"Before, I Asked My Mom, Now I Ask ChatGPT": Visual Privacy Management with Generative AI for Blind and Low-Vision People

Ayae Ide; Danna Gurari; Kelly Avery Mack; Leah Findlater; Lotus Zhang; Tanusree Sharma; Yang Wang; Yu-Yun Tseng

REVIEW 2 major objections 5 minor 115 references

"Before, I Asked My Mom, Now I Ask ChatGPT": Visual Privacy Management with Generative AI for Blind and Low-Vision People

T0 review · 2 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Blind and low-vision people are already using GenAI as a deliberate privacy-management strategy, choosing AI over human helpers for sensitive visual tasks.

desk verdict Solid empirical study of BLV people's use of GenAI for visual privacy, though the 'current practices' framing is a bit ahead of the protocol's ability to distinguish actual from imagined use. read the letter →

arxiv 2507.00286 v2 pith:UTH6AGYV submitted 2025-06-30 cs.HC cs.AIcs.ET

Tanusree Sharma , Yu-Yun Tseng , Lotus Zhang , Ayae Ide , Kelly Avery Mack , Leah Findlater , Danna Gurari , Yang Wang This is my paper

classification cs.HCcs.AIcs.ET

keywords visualprivacygenerativeAIblindandlowvisionaccessibilityinterviewstudybydesignemotionalagency

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper claims that blind and low-vision (BLV) people are already using generative AI tools to manage visual privacy, deliberately choosing ChatGPT, Be My AI, or Seeing AI over family, friends, or remote agents for tasks like reading pregnancy test results, checking a room for embarrassing clutter, or reviewing a photo before posting it. Based on interviews with 21 BLV participants, it argues that these tools let users balance privacy, efficiency, and emotional agency, and that the same users want future GenAI to process data on-device, guarantee zero retention, redact sensitive content, and offer tactile, multimodal feedback. The contribution is a map of current practices and design preferences that reframes visual privacy not just as preventing disclosure, but as supporting autonomy and accountability. If the findings hold, they imply that GenAI is already functioning as a desirable privacy-preserving alternative to human assistance for some BLV users, and that accessibility tools should treat privacy as a first-class design requirement.

What carries the argument

The scenario-driven interview protocol is the central instrument: six concrete privacy situations, grounded in prior work, that prompt participants to describe both what they do today and what they would want from future GenAI. Thematic analysis of the resulting 102 low-level codes into 29 themes converts these stories into claims about practices and preferences. The paper also relies on two analytical lenses: impression management (controlling how others perceive one's appearance and environment) and accountability (handling others' private content), which together define the expanded notion of visual privacy. The technical design recommendations—a local GenAI sandbox, a federated compliance-aware toolkit, a personalized appearance-feedback system, and visual disambiguation for shared spaces—are the material instantiation of this machinery.

What would settle it

A field deployment that logs actual GenAI privacy-related interactions of BLV users for a month, and compares the logged choices (e.g., which tool for reading a bank statement, whether they ask a friend instead) against the self-reported practices; if the logs show no systematic preference for GenAI over human help for sensitive content, the paper's central claim would be falsified.

Watch

Extended reading notes

Core claim

The central discovery is that visual privacy management for BLV users has shifted into everyday GenAI use: participants described using AI for tasks they once delegated to family, friends, volunteers, or remote agents, specifically to avoid social awkwardness, judgment, or data exposure. The paper organizes these practices around six scenarios—self-appearance checks, indoor space checks, social media photo review, document sharing with employers, handling others' documents as professionals, and outdoor navigation—and shows that in each, users weigh autonomy against accuracy and trust. A notable finding is that participants preferred GenAI over humans for deeply personal readings such as pregnancy tests, mammograms, and bank statements, valuing independence and privacy even when the AI gave inconsistent results. The paper also documents a second dimension of privacy: accountability for others' data, where blind professionals use GenAI to read client or student documents and want compliance-aware tooling. The authors propose design directions—on-device processing, zero-retention guarantees, redaction, privacy-aware appearance indicators, and tactile mirrored interaction—as direct consequences of these lived practices.

Load-bearing premise

The findings rest on self-reports from 21 US-based BLV participants who already use GenAI, and the interview scenarios were chosen by the researchers; if those scenarios omit important privacy contexts, or if participants' descriptions of their practices differ from what they actually do, the central descriptive claims about current usage may not generalize.

Editorial extensions

If this is right

If GenAI is already a desired privacy substitute for human help on sensitive visuals, accessibility apps should place privacy controls at the center of the interaction flow, not as an afterthought.
The finding that users prefer independence even when AI results are inaccurate implies that privacy-preserving local processing must not sacrifice the conversational, clarifying interaction style that GenAI offers.
The professional scenario shows that BLV employees need institutionally sanctioned, compliance-aware GenAI tooling rather than self-governed use when handling others' data.
The appearance and shared-space scenarios imply a demand for personalized, trainable object and identity recognition that distinguishes the user's own items from others'.
The emotional-confidant use points to a need for GenAI designs that guarantee no memory or retention as an explicit feature, not a hidden risk.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The authors restrict their scenario set to six contexts; a natural testable extension is a broader diary study capturing privacy episodes the interviews may have missed, such as medical appointments, dating, or legal document review.
The preference for on-device processing suggests a concrete product direction: local small language models could satisfy many of these needs today, and their privacy properties could be verified by network-traffic audits.
The paper's reframing of privacy as emotional agency implies that privacy research for BLV users should include affective outcomes like dignity, embarrassment, and independence as measured variables, not just disclosure risk.
If GenAI companies adopt zero-retention guarantees as a differentiator, the same design vocabulary could extend to other populations with heightened privacy sensitivity, such as survivors of abuse or people with stigmatized conditions.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Read this paper. It's a well-done interview study (n=21) on how blind and low-vision people use ChatGPT, Be My AI, Seeing AI, and similar tools to manage visual privacy: checking appearance, scanning rooms, reading medical or financial documents, and handling others' private content at work. The new contribution is real. Prior work covered visual privacy with human assistance and with traditional visual assistive technologies, but not GenAI specifically as a privacy-management tool. The themes of emotional agency and independence (e.g., reading a pregnancy test rather than asking family) are vivid and useful. The design recommendations (on-device processing, zero-retention, redaction, appearance indicators, tactile multimodal feedback) follow from the data rather than appearing out of nowhere. Method is standard for the field: semi-structured interviews, thematic analysis, 0.87 inter-coder reliability, full protocol in the appendix. Credit where it's due.

The main soft spot is the one the stress-test flagged. Section 2 of the protocol asks participants to imagine specific scenarios ('Imagine you are about to meet someone...') and then asks 'How would you currently use...?'. Section 4 then presents these responses as current practice. The paper never tells you which quotes came from the spontaneous Section 1 questions versus the scenario-prompted Section 2 questions. That matters because the scenario prompt supplies the privacy context and projects the GenAI-use frame. Some participants may be describing what they would do, not what they have done. The abstract's claim of 'a range of current practices ... across six key scenarios' is therefore a bit stronger than the evidence can support. This is not a fatal flaw—the quotes often refer to past behavior ('Before, I asked my mom...')—but it is a real conflation that should be addressed by reporting provenance and softening the claim if needed.

Other soft spots are minor for qualitative work: the six scenarios are researcher-selected, so the theme structure is partly imposed; no saturation analysis or member checking; no codebook or data artifacts. The sample is all GenAI users, US only, and skews toward adopters, which the limitations section partially acknowledges.

All told: this is a solid, citable empirical contribution. I'd send it to peer review without hesitation. Revisions should ask for a clear breakdown of which findings are current-use versus hypothetical, and moderate any 'first' claims relative to closely related prior work (e.g., Adnin & Das 2024). I'd also ask for the codebook as a supplement. Worth bringing to reading group—there's a good conversation to be had about elicitation validity in interview studies.

Referee Report

2 major / 5 minor

Summary. This paper reports a semi-structured interview study with 21 blind and low-vision (BLV) participants in the United States, examining how they currently use generative AI (GenAI) tools to manage visual privacy and what design improvements they envision. The study is organized around six researcher-defined privacy scenarios (self-presentation, indoor spatial privacy, social media sharing, document sharing, professional handling of others' content, and outdoor spatial privacy). Findings are presented as themes under two research questions: current practices (Section 4) and perceived limitations/expectations (Section 5). The paper concludes with design recommendations including on-device processing, zero-retention guarantees, redaction, privacy-aware appearance indicators, and multimodal tactile interaction. The manuscript includes the full interview protocol in an appendix and reports inter-coder reliability of 0.87.

Significance. If the descriptive claims are supported, this is a useful and timely contribution to accessibility and privacy research. The study is one of the first to focus specifically on GenAI-mediated visual privacy for BLV users, a population often underrepresented in privacy research. The paper's strengths include a transparently reported interview protocol, a participant sample with diverse blindness levels and GenAI tool usage, rich verbatim quotes, and concrete design directions grounded in participant accounts. The design recommendations, while not implemented or evaluated, are actionable and derive from stated user expectations. The main significance hinges on whether the reported 'current practices' are truly current behaviors or partly hypothetical responses to researcher-supplied scenarios; this distinction directly affects the paper's central claim that GenAI is already functioning as a privacy-preserving alternative to human assistance for some BLV users.

major comments (2)

[Section 3.1 and Appendix A.1 vs. Section 4] The manuscript conflates spontaneously reported current practices with responses elicited by hypothetical scenario prompts. The interview protocol in Appendix A.1 shows that Section 2 asks participants to 'Imagine you are about to meet someone...' and then asks both 'How would you currently use a Generative AI tool...' and 'How would you like this tool to be designed in the future...'. Section 4, however, presents quotes from these scenario-based discussions as evidence of current practice without indicating which quotes were volunteered in Section 1 (unprompted) versus produced only after the scenario prompt. For example, the P15 quote in Section 4.1 begins 'Say, I had to take a photo of myself...' and appears to be a scenario-elicited response, yet it is used to support the claim that participants 'replaced human assistance with GenAI.' The paper never reports the provenance of the quotes in Section 4 or Table 3. Because the central contribution is an account of current practices, the authors must either re-analyze their data to separate spontaneous reports from scenario-elicited hypotheticals and report this distinction, or substantially temper the abstract's claim that the findings reveal 'a range of current practices with GenAI.' Without this, the reader cannot tell whether the paper documents actual behavior or imagined use.
[Section 3.1 and Table 1] The six 'key scenarios' are researcher-defined, and the results are organized around them, so the finding that current practices span these six scenarios is partly an artifact of the instrument. The text states that the scenarios were 'inspired by prior work' and selected based on 'relevancy,' 'gap,' and 'diversity of context,' which is a reasonable design choice, but the paper does not discuss what themes or practices emerged outside these predefined scenarios. For instance, Section 1 of the protocol asks generally about GenAI use for visual information and privacy, yet Section 4 reports no themes that are explicitly identified as having arisen from that open-ended part of the interview. The manuscript also does not report a saturation analysis or member checking, and the limitations section (6.5) only mentions sample representativeness. The authors should acknowledge that the scenario set may not cover all relevant privacy contexts and should either present evidence of saturation or report themes that emerged independently of the scenario prompts.

minor comments (5)

[Appendix A.1, Privacy Scenario 4] The prompt says 'sharing a financial report with an employee,' but the scenario and Table 1 clearly refer to sharing with an employer; this wording should be corrected.
[Section 3.3] The analysis section reports 'approximately 29 themes' derived from 102 low-level codes, which is a large number for 21 interviews; the paper would benefit from explaining how these themes are hierarchically organized and how many are reported in the results.
[Section 4.4] The subsection on 'Sociopolitical Decision Making Factors' includes content about AI models repeating myths about blindness and about content-filter vagueness; these points are relevant but feel somewhat disconnected from the visual privacy focus, and a linking sentence would improve coherence.
[Section 6.4] The design implications are generally grounded in the findings, but the proposal for a 'fine-tuned multi-label visual classifier and customizable sensitivity profiles' (appearance feedback system) goes beyond the interview data into specific model choices; the authors should clarify which parts are directly supported by participant statements and which are expert design suggestions.
[Reference list] Reference [6] (Acquisti and Grossklags) is cited in the introduction to privacy decision-making but is not discussed in the related work; either integrate it or remove the citation to avoid a dangling reference.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: findings are grounded in participant interviews, and the authors' prior work is cited only as background.

full rationale

This is a qualitative interview study with no equations, fitted parameters, or formal derivation chain that could reduce to its own inputs. The central claims—current practices and design preferences of BLV individuals using GenAI for visual privacy—are supported by quoted participant responses and thematic analysis of interview transcripts. The paper's scenario prompts ask participants both how they currently use GenAI and how they would like it designed, and Section 4 reports participant-described practices; this is an elicitation design choice rather than a circular derivation. The authors cite their own prior work in the related work and scenario-selection sections, but those citations provide background and framing, not load-bearing evidence for the paper's empirical findings; no uniqueness theorem or fitted parameter is imported from prior work. Even the design recommendations in Section 6.4 are explicitly grounded in participants' stated limitations and expectations. A reader could question whether prompted hypothetical responses are over-represented in the 'current use' themes, but that is an external-validity or methodological concern, not a circularity of the kind where a predicted result is equivalent to its input by construction.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

This is a qualitative interview study, so the central claim rests on methodological assumptions rather than fitted parameters. No numerical free parameters and no invented entities. The key assumptions are about the validity of self-reports, the coverage of the predefined scenarios, and the sufficiency of the sample for saturation.

assumptions (4)

domain assumption Participants' self-reported usage accurately reflects their actual GenAI practices and privacy behaviors.
Data are interview transcripts only; no behavioral logs, diary studies, or observational data are used (Section 3.3).
ad hoc to paper The six researcher-defined scenarios adequately cover the range of visual privacy contexts relevant to BLV people.
Scenarios were chosen from prior literature and researcher judgment (Section 3.1), not derived from participants; findings are structured by these scenarios.
domain assumption The sample of 21 GenAI-using, US-based BLV participants is sufficient to reach stable themes.
The paper reports inter-coder reliability but does not provide a saturation analysis; the limitation section notes the sample may not represent the broader BLV community (Section 6.5).
domain assumption Thematic coding and the '20% coding plus Cohen's Kappa 0.87' procedure yield trustworthy themes.
The codebook is not included, so the coding quality cannot be independently audited (Section 3.3).

how reviews work

0 comments

Cite this review

Pith. "Pith review of "Before, I Asked My Mom, Now I Ask ChatGPT": Visual Privacy Management with Generative AI for Blind and Low-Vision People." pith.science (2026). https://pith.science/paper/UTH6AGYV

@misc{pith2026250700286,
  author       = {Pith},
  title        = {Pith review of: "Before, I Asked My Mom, Now I Ask ChatGPT": Visual Privacy Management with Generative AI for Blind and Low-Vision People},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UTH6AGYV}},
  note         = {Machine review of arXiv:2507.00286}
}

read the original abstract

Blind and low vision (BLV) individuals use Generative AI (GenAI) tools to interpret and manage visual content in their daily lives. While such tools can enhance the accessibility of visual content and so enable greater user independence, they also introduce complex challenges around visual privacy. In this paper, we investigate the current practices and future design preferences of blind and low vision individuals through an interview study with 21 participants. Our findings reveal a range of current practices with GenAI that balance privacy, efficiency, and emotional agency, with users accounting for privacy risks across six key scenarios, such as self-presentation, indoor/outdoor spatial privacy, social sharing, and handling professional content. Our findings reveal design preferences, including on-device processing, zero-retention guarantees, sensitive content redaction, privacy-aware appearance indicators, and multimodal tactile mirrored interaction methods. We conclude with actionable design recommendations to support user-centered visual privacy through GenAI, expanding the notion of privacy and responsible handling of others data.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

115 extracted references · 54 canonical work pages

[1]

[n. d.]. Employment (Title I) — archive.ada.gov. https://archive.ada.gov/ada_ title_I.htm. [Accessed 07-07-2025]

2025
[2]

[n. d.]. Envision - Perceive Possibility — letsenvision.com. https://www. letsenvision.com/. [Accessed 08-04-2025]

2025
[3]

[n. d.]. Home — bemyeyes.com. https://www.bemyeyes.com/. [Accessed 08-04-2025]

2025
[4]

[n. d.]. Seeing AI - Talking Camera for the Blind — seeingai.com. https://www. seeingai.com/. [Accessed 08-04-2025]

2025
[5]

Be My Eyes

2023. Be My Eyes. https://www.bemyeyes.com/. Accessed: 2023-9-02

2023
[6]

Alessandro Acquisti and Jens Grossklags. 2005. Privacy and rationality in individual decision making. IEEE security & privacy 3, 1 (2005), 26–33

2005
[7]

Disabilities Act ADA. 1990. Americans with disabilities act. Title II Public Services and Transportation (1990)

1990
[8]

Dustin Adams, Lourdes Morales, and Sri Kurniawan. 2013. A Qualitative Study to Support a Blind Photography Mobile Application. In Proceedings of the 6th International Conference on PErvasive Technologies Related to Assistive Environ- ments (Rhodes, Greece) (PETRA ’13). Association for Computing Machinery, New York, NY, USA, Article 25, 8 pages. doi:10.114...

work page arXiv 2013

Show all 115 references

[9]

I look at it as the king of knowledge

Rudaiba Adnin and Maitraye Das. 2024. " I look at it as the king of knowledge": How Blind People Use and Understand Generative AI Tools. InProceedings of the 26th International ACM SIGACCESS Conference on Computers and Accessibility . 1–14

2024
[10]

Tousif Ahmed, Roberto Hoyle, Kay Connelly, David Crandall, and Apu Kapadia
[11]

Tousif Ahmed, Patrick Shaffer, Kay Connelly, David Crandall, and Apu Kapadia
[12]

Aira. 2020. Aira. https://aira.io/

2020
[13]

Taslima Akter, Tousif Ahmed, Apu Kapadia, and Manohar Swaminathan. 2022. Shared privacy concerns of the visually impaired and sighted bystanders with camera-based assistive technologies. ACM Transactions on Accessible Computing (TACCESS) 15, 2 (2022), 1–33

2022
[14]

Taslima Akter, Bryan Dosono, Tousif Ahmed, Apu Kapadia, and Bryan Semaan
[15]

Jean-Baptiste Alayrac, Jeff Donahue, Pauline Luc, Antoine Miech, Iain Barr, Yana Hasson, Karel Lenc, Arthur Mensch, Katherine Millican, Malcolm Reynolds, et al. 2022. Flamingo: a visual language model for few-shot learning. Advances in neural information processing systems 35 ...

2022
[16]

Gus Alexiou. [n. d.]. Envision Adds ChatGPT AI Sight Assis- tance To Its Smart Glasses For The Blind — forbes.com. https: //www.forbes.com/sites/gusalexiou/2023/04/30/envision-adds-chatgpt- ai-sight-assistance-to-its-smart-glasses-for-the-blind/?sh=d443c835446f. [Accessed 08-04-2025]

2023
[17]

Brewer, and Sarita Schoenebeck

Rahaf Alharbi, Robin N. Brewer, and Sarita Schoenebeck. 2022. Understanding Emerging Obfuscation Technologies in Visual Description Services for Blind and Low Vision People. Proc. ACM Hum.-Comput. Interact. 6, CSCW2, Article 469 (nov 2022), 33 pages. doi:10.1145/3555570

2022 doi
[18]

Maryam Bandukda, Catherine Holloway, Aneesha Singh, and Nadia Berthouze
[19]

Hanoona Bangalath, Muhammad Maaz, Muhammad Uzair Khattak, Salman H Khan, and Fahad Shahbaz Khan. 2022. Bridging the gap between object and image-level representations for open-vocabulary detection. Advances in Neural Information Processing Systems 35 (2022), 33781–33794

2022
[20]

Margaret E Bausch, Joel E Mittler, Ted S Hasselbring, and Donald P Cross. 2005. The Assistive Technology Act of 2004: What Does It Say and What Does It Mean?. Physical Disabilities: Education and Related Services 23, 2 (2005), 59–67

2005
[21]

Andy Beatman. 2024. 6 ways generative AI helps improve accessibility for all with Azure. https://azure.microsoft.com/en-us/blog/6-ways-generative-ai- helps-improve-accessibility-for-all-with-azure/. [Accessed 08-04-2025]

2024
[22]

In Proceedings of the 22nd International ACM SIGACCESS Conference on Computers and Accessibility

PLACES: a framework for supporting blind and partially sighted peo- ple in outdoor leisure activities. In Proceedings of the 22nd International ACM SIGACCESS Conference on Computers and Accessibility . 1–13
[23]

Before, I Asked My Mom, Now I Ask ChatGPT

Cynthia L. Bennett, Jane E, Martez E. Mott, Edward Cutrell, and Meredith Ringel Morris. 2018. How Teens with Visual Impairments Take, Edit, and Share Photos on Social Media. In Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems (Montreal QC, Canada)(C...

2018
[24]

Jeffrey P Bigham, Chandrika Jayant, Hanjie Ji, Greg Little, Andrew Miller, Robert C Miller, Robin Miller, Aubrey Tatarowicz, Brandyn White, Samual White, et al . 2010. Vizwiz: nearly real-time answers to visual questions. In Proceedings of the 23nd annual ACM symposium on User...

2010
[25]

Bigham, Tom Yeh, Chandrika Jayant, Hanjie Ji, Greg Little, Andrew Miller, Robert C

Jeffrey P. Bigham, Tom Yeh, Chandrika Jayant, Hanjie Ji, Greg Little, Andrew Miller, Robert C. Miller, Aubrey Tatarowicz, Brandyn White, and Samuel White
[26]

Oliver Bendel. 2024. How Can Generative AI Enhance the Well-being of Blind?. In Proceedings of the AAAI Symposium Series , Vol. 3. 340–347

2024
[27]

Virginia Braun and Victoria Clarke. 2012. Thematic analysis. American Psycho- logical Association

2012
[28]

Finn Brunton and Helen Nissenbaum. 2015. Obfuscation: A User’s Guide for Privacy and Protest. The MIT Press

2015
[29]

Ruei-Che Chang, Yuxuan Liu, and Anhong Guo. 2024. WorldScribe: Towards Context-Aware Live Visual Descriptions. InProceedings of the 37th Annual ACM Symposium on User Interface Software and Technology . 1–18

2024
[30]

Chaoran Chen, Daodao Zhou, Yanfang Ye, Toby Jia-jun Li, and Yaxing Yao. 2025. CLEAR: Towards Contextual LLM-Empowered Privacy Policy Analysis and Risk Generation for Large Language Model Applications. In Proceedings of the 30th International Conference on Intelligent User Inte...

2025
[31]

Erin Brady, Meredith Ringel Morris, Yu Zhong, Samuel White, and Jeffrey P Bigham. 2013. Visual challenges in the everyday lives of blind people. InProceed- ings of the SIGCHI conference on human factors in computing systems. 2117–2126

2013
[32]

Wei-Lin Chiang, Zhuohan Li, Zi Lin, Ying Sheng, Zhanghao Wu, Hao Zhang, Lianmin Zheng, Siyuan Zhuang, Yonghao Zhuang, Joseph E Gonzalez, et al
[33]

Chun Jie Chong, Chenxi Hou, Zhihao Yao, and Seyed Mohammadjavad Seyed Talebi. 2024. Casper: Prompt Sanitization for Protecting User Privacy in Web- Based Large Language Models. arXiv preprint arXiv:2408.07004 (2024)

2024
[34]

Maitraye Das, Alexander J Fiannaca, Meredith Ringel Morris, Shaun K Kane, and Cynthia L Bennett. 2024. From provenance to aberrations: Image creator and screen reader user perspectives on alt text for AI-generated images. In Proceedings of the 2024 CHI Conference on Human Fact...

2024
[35]

Agebson Rocha Façanha, Ticianne Darin, Windson Viana, and Jaime Sánchez
[36]

Tiffany Chen, Cassandra Lee, Jessica R Mindel, Neska Elhaouij, and Rosalind Picard. 2023. Closer worlds: using generative AI to facilitate intimate conver- sations. In Extended Abstracts of the 2023 CHI Conference on Human Factors in Computing Systems. 1–15

2023
[37]

Jennifer Fereday and Eimear Muir-Cochrane. 2006. Demonstrating rigor using thematic analysis: A hybrid approach of inductive and deductive coding and theme development. International journal of qualitative methods 5, 1 (2006), 80–92

2006
[38]

Joseph L Fleiss, Bruce Levin, and Myunghee Cho Paik. 2013. Statistical methods for rates and proportions . john wiley & sons

2013
[39]

Claudia Flores-Saviaga, Benjamin V Hanrahan, Kashif Imteyaz, Steven Clarke, and Saiph Savage. 2025. The Impact of Generative AI Coding Assistants on Developers Who Are Visually Impaired. arXiv preprint arXiv:2503.16491 (2025)

2025 arXiv
[40]

Ricardo Gonzalez, Jazmin Collins, Shiri Azenkot, and Cynthia Bennett. 2024. Investigating Use Cases of AI-Powered Scene Description Applications for Blind and Low Vision People. arXiv preprint arXiv:2403.15604 (2024)

2024 arXiv
[41]

Gonzalez Penuela, Paul Vermette, Zihan Yan, Cheng Zhang, Keith Vertanen, and Shiri Azenkot

Ricardo E. Gonzalez Penuela, Paul Vermette, Zihan Yan, Cheng Zhang, Keith Vertanen, and Shiri Azenkot. 2022. Understanding How People with Visual Impairments Take Selfies: Experiences and Challenges. In Proceedings of the 24th International ACM SIGACCESS Conference on Computer...

2022
[42]

ACM Transactions on Accessible Computing (TACCESS) 13, 2 (2020), 1–42

O&M indoor virtual environments for people who are blind: A systematic literature review. ACM Transactions on Accessible Computing (TACCESS) 13, 2 (2020), 1–42

2020
[43]

Yuanyuan Feng, Abhilasha Ravichander, Yaxing Yao, Shikun Zhang, and Rex Chen. 2024. Understanding How to Inform Blind and{Low-Vision} Users about Data Privacy through Privacy Question Answering Assistants. In 33rd USENIX Security Symposium (USENIX Security 24) . 2065–2082

2024
[44]

Danna Gurari, Qing Li, Abigale J Stangl, Anhong Guo, Chi Lin, Kristen Grauman, Jiebo Luo, and Jeffrey P Bigham. 2018. Vizwiz grand challenge: Answering visual questions from blind people. In Proceedings of the IEEE conference on computer vision and pattern recognition . 3608–3617

2018
[45]

Yu Hao, Alexey Magay, Hao Huang, Shuaihang Yuan, Congcong Wen, and Yi Fang. 2024. ChatMap: A Wearable Platform Based on the Multi-modal Foun- dation Model to Augment Spatial Cognition for People with Blindness and Low Vision. In 2024 IEEE/RSJ International Conference on Intell...

2024
[46]

Adams, Sri Kurniawan, Hironobu Takagi, and Chieko Asakawa

Susumu Harada, Daisuke Sato, Dustin W. Adams, Sri Kurniawan, Hironobu Takagi, and Chieko Asakawa. 2013. Accessible Photo Album: Enhancing the Photo Sharing Experience for People with Visual Impairment. In Proceedings of the SIGCHI Conference on Human Factors in Computing Syste...

2013
[47]

Jordan Hayes, Smirity Kaushik, Charlotte Emily Price, and Yang Wang. 2019. Cooperative privacy and security: Learning from people with visual impairments and their allies. In Fifteenth Symposium on Usable Privacy and Security (SOUPS 2019). 1–20

2019
[48]

Orit Hazzan and Yael Erez. 2024. Generative AI in computer science education. In Proceedings of the 55th ACM Technical Symposium on Computer Science Education V. 2. 1899–1899

2024
[49]

Darren Guinness, Edward Cutrell, and Meredith Ringel Morris. 2018. Caption crawler: Enabling reusable alternative text descriptions using reverse image search. In Proceedings of the 2018 chi conference on human factors in computing systems. 1–11

2018
[50]

Danna Gurari, Qing Li, Chi Lin, Yinan Zhao, Anhong Guo, Abigale Stangl, and Jeffrey P. Bigham. 2019. VizWiz-Priv: A Dataset for Recognizing the Presence and Purpose of Private Visual Information in Images Taken by Blind People. In Proceedings of the IEEE/CVF Conference on Comp...

2019
[51]

Alekhya Jonnalagedda, Lucy Pei, Suryansh Saxena, Ming Wu, Byung-Cheol Min, Ermine A Teves, Aaron Steinfeld, and M Bernardine Dias. 2014. Enhancing the safety of visually impaired travelers in and around transit stations. Journal Contribution (2014), 1–31

2014
[52]

2023.{GuardLens}: Supporting safer online browsing for people with visual impairments

Smirity Kaushik, Natã M Barbosa, Yaman Yu, Tanusree Sharma, Zachary Kilhof- fer, JooYoung Seo, Sauvik Das, and Yang Wang. 2023.{GuardLens}: Supporting safer online browsing for people with visual impairments. In Nineteenth Sympo- sium on Usable Privacy and Security (SOUPS 2023...

2023
[53]

Majeed Kazemitabaar, Justin Chow, Carl Ka To Ma, Barbara J Ericson, David Weintrop, and Tovi Grossman. 2023. Studying the effect of AI code generators on supporting novice learners in introductory programming. In Proceedings of the 2023 CHI Conference on Human Factors in Compu...

2023
[54]

Seonghee Lee, Maho Kohga, Steve Landau, Sile O’Modhrain, and Hari Subra- monyam. 2024. AltCanvas: A Tile-Based Editor for Visual Content Creation with Generative AI for Blind or Visually Impaired People. In Proceedings of the 26th International ACM SIGACCESS Conference on Comp...

2024
[55]

Sooyeon Lee, Rui Yu, Jingyi Xie, Syed Masum Billah, and John M Carroll. 2022. Opportunities for human-AI collaboration in remote sighted assistance. In Proceedings of the 27th International Conference on Intelligent User Interfaces . 63–78

2022
[56]

Mina Huh, Yi-Hao Peng, and Amy Pavel. 2023. GenAssist: Making image generation accessible. In Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology . 1–17

2023
[57]

Nanna Inie, Jeanette Falk, and Steve Tanimoto. 2023. Designing participatory ai: Creative professionals’ worries and expectations about generative ai. In Extended Abstracts of the 2023 CHI Conference on Human Factors in Computing Systems. 1–8

2023
[58]

KunChang Li, Yinan He, Yi Wang, Yizhuo Li, Wenhai Wang, Ping Luo, Yali Wang, Limin Wang, and Yu Qiao. 2023. Videochat: Chat-centric video understanding. arXiv preprint arXiv:2305.06355 (2023)

2023 arXiv
[59]

Feng Liang, Bichen Wu, Xiaoliang Dai, Kunpeng Li, Yinan Zhao, Hang Zhang, Peizhao Zhang, Peter Vajda, and Diana Marculescu. 2023. Open-vocabulary semantic segmentation with mask-adapted clip. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition ...

2023
[60]

Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. 2024. Visual instruction tuning. Advances in neural information processing systems 36 (2024)

2024
[61]

Ying Ma, Shiquan Zhang, Dongju Yang, Zhanna Sarsenbayeva, Jarrod Knibbe, and Jorge Goncalves. 2025. Raising Awareness of Location Information Vulner- abilities in Social Media Photos using LLMs. arXiv preprint arXiv:2503.20226 (2025)

2025 arXiv
[62]

Muhammad Maaz, Hanoona Rasheed, Salman Khan, and Fahad Shahbaz Khan
[63]

It feels like taking a gamble

Franklin Mingzhe Li, Franchesca Spektor, Meng Xia, Mina Huh, Peter Cederberg, Yuqi Gong, Kristen Shinohara, and Patrick Carrington. 2022. “It feels like taking a gamble”: Exploring perceptions, practices, and challenges of using makeup and cosmetics for people with visual impa...

2022
[64]

Junnan Li, Dongxu Li, Silvio Savarese, and Steven Hoi. 2023. Blip-2: Boot- strapping language-image pre-training with frozen image encoders and large language models. In International conference on machine learning . PMLR, 19730– 19742

2023
[65]

National Federation of the Blind. [n. d.]. Blind Users Innovating and Leading Design. https://nfb.org/programs-services/center-excellence-nonvisual-access/ blind-users-innovating-and-leading-design
[66]

Trisha Neogi. 2024. Protecting People with Disabilities: A Guide for Non- Technical Committee Members in Understanding the Regulations Needed to Design Ethical AI. (2024)

2024
[67]

Bolin Ni, Houwen Peng, Minghao Chen, Songyang Zhang, Gaofeng Meng, Jianlong Fu, Shiming Xiang, and Haibin Ling. 2022. Expanding language-image pretrained models for general video recognition. In European Conference on Computer Vision. Springer, 1–18

2022
[68]

Helen Nissenbaum. 2004. Privacy as contextual integrity.Wash. L. Rev.79 (2004), 119

2004
[69]

CG Docket No. [n. d.]. PUBLIC NOTICE. ([n. d.])
[70]

arXiv preprint arXiv:2306.05424 (2023)

Video-chatgpt: Towards detailed video understanding via large vision and language models. arXiv preprint arXiv:2306.05424 (2023)

2023 arXiv
[71]

Philippe Massonet, Syed Naqvi, Christophe Ponsard, Joseph Latanicki, Benny Rochwerger, and Massimo Villari. 2011. A monitoring and audit logging archi- tecture for data location compliance in federated cloud infrastructures. In 2011 IEEE international symposium on parallel and...

2011
[72]

With most of it being pictures now, I rarely use it

Meredith Ringel Morris, Annuska Zolyomi, Catherine Yao, Sina Bahram, Jef- frey P Bigham, and Shaun K Kane. 2016. " With most of it being pictures now, I rarely use it" Understanding Twitter’s Evolving Accessibility to Blind Users. In Proceedings of the 2016 CHI conference on h...

2016
[73]

Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sand- hini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al
[74]

Gayatri Raman and Erin Brady. 2024. Exploring Use and Perceptions of Genera- tive AI Art Tools by Blind Artists. arXiv preprint arXiv:2409.08226 (2024)

2024 arXiv
[75]

Hanoona Rasheed, Muhammad Uzair Khattak, Muhammad Maaz, Salman Khan, and Fahad Shahbaz Khan. 2023. Fine-tuned clip models are efficient video learners. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 6545–6554

2023
[76]

David Rozenberszki, Or Litany, and Angela Dai. 2022. Language-grounded in- door 3d semantic segmentation in the wild. InEuropean Conference on Computer Vision. Springer, 125–141

2022
[77]

I’m not con- vinced that they don’t collect more than is necessary

Tanusree Sharma, Lin Kyi, Yang Wang, and Asia J Biega. 2024. " I’m not con- vinced that they don’t collect more than is necessary":{User-Controlled} Data Minimization Design in Search Engines. In 33rd USENIX Security Symposium (USENIX Security 24). 2797–2812

2024
[78]

Soobin Park. 2020. Supporting Selfie Editing Experiences for People with Visual Impairments. In Proceedings of the 22nd International ACM SIGACCESS Conference on Computers and Accessibility (Virtual Event, Greece) (ASSETS ’20). Association for Computing Machinery, New York, NY...

2020
[79]

Yuxin Peng and Chong-Wah Ngo. 2006. Clip-based similarity measure for query- dependent clip retrieval and video summarization. IEEE Transactions on Circuits and Systems for Video Technology 16, 5 (2006), 612–627

2006
[80]

Ricardo E Gonzalez Penuela, Ruiying Hu, Sharon Lin, Tanisha Shende, and Shiri Azenkot. 2025. Towards Understanding the Use of MLLM-Enabled Applica- tions for Visual Interpretation by Blind and Low Vision People. arXiv preprint arXiv:2503.05899 (2025)

2025 arXiv
[81]

Person, Shoes, Tree. Is the Person Naked?

Abigale Stangl, Meredith Ringel Morris, and Danna Gurari. 2020. " Person, Shoes, Tree. Is the Person Naked?" What People with Vision Impairments Want in Image Descriptions. In Proceedings of the 2020 chi conference on human factors in computing systems. 1–13

2020
[82]

Dump it, Destroy it, Send it to Data Heaven

Abigale Stangl, Emma Sadjo, Pardis Emami-Naeini, Yang Wang, Danna Gurari, and Leah Findlater. 2023. “Dump it, Destroy it, Send it to Data Heaven”: Blind People’s Expectations for Visual Privacy in Visual Assistance Technologies. In Proceedings of the 20th International Web for...

2023
[83]

Abigale Stangl, Kristina Shiroma, Bo Xie, Kenneth R Fleischmann, and Danna Gurari. 2020. Visual content considered private by people who are blind. In Proceedings of the 22nd International ACM SIGACCESS Conference on Computers and Accessibility. 1–12

2020
[84]

Bolun Sun, Yifan Zhou, and Haiyun Jiang. 2024. Empowering Users in Digital Privacy Management through Interactive LLM-Based Agents. arXiv preprint arXiv:2410.11906 (2024)

2024 arXiv
[85]

Issatay Tokmurziyev, Miguel Altamirano Cabrera, Muhammad Haris Khan, Yara Mahmoud, Luis Moreno, and Dzmitry Tsetserukou. 2025. LLM-Glasses: GenAI- driven Glasses with Haptic Feedback for Navigation of Visually Impaired People. arXiv preprint arXiv:2503.16475 (2025)

2025
[86]

Efrat Tseëlon. 1992. Is the presented self sincere? Goffman, impression manage- ment and the postmodern self. Theory, culture & society 9, 2 (1992), 115–128

1992
[87]

Tanusree Sharma, Vivek C Nair, Henry Wang, Yang Wang, and Dawn Song
[88]

Sheena Vasani. [n. d.]. Be My Eyes AI offers GPT-4-powered support for blind Microsoft customers — theverge.com. https://www.theverge.com/2023/11/15/ 23962709/microsoft-blind-users-open-ai-chatgpt-4-be-my-eyes. [Accessed 08-04-2025]

2023
[89]

Tanusree Sharma, Abigale Stangl, Lotus Zhang, Yu-Yun Tseng, Inan Xu, Leah Findlater, Danna Gurari, and Yang Wang. 2023. Disability-first design and creation of a dataset showing private visual information collected with people who are blind. In Proceedings of the 2023 CHI Conf...

2023
[90]

Fangxun Shu, Lei Zhang, Hao Jiang, and Cihang Xie. 2023. Audio-visual llm for video understanding. arXiv preprint arXiv:2312.06720 (2023)

2023 arXiv
[91]

Herman Wahidin, Jenny Waycott, and Steven Baker. 2018. The challenges in adopting assistive technologies in the workplace for people with visual impairments. In Proceedings of the 30th Australian Conference on Computer- Human Interaction. 432–442

2018
[92]

Bryan Wang, Yuliang Li, Zhaoyang Lv, Haijun Xia, Yan Xu, and Raj Sodhi. 2024. LAVE: LLM-powered agent assistance and language augmentation for video editing. In Proceedings of the 29th International Conference on Intelligent User Interfaces. 699–714

2024
[93]

Mengmeng Wang, Jiazheng Xing, and Yong Liu. 2021. Actionclip: A new para- digm for video action recognition. arXiv preprint arXiv:2109.08472 (2021)

2021 arXiv
[94]

Qi Wu, Peng Wang, Chunhua Shen, Anthony Dick, and Anton Van Den Hengel
[95]

Shaomei Wu, Jeffrey Wieland, Omid Farivar, and Julie Schiller. 2017. Automatic alt-text: Computer-generated image descriptions for blind users on a social network service. Inproceedings of the 2017 ACM conference on computer supported cooperative work and social computing . 1180–1192

2017
[96]

Jingyi Xie, Madison Reddie, Sooyeon Lee, Syed Masum Billah, Zihan Zhou, Chun-Hua Tsai, and John M Carroll. 2022. Iterative design and prototyping of computer vision mediated remote sighted assistance. ACM Transactions on Computer-Human Interaction (TOCHI) 29, 4 (2022), 1–40

2022
[97]

Yu-Yun Tseng, Tanusree Sharma, Lotus Zhang, Abigale Stangl, Leah Findlater, Yang Wang, and Danna Gurari. 2024. BIV-Priv-Seg: Locating Private Con- tent in Images Taken by People With Visual Impairments. arXiv preprint arXiv:2407.18243 (2024)

2024 arXiv
[98]

Jingyi Xie, Rui Yu, He Zhang, Sooyeon Lee, Syed Masum Billah, and John M Carroll. 2024. Bubblecam: Engaging privacy in remote sighted assistance. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems . 1–16

2024
[99]

Violeta Voykinska, Shiri Azenkot, Shaomei Wu, and Gilly Leshed. 2016. How Blind People Interact with Visual Content on Social Networking Services. In Proceedings of the 19th ACM Conference on Computer-Supported Cooperative Work & Social Computing (San Francisco, California...

2016
[100]

Violeta Voykinska, Shiri Azenkot, Shaomei Wu, and Gilly Leshed. 2016. How blind people interact with visual content on social networking services. In Proceedings of the 19th acm conference on computer-supported cooperative work & social computing. 1584–1595

2016
[101]

Xinyu Zhang, Huiyu Xu, Zhongjie Ba, Zhibo Wang, Yuan Hong, Jian Liu, Zhan Qin, and Kui Ren. 2024. Privacyasst: Safeguarding user privacy in tool-using large language model agents. IEEE Transactions on Dependable and Secure Computing (2024)

2024
[102]

Zhuohao (Jerry) Zhang, Smirity Kaushik, JooYoung Seo, Haolin Yuan, Sauvik Das, Leah Findlater, Danna Gurari, Abigale Stangl, and Yang Wang. 2023. Im- ageAlly: A Human-AI Hybrid Approach to Support Blind People in Detecting and Redacting Private Image Content. In Nineteenth Sym...

2023
[103]

Jijie Zhou, Eryue Xu, Yaoyao Wu, and Tianshi Li. 2024. Rescriber: Smaller- LLM-Powered User-Led Data Minimization for Navigating Privacy Trade-offs in LLM-Based Conversational Agent. arXiv preprint arXiv:2410.11876 (2024)

2024 arXiv
[104]

Before, I Asked My Mom, Now I Ask ChatGPT

Deyao Zhu, Jun Chen, Xiaoqian Shen, Xiang Li, and Mohamed Elhoseiny. 2023. Minigpt-4: Enhancing vision-language understanding with advanced large lan- guage models. arXiv preprint arXiv:2304.10592 (2023). “Before, I Asked My Mom, Now I Ask ChatGPT”: Visual Privacy Management w...

2023 arXiv
[105]

In Proceedings of the IEEE conference on computer vision and pattern recognition

Ask me anything: Free-form visual question answering based on knowl- edge from external sources. In Proceedings of the IEEE conference on computer vision and pattern recognition . 4622–4630
[108]

Jingyi Xie, Rui Yu, He Zhang, Syed Masum Billah, Sooyeon Lee, and John M Carroll. 2025. Beyond Visual Perception: Insights from Smartphone Interaction of Visually Impaired Users with Large Multimodal Models. arXiv preprint arXiv:2502.16098 (2025)

2025 arXiv
[110]

Lotus Zhang, Abigale Stangl, Tanusree Sharma, Yu-Yun Tseng, Inan Xu, Danna Gurari, Yang Wang, and Leah Findlater. 2024. Designing Accessible Obfuscation Support for Blind Individuals’ Visual Privacy Management. In Proceedings of the 2024 CHI Conference on Human Factors in Comp...

2024
[111]

Shuning Zhang, Xin Yi, Haobin Xing, Lyumanshan Ye, Yongquan Hu, and Hewu Li. 2024. Adanonymizer: Interactively Navigating and Balancing the Duality of Privacy and Output Performance in Human-LLM Interaction. arXiv preprint arXiv:2410.15044 (2024)

2024 arXiv
[2010]

In Proceedings of the 2010 International Cross Disciplinary Conference on Web Accessibility (W4A) - W4A ’10

VizWiz: nearly real-time answers to visual questions. In Proceedings of the 2010 International Cross Disciplinary Conference on Web Accessibility (W4A) - W4A ’10. ACM Press, Raleigh, North Carolina, 1. doi:10.1145/1805986.1806020

2010
[2015]

In Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems

Privacy concerns and behaviors of people with visual impairments. In Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems. 3523–3532
[2016]

In Twelfth Symposium on Usable Privacy and Security (SOUPS 2016)

Addressing physical safety, security, and privacy for people with visual impairments. In Twelfth Symposium on Usable Privacy and Security (SOUPS 2016). 341–354

2016
[2020]

I am uncomfortable sharing what I can’t see

" I am uncomfortable sharing what I can’t see": Privacy Concerns of the Visually Impaired with Camera Based Assistive Applications. In 29th USENIX Security Symposium (USENIX Security 20) . 1929–1948

1929
[2021]

In International conference on machine learning

Learning transferable visual models from natural language supervision. In International conference on machine learning . PMLR, 8748–8763
[2023]

See https://vicuna

Vicuna: An open-source chatbot impressing gpt-4 with 90%* chatgpt quality. See https://vicuna. lmsys. org (accessed 14 April 2023) 2, 3 (2023), 6

2023
[2024]

I Can’t Believe It’s Not Custodial!

“I Can’t Believe It’s Not Custodial!”: Usable Trustless Decentralized Key Management. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems. 1–16

2024

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

[n. d.]. Employment (Title I) — archive.ada.gov. https://archive.ada.gov/ada_ title_I.htm. [Accessed 07-07-2025]

2025

[2] [2]

[n. d.]. Envision - Perceive Possibility — letsenvision.com. https://www. letsenvision.com/. [Accessed 08-04-2025]

2025

[3] [3]

[n. d.]. Home — bemyeyes.com. https://www.bemyeyes.com/. [Accessed 08-04-2025]

2025

[4] [4]

[n. d.]. Seeing AI - Talking Camera for the Blind — seeingai.com. https://www. seeingai.com/. [Accessed 08-04-2025]

2025

[5] [5]

Be My Eyes

2023. Be My Eyes. https://www.bemyeyes.com/. Accessed: 2023-9-02

2023

[6] [6]

Alessandro Acquisti and Jens Grossklags. 2005. Privacy and rationality in individual decision making. IEEE security & privacy 3, 1 (2005), 26–33

2005

[7] [7]

Disabilities Act ADA. 1990. Americans with disabilities act. Title II Public Services and Transportation (1990)

1990

[8] [8]

Dustin Adams, Lourdes Morales, and Sri Kurniawan. 2013. A Qualitative Study to Support a Blind Photography Mobile Application. In Proceedings of the 6th International Conference on PErvasive Technologies Related to Assistive Environ- ments (Rhodes, Greece) (PETRA ’13). Association for Computing Machinery, New York, NY, USA, Article 25, 8 pages. doi:10.114...

work page arXiv 2013