pith. machine review for the scientific record. sign in

arxiv: 2604.27744 · v1 · submitted 2026-04-30 · 💻 cs.AI

Recognition: unknown

Consumer Attitudes Towards AI in Digital Health: A Mixed-Methods Survey in Australia

Wei Zhou , Rashina Hoda , Joycelyn Ling
Authors on Pith no claims yet

Pith reviewed 2026-05-07 05:19 UTC · model grok-4.3

classification 💻 cs.AI
keywords consumer attitudesAI in healthcaredigital healthmixed-methods surveyAustraliaconsultation summariestrust in AIacceptance of AI
0
0 comments X

The pith

Australian consumers prefer AI-generated consultation summaries for quality and empathy, yet identify them as AI only at chance levels.

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

The paper reports results from a mixed-methods survey of 275 Australian consumers on their readiness, acceptance, trust, and risk perceptions toward AI in digital health. Participants voiced moderate optimism and high ratings for usefulness and ease of use, but also substantial worries about accuracy, safety, and data handling. In a concrete scenario task, the AI-generated summary outperformed the clinician-written version on quality, empathy, and overall usefulness, while correct identification of the AI version occurred near chance. A sympathetic reader would care because the findings indicate that successful deployment depends less on technical performance alone and more on producing outputs that demonstrate clear communication quality and human governance.

Core claim

In the scenario-based evaluation, the AI-generated consultation summary was strongly preferred for quality, empathy, and overall usefulness, yet identification of the AI summary was near chance. Combined with the broader survey results showing moderate optimism alongside concerns about accuracy, safety, and data use, the study establishes that consumers judge AI in healthcare through concrete communication quality and visible human governance, underscoring the need for clinically supervised deployment frameworks beyond technical performance alone.

What carries the argument

The scenario-based evaluation task, in which participants compared an AI-generated consultation summary against a clinician-written one on quality, empathy, usefulness, and source identification.

If this is right

  • Clinically supervised deployment frameworks are required to address consumer concerns about accuracy and safety.
  • Visible human governance in AI outputs can support acceptance even when users cannot reliably detect AI involvement.
  • Consumer judgments of AI depend more on tangible output qualities such as empathy and usefulness than on abstract knowledge that AI is in use.
  • Technical performance metrics alone are insufficient to drive adoption; user-facing communication aspects must be prioritized.

Where Pith is reading between the lines

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

  • The same pattern of quality-driven preference might appear in other patient-facing tools such as AI chatbots or symptom checkers if they maintain high communication standards.
  • Real-world usage studies could test whether the survey-measured optimism and preferences translate into sustained adoption or drop off once privacy risks become concrete.
  • Mandating disclosure of AI use might have limited effect on user attitudes if output quality remains high.

Load-bearing premise

That preferences and attitudes observed in one scenario-based evaluation of a single consultation summary generalize to other AI applications in digital health and that self-reported survey responses accurately predict real-world acceptance, trust, and behavior.

What would settle it

A larger study in which participants use AI-generated health summaries in actual clinical encounters and report lower acceptance or higher rejection rates than predicted by this survey would falsify the central claims.

read the original abstract

AI applications are increasingly being introduced into digital health. While technical performance has advanced rapidly, successful deployment mainly depends on consumer attitudes, especially to patient-facing applications. However, most existing research examines consumer attitudes towards healthcare AI at an abstract level rather than in response to concrete artefacts. We report a mixed-methods survey study in Australia (N=275) examining consumer readiness, acceptance, trust, and risk perceptions of healthcare AI, combined with a scenario-based evaluation of an AI-generated versus clinician-written consultation summary. Participants expressed moderate optimism and strong perceived usefulness and ease of use, but also substantial concerns about accuracy, safety, and data use. In the scenario task, the AI-generated summary was strongly preferred for quality, empathy, and overall usefulness, yet identification of the AI summary was near chance. Findings show that consumers judge AI through concrete communication quality and visible human governance, underscoring the need for clinically supervised deployment frameworks beyond technical performance alone.

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 / 3 minor

Summary. The manuscript reports a mixed-methods survey of N=275 Australian consumers examining attitudes toward AI in digital health, including readiness, acceptance, trust, and risk perceptions, alongside a scenario-based evaluation comparing an AI-generated consultation summary to a clinician-written one. Key findings include moderate optimism with concerns about accuracy, safety, and data privacy; in the scenario, the AI summary was strongly preferred on quality, empathy, and usefulness metrics, while identification of the AI version occurred at near-chance levels. The authors conclude that consumers evaluate AI primarily through concrete communication quality and visible human governance, calling for clinically supervised deployment frameworks beyond technical performance.

Significance. If the core empirical patterns hold, the study contributes useful data on consumer responses to concrete AI artifacts in healthcare rather than abstract attitudes alone. The mixed-methods design and scenario task are strengths, as they move beyond purely hypothetical questions and reveal a preference for the AI-generated summary despite general risk concerns. This could help guide patient-facing AI design by emphasizing output quality and oversight transparency. However, the single-scenario scope and reliance on self-report limit the strength of broader inferences about deployment frameworks.

major comments (3)
  1. [Results (scenario task) and Discussion] The central claim that consumers judge AI 'through concrete communication quality and visible human governance' rests on the scenario task results (AI summary preferred on quality/empathy/usefulness, identification near chance). However, the study provides no direct evidence linking participants' survey-reported governance or oversight concerns to their scenario preferences; the two appear analyzed separately, making the inference from data to this joint mechanism unsupported.
  2. [Discussion and Conclusion] The recommendation for 'clinically supervised deployment frameworks' generalizes from a single consultation-summary scenario to AI applications in digital health broadly. No additional scenarios, error conditions, or application types (e.g., diagnostic tools or monitoring apps) were tested, so the load-bearing extrapolation from one artifact to deployment policy lacks robustness testing.
  3. [Methods and Results] The manuscript relies exclusively on self-reported attitudes, preferences, and behavioral intentions without any behavioral validation measures (e.g., actual willingness to share data or use the summary in a real consultation). This weakens support for claims about real-world acceptance and trust, which are central to the deployment recommendations.
minor comments (3)
  1. [Abstract and Results] The abstract and results sections should more explicitly separate quantitative findings from the general survey versus the scenario task to prevent readers from conflating moderate overall optimism with the specific preference for the AI artifact.
  2. [Methods] Details on the AI model, prompting strategy, and any post-generation human review used to create the 'AI-generated' summary are needed to interpret the 'visible human governance' aspect and to allow replication.
  3. [Results] Tables or figures reporting preference ratings should include effect sizes, confidence intervals, and exact statistical tests used (e.g., for the quality/empathy comparisons) to strengthen the quantitative claims.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive comments on our manuscript. We have carefully considered each point and made revisions to strengthen the paper, particularly by clarifying the interpretive nature of our conclusions and expanding the limitations section. Below we provide point-by-point responses.

read point-by-point responses

  1. Referee: The central claim that consumers judge AI 'through concrete communication quality and visible human governance' rests on the scenario task results (AI summary preferred on quality/empathy/usefulness, identification near chance). However, the study provides no direct evidence linking participants' survey-reported governance or oversight concerns to their scenario preferences; the two appear analyzed separately, making the inference from data to this joint mechanism unsupported.

    Authors: We appreciate the referee's point that the survey and scenario data were not directly linked through statistical analysis. The claim in the manuscript is presented as an integrated interpretation of the mixed-methods findings: participants reported concerns about oversight and data use in the survey, yet demonstrated a clear preference for the AI-generated summary based on its perceived quality, empathy, and usefulness in the scenario, with source identification at chance levels. This pattern suggests that concrete quality can outweigh abstract concerns when human governance is implied by the clinical context. To address the concern, we have revised the Discussion to explicitly describe this as a synthesis of findings rather than a tested mechanism, and we have added text noting the absence of direct moderation analysis as a limitation while highlighting the value of the combined design. revision: partial

  2. Referee: The recommendation for 'clinically supervised deployment frameworks' generalizes from a single consultation-summary scenario to AI applications in digital health broadly. No additional scenarios, error conditions, or application types (e.g., diagnostic tools or monitoring apps) were tested, so the load-bearing extrapolation from one artifact to deployment policy lacks robustness testing.

    Authors: We agree that the study is limited to one scenario type and that broader generalizations require caution. The consultation summary scenario was chosen as it represents a key patient-facing application where communication quality directly impacts user experience and trust. We have revised the Conclusion and Discussion sections to narrow the scope of the recommendation to AI systems for clinical documentation and patient communication, emphasizing the need for human oversight in these areas. We have also added a call for future research to test similar designs across other AI applications, such as diagnostic aids or continuous monitoring tools, to assess the generalizability of these consumer attitudes. revision: partial

  3. Referee: The manuscript relies exclusively on self-reported attitudes, preferences, and behavioral intentions without any behavioral validation measures (e.g., actual willingness to share data or use the summary in a real consultation). This weakens support for claims about real-world acceptance and trust, which are central to the deployment recommendations.

    Authors: This is a fair critique of survey methodology. Our study employs a mixed-methods approach with validated scales for attitudes and a scenario-based task to move beyond purely hypothetical questions, allowing participants to evaluate concrete examples of AI output. However, we recognize that self-reported preferences do not substitute for observed behavior in real clinical settings. We have updated the Limitations section to explicitly acknowledge this gap and recommend that future studies incorporate behavioral measures, such as simulated or actual decisions to use AI summaries or share health data under different governance conditions. The current claims are framed around attitudes and stated preferences, which we believe remain valuable for informing design and policy. revision: yes

Circularity Check

0 steps flagged

No circularity: purely empirical survey with direct observations

full rationale

The paper is a mixed-methods survey (N=275) reporting participant responses on attitudes, trust, and a single scenario comparison of AI vs. clinician summaries. No equations, derivations, fitted parameters, or predictive models are present. All results are stated as direct empirical findings from the data collected. No self-citation chains, uniqueness theorems, or ansatzes are invoked as load-bearing steps. The central claims rest on observed preferences and survey scores rather than any reduction to prior inputs by construction. This is a standard empirical study whose validity concerns (generalization, self-report bias) fall outside circularity analysis.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claims rest on standard survey methodology assumptions rather than new mathematical constructs; no free parameters or invented entities are introduced.

axioms (2)
  • domain assumption Self-reported survey responses accurately reflect participants' true attitudes, trust, and risk perceptions
    Invoked throughout the interpretation of moderate optimism, usefulness ratings, and concerns about accuracy and data use.
  • ad hoc to paper The single scenario-based comparison of one AI-generated versus clinician-written summary is representative of consumer judgment of AI in broader digital health contexts
    Used to support the conclusion that consumers judge AI through concrete communication quality and the call for clinically supervised frameworks.

pith-pipeline@v0.9.0 · 5460 in / 1547 out tokens · 87606 ms · 2026-05-07T05:19:59.646440+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

40 extracted references · 2 canonical work pages

  1. [1]

    & Williams, B

    Bajwa, J., Munir, U., Nori, A. & Williams, B. Artificial intelligence in healthcare: transforming the practice of medicine.Future healthcare journal8, e188–e194 (2021)

    2021

  2. [2]

    Journal of personalized medicine13, 951 (2023)

    Al Kuwaiti, A.et al.A review of the role of artificial intelligence in healthcare. Journal of personalized medicine13, 951 (2023)

    2023

  3. [3]

    Rashid, Z., Ahmed, H., Nadeem, N., Zafar, S. B. & Yousaf, M. Z. The paradigm of digital health: Ai applications and transformative trends.Neural Computing and Applications1–32 (2025)

    2025

  4. [4]

    C., Nickel, G

    Kwong, J. C., Nickel, G. C., Wang, S. C. & Kvedar, J. C. Integrating artificial intelligence into healthcare systems: more than just the algorithm.NPJ Digital Medicine7, 52 (2024)

    2024

  5. [5]

    Silcox, C.et al.The potential for artificial intelligence to transform healthcare: perspectives from international health leaders.NPJ Digital Medicine7, 88 (2024)

    2024

  6. [6]

    Chen, X.et al.Exploring the opportunities of large language models for sum- marizing palliative care consultations: A pilot comparative study.Digital Health 10, 20552076241293932 (2024)

    2024

  7. [7]

    Lee, C., Vogt, K. A. & Kumar, S. Prospects for ai clinical summarization to reduce the burden of patient chart review.Frontiers in Digital Health6, 1475092 (2024)

    2024

  8. [8]

    Clough, R. A. J.et al.Transforming healthcare documentation: harnessing the potential of ai to generate discharge summaries.BJGP open8(2024)

    2024

  9. [9]

    Chua, C. E.et al.Integration of customised llm for discharge summary generation in real-world clinical settings: a pilot study on russell gpt.The Lancet Regional Health–Western Pacific51(2024)

    2024

  10. [10]

    Shemtob, L.et al.Comparing ai-versus clinician-authored summaries of simulated primary care electronic health records.medRxiv2025–02 (2025)

    2025

  11. [11]

    Fraile Navarro, D.et al.Expert evaluation of large language models for clinical dialogue summarization.Scientific Reports15, 1195 (2025)

    2025

  12. [12]

    Koh, M. C. Y.et al.Using chatgpt for writing hospital inpatient discharge summaries–perspectives from an inpatient infectious diseases service.BMC Health Services Research25, 221 (2025)

    2025

  13. [13]

    T., Amara, D., Bhattacharya, A

    Young, A. T., Amara, D., Bhattacharya, A. & Wei, M. L. Patient and general pub- lic attitudes towards clinical artificial intelligence: a mixed methods systematic review.The lancet digital health3, e599–e611 (2021). 24

    2021

  14. [14]

    & Dharanikota, S

    Esmaeilzadeh, P., Mirzaei, T. & Dharanikota, S. Patients’ perceptions toward human–artificial intelligence interaction in health care: experimental study. Journal of medical Internet research23, e25856 (2021)

    2021

  15. [15]

    Nuccetelli, F.et al.The use of artificial intelligence in healthcare as perceived by the citizens and patients: a narrative review of the literature.European Journal of Public Health35, 1092–1099 (2025)

    2025

  16. [16]

    Williamson, S. M. & Prybutok, V. Balancing privacy and progress: a review of privacy challenges, systemic oversight, and patient perceptions in ai-driven healthcare.Applied Sciences14, 675 (2024)

    2024

  17. [17]

    Ethicsandgovernanceofartificialintelligenceforhealth: Whoguidance

    World Health Organization. Ethics and governance of artificial intelligence for health (2021). URL https://www.who.int/publications/i/item/9789240029200

    work page arXiv 2021

  18. [18]

    Gundlack, J.et al.Patients’ perceptions of artificial intelligence acceptance, challenges, and use in medical care: Qualitative study.Journal of Medical Internet Research27, e70487 (2025)

    2025

  19. [19]

    Journal of Participatory Medicine17, e69564 (2025)

    Foresman, G.et al.Patient perspectives on artificial intelligence in health care: focus group study for diagnostic communication and tool implementation. Journal of Participatory Medicine17, e69564 (2025)

    2025

  20. [20]

    & Kunde, W

    Reis, M., Reis, F. & Kunde, W. Public perception of physicians who use artificial intelligence.JAMA Network Open8, e2521643–e2521643 (2025)

    2025

  21. [21]

    & Xing, C

    Ding, X. & Xing, C. Trust in ai vs. human doctors: The roles of subjective understanding, perceived epistemic authority and social proof.Acta Psychologica 261, 105945 (2025)

    2025

  22. [22]

    Cinalioglu, K.et al.Exploring differential perceptions of artificial intelligence in health care among younger versus older canadians: results from the 2021 canadian digital health survey.Journal of Medical Internet Research25, e38169 (2023)

    2021

  23. [23]

    & Kunde, W

    Reis, M., Reis, F. & Kunde, W. Influence of believed ai involvement on the perception of digital medical advice.Nature Medicine1–3 (2024)

    2024

  24. [24]

    A., Schaarup, J

    Isaksen, A. A., Schaarup, J. R., Bjerg, L. & Hulman, A. Changes in public perception of artificial intelligence in healthcare after exposure to chatgpt.npj Digital Medicine(2025)

    2025

  25. [25]

    Chen, D.et al.Patient perceptions of empathy in physician and artificial intelli- gence chatbot responses to patient questions about cancer.npj Digital Medicine 8, 275 (2025)

    2025

  26. [26]

    Ovsyannikova, D., de Mello, V. O. & Inzlicht, M. Third-party evaluators perceive ai as more compassionate than expert humans.Communications Psychology3, 25 4 (2025)

    2025

  27. [27]

    E., Paul, H

    Goodman, K. E., Paul, H. Y. & Morgan, D. J. Ai-generated clinical summaries require more than accuracy.JAMA331, 637–638 (2024)

    2024

  28. [28]

    Australia’s ai ethics principles (2019)

    Australian Government, Department of Industry, Science and Resources. Australia’s ai ethics principles (2019). URL https://www.industry.gov.au/ publications/australias-artificial-intelligence-ethics-framework

    2019

  29. [29]

    National framework for the assurance of artificial intelligence in government (2024)

    Australian Government, Department of Finance. National framework for the assurance of artificial intelligence in government (2024). URL https://www. finance.gov.au/government/public-data/data-and-digital-ministers-meeting/ national-framework-assurance-artificial-intelligence-government

    2024

  30. [30]

    Arti- ficial intelligence (ai) and medical device software (2025)

    Australian Government, Department of Health, Disability and Ageing. Arti- ficial intelligence (ai) and medical device software (2025). URL https:// www.tga.gov.au/products/medical-devices/software-and-artificial-intelligence/ manufacturing/artificial-intelligence-ai-and-medical-device-software

    2025

  31. [31]

    Ai principles (2019)

    OECD. Ai principles (2019). URL https://www.oecd.org/en/topics/sub-issues/ ai-principles.html

    2019

  32. [32]

    Lekadir, K.et al.Future-ai: International consensus guideline for trustworthy and deployable artificial intelligence in healthcare.bmj388(2025)

    2025

  33. [33]

    Chew, H. S. J. & Achananuparp, P. Perceptions and needs of artificial intelligence in health care to increase adoption: scoping review.Journal of medical Internet research24, e32939 (2022)

    2022

  34. [34]

    A., Lungren, M

    Jindal, J. A., Lungren, M. P. & Shah, N. H. Ensuring useful adoption of generative artificial intelligence in healthcare.Journal of the American Medical Informatics Association31, 1441–1444 (2024)

    2024

  35. [35]

    J.et al.Barriers and facilitators to utilizing digital health technologies by healthcare professionals.NPJ digital medicine6, 161 (2023)

    Borges do Nascimento, I. J.et al.Barriers and facilitators to utilizing digital health technologies by healthcare professionals.NPJ digital medicine6, 161 (2023)

    2023

  36. [36]

    Zhou, W.et al.From framework to practice: Designing a real-world telehealth application for palliative care.arXiv preprint arXiv:2512.13693(2025)

    work page arXiv 2025

  37. [37]

    & Garc´ ıa-G´ omez, J

    S´ aez, C., Ferri, P. & Garc´ ıa-G´ omez, J. M. Resilient artificial intelligence in health: synthesis and research agenda toward next-generation trustworthy clinical decision support.Journal of Medical Internet Research26, e50295 (2024)

    2024

  38. [38]

    K., Mabitsela, T

    Khoza, T. K., Mabitsela, T. & Nel, P. Technology readiness, technology accep- tance, and work engagement: A mediational analysis.SA Journal of Industrial Psychology50, 2131 (2024). 26

    2024

  39. [39]

    Fam, K.-S.et al.Modeling new technology readiness and acceptance in the case of b2b marketing employees.Journal of Business-to-Business Marketing32, 1–30 (2025)

    2025

  40. [40]

    & Baldassarre, M

    Storey, M.-A., Hoda, R., Maciel Paz Milani, A. & Baldassarre, M. T. Guiding principles for mixed methods research in software engineering.Empirical Software Engineering30, 138 (2025). 27

    2025