 (2025), 1–30

    2025

  64. [64]

    Nora McDonald, Sarita Schoenebeck, and Andrea Forte. 2019. Reliability and inter-rater reliability in qualitative research: Norms and guidelines for CSCW and HCI practice.Proceedings of the ACM on human-computer interaction3, CSCW (2019), 1–23

    2019

  65. [65]

    Niloofar Mireshghallah, Maria Antoniak, Yash More, Yejin Choi, and Golnoosh Farnadi. 2024. Trust No Bot: Discovering Personal Disclosures in Human-LLM Conversations in the Wild. InFirst Conference on Language Modeling

    2024

  66. [66]

    Kyzyl Monteiro, Yuchen Wu, and Sauvik Das. 2025. Imago Obscura: An Im- age Privacy AI Co-pilot to Enable Identification and Mitigation of Risks. In Proceedings of the 38th Annual ACM Symposium on User Interface Software and Technology. 1–26

    2025

  67. [67]

    Miryam Naddaf. 2025. AI is transforming peer review—and many scientists are worried.Nature639, 8056 (2025), 852–854

    2025

  68. [68]

    Nima Naderloui, Shenao Yan, Binghui Wang, Jie Fu, Wendy Hui Wang, Weiran Liu, and Yuan Hong. 2025. Rectifying privacy and efficacy measurements in machine unlearning: A new inference attack perspective. In34th USENIX Security Symposium (USENIX Security 25). 5545–5564

    2025

  69. [69]

    Bei Yi Ng, Jiarui Li, Xinyuan Tong, Kevin Ye, Gauthami Yenne, Varun Chan- drasekaran, and Jingjie Li. 2025. Analyzing security and privacy chal- lenges in generative ai usage guidelines for higher education.arXiv preprint arXiv:2506.20463(2025)

    work page arXiv 2025

  70. [70]

    Andy Nguyen, Yvonne Hong, Belle Dang, and Xiaoshan Huang. 2024. Human- AI collaboration patterns in AI-assisted academic writing.Studies in Higher Education49, 5 (2024), 847–864

    2024

  71. [71]

    Sanghak Oh, Kiho Lee, Seonhye Park, Doowon Kim, and Hyoungshick Kim

  72. [72]

    In2024 IEEE Symposium on Security and Privacy (SP)

    Poisoned chatgpt finds work for idle hands: Exploring developers’ coding practices with insecure suggestions from poisoned ai models. In2024 IEEE Symposium on Security and Privacy (SP). IEEE, 1141–1159

  73. [73]

    Sebastian Porsdam Mann, Anuraag A Vazirani, Mateo Aboy, Brian D Earp, Timo Minssen, I Glenn Cohen, and Julian Savulescu. 2024. Guidelines for ethical use and acknowledgement of large language models in academic writing.Nature Machine Intelligence6, 11 (2024), 1272–1274

    2024

  74. [74]

    James Prather, Juho Leinonen, Natalie Kiesler, Jamie Gorson Benario, Sam Lau, Stephen MacNeil, Narges Norouzi, Simone Opel, Vee Pettit, Leo Porter, et al. 2025. Beyond the hype: A comprehensive review of current trends in generative ai research, teaching practices, and tools.2024 Working Group Reports on Innovation and Technology in Computer Science Educa...

    2025

  75. [75]

    Maxwell Prybylo, Sara Haghighi, Sai Teja Peddinti, and Sepideh Ghanavati

  76. [76]

    InTwentieth Symposium on Usable Privacy and Security (SOUPS 2024)

    Evaluating privacy perceptions, experience, and behavior of software development teams. InTwentieth Symposium on Usable Privacy and Security (SOUPS 2024). 101–120

    2024

  77. [77]

    2021.Cybersecurity and third-party risk: Third party threat hunting

    Gregory C Rasner. 2021.Cybersecurity and third-party risk: Third party threat hunting. John Wiley & Sons

    2021

  78. [78]

    Ankita Priti Roy, Kerena Anand, N Elangovan, and D Halaswamy. 2024. Priori- tizing Risks in AI-Enabled EdTech Platforms: An Analytic Hierarchy Process Approach. InInternational Conference on Artificial Intelligence on Textile and Apparel. Springer, 549–564

    2024

  79. [79]

    Rutgers Office for Research. n.d.. Human Research Protection Program Toolkit. https://research.rutgers.edu/faculty-staff/compliance/human-research- protection/toolkit. [Accessed: 2026-04-29]

    2026

  80. [80]

    Rutgers Office for Research. n.d.. Interactive IRB Tools. https://research. rutgers.edu/faculty-staff/compliance/human-research-protection/interactive- irb-tools. Accessed: 2026-04-29

    2026

Showing first 80 references.