The reviewed record of science sign in
Pith

arxiv: 2606.05055 · v1 · pith:WMODK2GN · submitted 2026-06-03 · cs.HC

"A Glimpse, Not a Gaze": Using Generative AI to Balance Privacy and Awareness in Inter-generational Caregiving

Reviewed by Pith2026-06-28 04:02 UTCgrok-4.3pith:WMODK2GNopen to challenge →

classification cs.HC
keywords generative AIprivacy preservationinter-generational caregivingvisual abstractionexperience sampling methodolder adultsadult childrenawareness tools
0
0 comments X

The pith

Generative AI can produce abstract visual summaries that allow adult children to maintain awareness of older adults' daily activities while respecting their privacy.

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

This paper investigates the use of generative AI to create abstract sketches of daily activities as a way to resolve the tension between caregivers' need for information and older adults' need for privacy. It describes a planned 10-day experience sampling study involving pairs of older adults and their adult children, where participants evaluate pre-generated AI sketches and report their willingness to share or receive them. The study seeks to measure generational differences in privacy boundaries and to develop concrete design guidelines for AI tools in caregiving. If successful, this approach could replace intrusive video monitoring with less invasive visual abstractions that still convey essential context.

Core claim

The central claim is that generative AI can generate privacy-preserving visual summaries of daily life that support inter-generational caregiving by providing sufficient awareness without compromising the older adult's autonomy and dignity.

What carries the argument

Pre-generated AI sketches evaluated through daily ESM prompts for willingness to share or receive, serving as visual summaries that abstract away raw details.

If this is right

  • Quantifying the privacy mismatch will reveal specific differences in what each generation finds acceptable.
  • Actionable design guidelines will emerge for the level of abstraction needed in such tools.
  • AI-mediated tools using this method could support connection while protecting dignity better than traditional monitoring.
  • Boundary-setting behaviors identified in interviews will inform how to implement these systems in practice.

Where Pith is reading between the lines

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

  • These guidelines might extend to other family monitoring contexts beyond caregiving, such as checking on distant relatives.
  • Testing the sketches in actual deployed systems rather than pre-generated ones could reveal additional practical issues.
  • Broader adoption could shift norms around privacy in aging-in-place technologies.

Load-bearing premise

That participants' self-reports on willingness to share or receive the AI sketches in the study prompts will accurately reflect their real-world boundary-setting behaviors in caregiving situations.

What would settle it

A finding that participants are unwilling to share or receive the sketches at similar rates as they would raw video feeds, or that the sketches fail to provide enough awareness according to the adult children, would indicate the method does not achieve the intended balance.

Figures

Figures reproduced from arXiv: 2606.05055 by James R. Wallace, Keiko Katsuragawa, Zixi Christina Li.

Figure 1
Figure 1. Figure 1: Sample of AI generated sketches of participants [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
read the original abstract

As older adults increasingly prefer to age in place, their adult children often assume the role of informal caregivers. This dynamic creates a distinct tension between the adult child's need for awareness and the older adult's fundamental right to privacy. Traditional monitoring technologies, such as raw video feeds, often compromise the older adult's autonomy. To address this challenge, this study explores the use of generative Artificial Intelligence (GenAI) to create abstract, privacy-preserving ``visual summaries'' of daily activities. We design a 10-day Experience Sampling Method (ESM) study with dyads consisting of older adults and their adult children. Through daily smartphone prompts, participants report their current context and evaluate pre-generated AI sketches, indicating their willingness to share or receive these images. Follow-up interviews will further investigate participants' boundary-setting behaviours. This research aims to quantify the privacy mismatch between generations and provide actionable design guidelines for applying visual abstraction in AI-mediated caregiving tools, ultimately supporting inter-generational connection while protecting user dignity.

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

1 major / 2 minor

Summary. The manuscript presents a study protocol for a 10-day Experience Sampling Method (ESM) investigation with older adult–adult child dyads. Daily smartphone prompts ask participants to report context and rate pre-generated generative-AI sketches for willingness to share or receive; follow-up interviews explore boundary-setting. The stated aims are to quantify generational privacy mismatch and produce actionable design guidelines for visual-abstraction techniques in AI-mediated caregiving tools.

Significance. If executed and the measured responses prove predictive, the work could supply empirical grounding for privacy-preserving visual summaries in informal caregiving—an area of clear societal relevance. The dyadic, two-generation design and explicit focus on both awareness and dignity are constructive. Because the manuscript contains no collected data, the significance remains prospective and hinges on whether the protocol’s measurement approach can support the intended claims.

major comments (1)
  1. The protocol’s central claim—to quantify the privacy mismatch and derive design guidelines—rests on the assumption that ESM self-reports of willingness to share or receive pre-generated sketches will validly proxy real-world boundary-setting behavior. The described method uses non-context-matched, pre-generated images and contains no actual sharing consequences or real-time capture, creating a gap between the measured responses and the consequential caregiving scenarios the guidelines are intended to address.
minor comments (2)
  1. The method description does not specify the source, selection criteria, or generation parameters for the pre-generated sketches, leaving unclear how visual abstraction levels will be controlled or varied across prompts.
  2. No sample-size justification, power analysis, or recruitment targets for the dyads are provided, which affects the feasibility of the planned quantification of generational differences.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive review of our study protocol manuscript. The major comment raises an important point about the validity of our measurement approach, which we address directly below. We have revised the manuscript to better articulate the scope and limitations of the proposed method.

read point-by-point responses
  1. Referee: The protocol’s central claim—to quantify the privacy mismatch and derive design guidelines—rests on the assumption that ESM self-reports of willingness to share or receive pre-generated sketches will validly proxy real-world boundary-setting behavior. The described method uses non-context-matched, pre-generated images and contains no actual sharing consequences or real-time capture, creating a gap between the measured responses and the consequential caregiving scenarios the guidelines are intended to address.

    Authors: We acknowledge that self-reported willingness ratings on pre-generated sketches cannot fully replicate the dynamics of real-world boundary-setting, where actual consequences, real-time capture, and matched contexts influence decisions. This is an inherent limitation of any perception-based protocol that prioritizes participant safety and ethical constraints over direct observation of sharing. Our design uses ESM to sample responses in participants' natural environments while they report their current context, providing a degree of ecological grounding; pre-generated sketches ensure stimulus consistency and avoid privacy risks from live capture. Follow-up interviews are intended to probe the reasoning behind ratings. We agree the original framing overstated the direct applicability to consequential scenarios. The revised manuscript now (1) explicitly states that the study measures reported willingness rather than observed behavior, (2) qualifies the resulting design guidelines as preliminary and perception-informed, and (3) adds a dedicated limitations subsection discussing the proxy gap and calling for future validation with deployed systems. revision: yes

Circularity Check

0 steps flagged

No circularity: descriptive ESM protocol with no derivations or self-referential modeling

full rationale

The paper presents a study protocol using Experience Sampling Method prompts and follow-up interviews to collect participant self-reports on willingness to share/receive pre-generated AI sketches. No equations, fitted parameters, predictions derived from inputs, or self-citation chains appear in the provided text. The central aim of quantifying generational privacy mismatch is framed as an empirical outcome of the data collection rather than a quantity that reduces to its own assumptions by construction. This matches the default expectation for non-circular papers; the work is self-contained as a descriptive protocol without load-bearing reductions.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is a study protocol paper with no mathematical content. No free parameters, axioms, or invented entities are introduced because there are no derivations or quantitative models.

pith-pipeline@v0.9.1-grok · 5714 in / 939 out tokens · 27265 ms · 2026-06-28T04:02:39.867424+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

21 extracted references · 6 canonical work pages

  1. [1]

    The Computational Geometry Algorithms Library , author =

  2. [2]

    Menelaos Karavelas , subtitle =

  3. [3]

    The Computational Geometry Algorithms Library , subtitle =

    Menelaos Karavelas , editor =. The Computational Geometry Algorithms Library , subtitle =

  4. [4]

    The Parmap library , author =

  5. [5]

    Christopher Anderson and Sophia Drossopoulou , title =

  6. [6]

    2026 , url =

    Qualtrics. 2026 , url =

  7. [7]

    , title =

    Homaeian, Leila and Duong, Vanessa and Katsuragawa, Keiko and Wallace, James R. , title =. 2025 , issue_date =. doi:10.1145/3701187 , journal =

  8. [8]

    Designing for elders: exploring the complexity of relationships in later life , year =

    Lindley, Si\^. Designing for elders: exploring the complexity of relationships in later life , year =. Proceedings of the 22nd British HCI Group Annual Conference on People and Computers: Culture, Creativity, Interaction - Volume 1 , pages =

  9. [9]

    Psychology and aging , volume =

    Longitudinal study of social support and meaning in life , author =. Psychology and aging , volume =

  10. [10]

    Exploring

    Yun, Sojeong and Sohn, Yongjae and Lim, Youn-kyung , month = jul, year =. Exploring. Proceedings of the 2025. doi:10.1145/3715336.3735826 , abstract =

  11. [11]

    Sarma, M

    Park, Soobin and Kim, Hankyung and Lim, Youn-kyung , month = apr, year =. Reimagining. Proceedings of the 2025. doi:10.1145/3706598.3713722 , abstract =

  12. [12]

    Investigating

    Swienty, Lena and Panskus, Rebecca and Miketta, Stefanie and Kircili, Beyza and Cheng, Sen and Marky, Karola , month = dec, year =. Investigating. Proceedings of the 24th. doi:10.1145/3771882.3771889 , language =

  13. [13]

    IEEE Transactions on Visualization and Computer Graphics , author =

    Casual. IEEE Transactions on Visualization and Computer Graphics , author =. 2007 , keywords =. doi:10.1109/TVCG.2007.70541 , abstract =

  14. [14]

    Proceedings of the ACM on Human-Computer Interaction , author =

    Privacy vs. Proceedings of the ACM on Human-Computer Interaction , author =. 2023 , pages =. doi:10.1145/3610202 , abstract =

  15. [15]

    Proceedings of the CHI Conference on Human Factors in Computing Systems , pages =

    ‘To Click or not to Click’: Back to Basic for Experience Sampling for Office Well-being in Shared Office Spaces , author =. Proceedings of the CHI Conference on Human Factors in Computing Systems , pages =. 2024 , doi =

  16. [16]

    Designing Interactive Systems Conference , pages =

    Flexibility and Social Disconnectedness: Assessing University Students’ Well-Being Using an Experience Sampling Chatbot and Surveys Over Two Years of COVID-19 , author =. Designing Interactive Systems Conference , pages =. 2022 , doi =

  17. [17]

    Proceedings of the ACM on Human-Computer Interaction , volume =

    The Experience Sampling Method and its Tools: A Review for Developers, Study Administrators, and Participants , author =. Proceedings of the ACM on Human-Computer Interaction , volume =. 2023 , doi =

  18. [18]

    Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies , volume =

    How Much is Too Much?: Understanding the Information Needs of Parents of Young Children , author =. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies , volume =. 2019 , doi =

  19. [19]

    Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies , volume =

    End-User Development of Experience Sampling Smartphone Apps -Recommendations and Requirements , author =. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies , volume =. 2020 , doi =

  20. [20]

    CHI Conference on Human Factors in Computing Systems , pages =

    Sad or just jealous? Using Experience Sampling to Understand and Detect Negative Affective Experiences on Instagram , author =. CHI Conference on Human Factors in Computing Systems , pages =. 2022 , doi =

  21. [21]

    The Gerontologist , volume=

    The meaning of ``aging in place'' to older people , author=. The Gerontologist , volume=. 2012 , publisher=