Pith. sign in

REVIEW 4 major objections 4 minor 43 references

The paper claims that, in a high-trust civic AI setting, neither transparency nor control mechanisms shift data donation behavior, with donation rates uniformly high at 91.7% and Bayesian checks supporting genuine null effects.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 08:37 UTC pith:GG7DRMIW

load-bearing objection A clean, honestly reported null result that is more interesting for what the control condition reveals than for what the headline claims; the Bayesian 'evidence of absence' is prior-sensitive and overstates the case. the 4 major comments →

arxiv 2607.21044 v1 pith:GG7DRMIW submitted 2026-07-23 stat.AP econ.EMstat.OT

Trust by Context, Not by Design? A Quantitative Study of Data Donation Willingness for Open-Source Civic AI in Switzerland

Sabine Wildemann , Daniel Ambach This is my paper
classification stat.AP econ.EMstat.OT
keywords data donationcivic AIData Nutrition Labelgranular consent dashboardprivacy calculusdynamic consentmanipulation checkSwitzerland
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper aims to establish that interface design—specifically a Data Nutrition Label for transparency and a granular consent dashboard for control—has little causal leverage on Swiss residents' willingness to donate chatbot data to an open-source civic AI. Across 205 participants in a 2x2 experiment, donation rates were uniformly high (91.7%) regardless of condition, and Bayesian robustness checks supported genuine null effects. The dashboard did raise perceived control, but donors used it to restrict how their data could be used rather than to donate more. The authors interpret this through the privacy calculus: perceived public benefit was high and perceived risk low because institutional trust was high and anonymized queries were seen as non-personal. The study matters because it suggests that in trustworthy civic settings, consent design is less a tool for raising participation than a mechanism for letting residents negotiate the terms of contribution.

Core claim

On the paper's own terms, the central discovery is a null result with positive content: in a high-trust civic setting, data donation decisions are driven by context rather than by interface design. In a 2x2 experiment with 205 Swiss residents, a Data Nutrition Label (transparency) and a granular consent dashboard (control) produced no significant differences in donation rates, which hovered between 90.0% and 93.3% across conditions (overall 91.7%). Logistic regressions gave odds ratios near 1 (transparency OR=1.02; control OR=1.46; interaction OR=0.89), and beta-binomial Bayes factors provided strong to extreme evidence for the null. The control manipulation did increase perceived control, a

What carries the argument

The experimental apparatus consists of two stimuli plus a behavioral outcome. The Data Nutrition Label is a visual, icon-based summary of model provenance, data sources, privacy safeguards, and output generation—designed to trigger heuristic transparency. The granular consent dashboard is a sequential configuration interface (data scope, purpose, storage location, retention period) operationalizing dynamic consent. The decision context is a functional civic AI chatbot, grounded in official Swiss ballot data, which gives the donation decision ecological validity. Analysis uses chi-square tests, nested logistic regressions, and beta-binomial Bayes factors, with manipulation checks (perceived t

Load-bearing premise

The load-bearing premise is that the transparency manipulation actually worked—that the Data Nutrition Label increased perceived transparency—but the manipulation check found no such increase (p=.480), so the paper's conclusion that transparency does not affect donation rests on an untested manipulation.

What would settle it

Run the same 2x2 experiment in a low-trust or commercial frame—say, a private company or a foreign-hosted service—with real (non-simulated) data transfer and a graded willingness scale; if donation drops well below 91.7% or the label/dashboard show measurable effects, the 'context, not design' conclusion would be falsified. A simpler check: measure donation rates when the identical interface is paired with commercial rather than civic framing.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • If the paper is right, adding richer transparency and control widgets to civic AI consent screens will not increase donation rates in settings with high institutional trust and salient public benefit.
  • The value of granular consent shifts from persuading users to donate to letting them negotiate scope, purpose, storage, and retention; designers should measure configuration choices, not just yes/no donation.
  • Because the control manipulation worked while donation did not change, high donation and boundary-setting can coexist; privacy-preserving defaults matter even at near-universal donation rates.
  • The Bayesian evidence (BF01 of 7.99–428.76) indicates the null effects are not merely underpowered; they are genuine absences of treatment effects in this context.
  • The findings imply that institutional trust and framing do more work than interface design, so consent-design experiments in low-trust or commercial contexts may yield different results.

Where Pith is reading between the lines

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

  • Editorial inference: Because the transparency manipulation check failed (MC-T p=.480), the paper does not actually test whether transparency can move donation; it tests a label that participants did not perceive as more transparent. A stronger or more novel transparency stimulus could still show an effect.
  • Editorial inference: The simulated donation and civic academic framing may create demand characteristics; a replication with real data transfer or a neutral, non-civic frame would reveal how much of the 91.7% ceiling is an artifact of the setting.
  • Editorial inference: The dashboard data suggest sovereignty is the dimension that matters: 68.6% chose Swiss or EU storage and 67.6% academic-only purpose. A direct test would vary server location or purpose while holding the interface identical and measure donation willingness.
  • Editorial inference: The privacy-calculus explanation could be probed by varying query sensitivity (e.g., innocuous vs. politically charged topics) to see whether the ceiling breaks when perceived risk rises; that would convert the post-hoc interpretation into a testable prediction.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper reports a 2x2 between-subjects experiment (N=205) in which Swiss residents interacted with a chatbot based on the Apertus-70B model and then chose whether to donate their anonymized conversation for training an open-source civic AI. The two factors were presence of a Data Nutrition Label and a granular consent dashboard. Donation rates were uniformly high (91.7% overall) with no significant main or interaction effects; the dashboard increased perceived control but not donation. Bayesian robustness checks (Beta(1,1) priors) were interpreted as strong evidence of absence. Open-ended responses emphasized public benefit and low perceived sensitivity. The authors conclude that in high-trust civic contexts, context matters more than interface design, and control functions mainly to let donors define terms of use.

Significance. If the null effects were robust, the paper would be a useful contribution to consent-mechanism research in civic AI, showing ceiling effects in high-trust settings and documenting how users exercise granular restrictions. The strengths include the embedded functional chatbot with sovereign infrastructure, the structured analysis plan, the reproducible codebook for qualitative coding, and the use of a real ballot-information task. However, the primary interpretive claims go beyond what the design can support: the transparency manipulation failed its manipulation check, the Bayesian absence evidence is prior-sensitive under ceiling conditions, and the study is underpowered for plausible small effects. The descriptive results on dashboard configuration choices and qualitative themes remain informative despite these limitations.

major comments (4)
  1. [Section 4.2, Table 5; Section 4.3; Abstract] The transparency manipulation failed its manipulation check (U=5518.0, p=.480, rank-biserial r=0.06). Consequently, H1 is not merely unsupported; it is untested. The null behavioral effect for the Data Nutrition Label cannot be interpreted as evidence that transparency has no effect because participants in the label conditions did not perceive greater transparency. The abstract and conclusion nevertheless state that 'neither transparency nor control significantly affected donation' and that 'Bayesian checks confirmed the absence of treatment effects.' The limitation in Section 5.2 partially acknowledges the failed manipulation, but the central claim is not qualified accordingly. This is load-bearing for the paper's conclusion that interface design has little leverage.
  2. [Section 4.3, 'Bayesian robustness check'] The Bayes factors are computed with Beta(1,1) priors on cell probabilities. With observed rates near 0.92 and cell sizes of 45–57, the independent-rates alternative under Beta(1,1) places most prior mass on very large probability differences (rates near 0 or 1), making the common-rate model look overwhelmingly better. BF01 values of 10.30, 7.99, and 428.76 are therefore prior-sensitive artifacts, not calibrated evidence about effect sizes of interest. A sensitivity analysis with informative priors centered on small effects, or a prior directly on the odds ratio, is required before claiming 'genuine absence.' This is load-bearing because the 'no leverage' conclusion depends on converting non-significance into evidence of absence.
  3. [Section 3.5 and Table 2] The power analysis assumed OR>=2.0, but with a baseline donation rate of 90%, OR=2.0 corresponds to roughly a 4.7 percentage-point increase. With n≈50 per arm, the design has low power to detect such effects, and the authors acknowledge power is only about 0.75 for this assumed size. Thus the experiment cannot rule out practically meaningful smaller effects. The Bayesian claim does not repair this problem because it uses the same low-information data. The conclusion that 'interface design has little leverage' exceeds what this design can establish.
  4. [Title, Section 5.1, Section 6] The study uses a single high-trust civic context (Swiss academic, non-commercial, Apertus infrastructure) and does not manipulate context. The claim encapsulated in the title, 'Trust by Context, Not by Design?', is therefore not directly tested. The high baseline rate could be due to social desirability or the simulated donation (both acknowledged in Section 5.2), and the single-cell design cannot compare context against design. The conclusion should be reframed as a descriptive finding within this particular high-trust context, not as a comparative claim that context matters more than design.
minor comments (4)
  1. [Section 3.4, Figure 4] The stimulus is described as an independently designed artifact following the Data Nutrition Project framework, but the text sometimes refers to it simply as 'the Data Nutrition Label.' Clarify early that this is an adapted, context-specific version, not the original MIT label, to avoid confusion for readers familiar with the original.
  2. [Table 12] The two storage rows ('Swiss servers only' and 'Swiss or EU servers') are not mutually exclusive, as the note states. Please clarify the response-option structure: is 'Swiss or EU' a separate option, or a collapsed category? The current presentation makes the 68.6% figure ambiguous.
  3. [Section 4.4, qualitative coding] The deterministic keyword-matching approach is reproducible, but it may miss paraphrases and context-dependent meanings. The paper states this only implicitly; a sentence acknowledging this limitation and its effect on theme frequencies would be helpful.
  4. [Abstract and Introduction] The phrase 'Bayesian checks confirmed the absence of treatment effects' is too strong given the prior sensitivity described above. Even after revision, the abstract should be worded to reflect that these checks are consistent with null effects under diffuse priors, not that absence is established.

Circularity Check

0 steps flagged

No significant circularity: the analysis is an empirical null-result study with no fitted predictions, no self-citation chain, and no result that reduces to its inputs by construction.

full rationale

The paper's claims are empirical and self-contained: donation rates are directly measured, hypotheses are tested with chi-square tests and logistic regression, and no parameter is fitted to one part of the data and then 'predicted' for a closely related quantity. The Bayesian robustness check uses a stated Beta(1,1) prior and is therefore a transparent statistical assumption, not a hidden fit; whether that prior is well chosen is a statistical validity question, not circularity. The manipulation-check failure for transparency (MC-T p=.480) is explicitly acknowledged as limiting interpretation of H1, which weakens the paper's claim but does not constitute circular reasoning. The 'context, not design' interpretation is an explanatory narrative built around the null result, not a derivation from the hypotheses. There are no self-citations among the references and no imported uniqueness theorems. The study's ceiling effect is a real limitation, but it does not make the analysis circular: the outcome was not defined in terms of the explanation, nor was the conclusion forced by a fitted input.

Axiom & Free-Parameter Ledger

2 free parameters · 4 axioms · 0 invented entities

This is an empirical behavioral study, not a derivation; the central claims rest on statistical modeling and domain assumptions, not on fitted parameters. The main choices are the prior in the Bayesian robustness check and the a-priori effect size in the power analysis.

free parameters (2)
  • Beta(1,1) prior in Bayesian robustness check
    The Bayes factors (BF01 = 10.30, 7.99, 428.76) are computed under a Beta(1,1) prior, chosen ad hoc; the 'strong evidence of absence' conclusions are conditional on this prior and no sensitivity analysis is reported.
  • Assumed effect size OR >= 2.0 in power analysis
    The target N=200 was chosen assuming a medium effect (OR>=2.0), giving power ~0.75, below 0.80; this assumption underlies the interpretation that non-significance is evidence, but ceiling further reduces power.
axioms (4)
  • domain assumption Simulated donation decisions measure the same latent construct as real data donation
    The paper's central behavioral outcome is a simulated, non-stored donation; generalizing to real donation behavior is assumed (Section 3.1, Phase 9; Section 5.2).
  • domain assumption Privacy calculus (risk/benefit tradeoff) is the operative decision model in this civic context
    The interpretation that high perceived benefit and low perceived risk explain 91.7% donation presupposes the privacy calculus model (Section 2.1), which is not itself tested.
  • domain assumption The chatbot interaction was sufficiently realistic to evoke genuine donation intentions
    Participants were required to ask 2-3 questions; the paper assumes this minimal interaction is enough to ground a meaningful behavioral decision (Phase 5, Section 3.1).
  • standard math Beta-binomial conjugate model is an appropriate statistical model for computing the reported Bayes factors
    The paper invokes a beta-binomial conjugate model without deriving it; for the four-condition comparison the appropriate model would be multinomial, and the paper does not justify the beta-binomial choice (Section 4.3).

pith-pipeline@v1.3.0-alltime-deepseek · 19185 in / 15276 out tokens · 158723 ms · 2026-08-01T08:37:21.493589+00:00 · methodology

0 comments
read the original abstract

Civic AI systems increasingly support democratic participation, yet interactions with them may reveal sensitive political views, creating tension between improving AI models and residents' expectations of privacy and consent. This study examines the conditions of transparency and user control under which Swiss residents are willing to donate their anonymized chatbot conversations to train an open-source AI model. A 2x2 between-subjects factorial design evaluated how a Data Nutrition Label and a granular consent dashboard influence donation decisions. The experiment was delivered via a multilingual online survey featuring a custom chatbot powered by the Apertus-70B model. Analysis of the 205 participants revealed that neither transparency nor control significantly affected donation behavior. Rates were uniformly high (91.7% overall), producing a ceiling effect, and Bayesian checks confirmed the absence of treatment effects. The dashboard raised perceived control but not donation, and high-control participants actively restricted their data-use settings. A qualitative analysis of 120 open-ended responses indicates that residents framed donation as a contribution to the public good, motivated by democratic participation, an open-source model, and research, while many regarded their anonymized queries as non-personal and therefore low in risk. Interpreted through the privacy calculus, a high perceived benefit coincided with a low perceived risk under high institutional trust, so both sides of the trade-off aligned and interface design had little leverage. Offering control served less to raise donation than to let residents define the terms of their contribution.

Figures

Figures reproduced from arXiv: 2607.21044 by Daniel Ambach, Sabine Wildemann.

Figure 1
Figure 1. Figure 1: The 2 × 2 factorial design with stimulus assignment and the resulting risk pattern across the four conditions 6 [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The nine-phase study protocol, from eligibility screening and informed consent through the chat interaction [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Overview of the system architecture and the data flow between the frontend, backend, ballot-data service, and [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: The Data Nutrition Label displayed in conditions B and D [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: The granular consent dashboard displayed in conditions C and D [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Sample demographics: age, gender, primary language, education level, and voting eligibility ( [PITH_FULL_IMAGE:figures/full_fig_p013_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Data donation rates by condition with 95% confidence intervals. The dashed line marks the overall rate of [PITH_FULL_IMAGE:figures/full_fig_p014_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Manipulation check results for perceived transparency and perceived control by factor level. Error bars [PITH_FULL_IMAGE:figures/full_fig_p016_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Frequencies of the open-ended response themes ( [PITH_FULL_IMAGE:figures/full_fig_p017_9.png] view at source ↗

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

43 extracted references

  1. [1]

    Phi-4 technical report , year =

    Abdin, Marah and Aneja, Jyoti and Behl, Harkirat and Bubeck, S. Phi-4 technical report , year =

  2. [2]

    Science , year =

    Acquisti, Alessandro and Brandimarte, Laura and Loewenstein, George , title =. Science , year =

  3. [3]

    IEEE Security & Privacy , year =

    Acquisti, Alessandro and Grossklags, Jens , title =. IEEE Security & Privacy , year =

  4. [4]

    and Adaji, Ifeoma and Ituma, Chidinma and Ricciardelli, Rosemary , title =

    Andreotta, Adam J. and Adaji, Ifeoma and Ituma, Chidinma and Ricciardelli, Rosemary , title =. Ethics and Information Technology , year =

  5. [5]

    2025 , publisher =

    Codebuch f. 2025 , publisher =

    2025

  6. [6]

    and Busby, Ethan C

    Argyle, Lisa P. and Busby, Ethan C. and Fulda, Nancy and Gubler, Joshua R. and Rytting, Christopher and Wingate, David , title =. Political Analysis , year =

  7. [7]

    , title =

    Baum, Howell S. , title =. International Encyclopedia of the Social & Behavioral Sciences , edition =. 2001 , doi =

    2001

  8. [8]

    Social Psychological and Personality Science , year =

    Brandimarte, Laura and Acquisti, Alessandro and Loewenstein, George , title =. Social Psychological and Personality Science , year =

  9. [9]

    Journal of Personality and Social Psychology , year =

    Chaiken, Shelly , title =. Journal of Personality and Social Psychology , year =

  10. [10]

    2022 , howpublished =

    Chmielinski, Kasia and Newman, Sarah and Taylor, Matt and Joseph, Josh and Thomas, Kemi and Yurkofsky, Jessica and Holland, Sarah , title =. 2022 , howpublished =

    2022

  11. [11]

    and Armstrong, Pamela K

    Culnan, Mary J. and Armstrong, Pamela K. , title =. Organization Science , year =

  12. [12]

    2021 , howpublished =

    Delgado, Fernando and Yang, Stephen and Madaio, Michael and Yang, Qian , title =. 2021 , howpublished =

    2021

  13. [13]

    Information Systems Research , year =

    Dinev, Tamara and Hart, Paul , title =. Information Systems Research , year =

  14. [14]

    Regulation (EU) 2016/679 (General Data Protection Regulation) , year =

    2016

  15. [15]

    2023 , howpublished =

    Gao, Yunfan and Xiong, Yun and Gao, Xinyu and Jia, Kangxiang and Pan, Jinliu and Bi, Yuxi and Dai, Yi and Sun, Jiawei and Wang, Meng and Wang, Haofen , title =. 2023 , howpublished =

    2023

  16. [16]

    Apertus: Democratizing open and compliant

    Hern. Apertus: Democratizing open and compliant. 2025 , howpublished =

    2025

  17. [17]

    and Lutz, Christoph , title =

    Hoffmann, Christian P. and Lutz, Christoph , title =. European Journal of Communication , year =

  18. [18]

    2018 , howpublished =

    Holland, Sarah and Hosny, Ahmed and Newman, Sarah and Joseph, Joshua and Chmielinski, Kasia , title =. 2018 , howpublished =

    2018

  19. [19]

    Jeffreys, Harold , title =

  20. [20]

    Nature Machine Intelligence , year =

    Jobin, Anna and Ienca, Marcello and Vayena, Effy , title =. Nature Machine Intelligence , year =

  21. [21]

    and Lund, David and Morrison, Michael and Teare, Harriet and Melham, Karen , title =

    Kaye, Jane and Whitley, Edgar A. and Lund, David and Morrison, Michael and Teare, Harriet and Melham, Karen , title =. European Journal of Human Genetics , year =

  22. [22]

    Information Systems Journal , year =

    Kehr, Flavius and Kowatsch, Tobias and Wentzel, Daniel and Fleisch, Elgar , title =. Information Systems Journal , year =

  23. [23]

    , title =

    Lee, Min Kyung and Kusbit, Daniel and Kahng, Anson and Kim, Ji Tae and Yuan, Xinran and Chan, Allissa and See, Daniel and Noothigattu, Ritesh and Lee, Siheon and Psomas, Alexandros and Procaccia, Ariel D. , title =. Proceedings of the ACM on Human-Computer Interaction , year =

  24. [24]

    , title =

    Leonard, Peter G. , title =. 2018 , howpublished =

    2018

  25. [25]

    Ethics in Design and Communication: Critical Perspectives , edition =

    Madaio, Michael and Martin, Shari , title =. Ethics in Design and Communication: Critical Perspectives , edition =

  26. [26]

    and Cranor, Lorrie F

    McDonald, Aleecia M. and Cranor, Lorrie F. , title =. I/S: A Journal of Law and Policy for the Information Society , year =

  27. [27]

    2025 , howpublished =

    Murray-Rust, Dave and Alfrink, Kars and Zaga, Cristina , title =. 2025 , howpublished =

    2025

  28. [28]

    MixtureVitae: Open web-scale pre-training dataset with high quality instruction and reasoning data built from permissive-first text sources , year =

    Nguyen, Huu and May, Victor and Raj, Harsh and Nezhurina, Marianna and Wang, Yishan and Luo, Yanqi and Vu, Minh Chien and Nakamura, Taishi and Tsui, Ken and Nguyen, Van Khue and Salinas, David and Krasnod. MixtureVitae: Open web-scale pre-training dataset with high quality instruction and reasoning data built from permissive-first text sources , year =

  29. [29]

    and Horne, Daniel R

    Norberg, Patricia A. and Horne, Daniel R. and Horne, David A. , title =. Journal of Consumer Affairs , year =

  30. [30]

    Governing with artificial intelligence: The state of play and way forward in core government functions , year =

  31. [31]

    Ouyang, Long and Wu, Jeffrey and Jiang, Xu and Almeida, Diogo and Wainwright, Carroll L. and Mishkin, Pamela and Zhang, Chong and Agarwal, Sandhini and Slama, Katarina and Ray, Alex and Schulman, John and Hilton, Jacob and Kelton, Fraser and Miller, Luke and Simens, Maddie and Askell, Amanda and Welinder, Peter and Christiano, Paul and Leike, Jan and Lowe...

  32. [32]

    2025 , howpublished =

    Overney, Cassandra , title =. 2025 , howpublished =

    2025

  33. [33]

    , title =

    Peck, Joann and Shu, Suzanne B. , title =. Journal of Consumer Research , year =

  34. [34]

    and Kostova, Tatiana and Dirks, Kurt T

    Pierce, Jon L. and Kostova, Tatiana and Dirks, Kurt T. , title =. Review of General Psychology , year =

  35. [35]

    Shklovski, Irina and Mainwaring, Scott D. and Sk. Leakiness and creepiness in app economies: Investigating experiences of surveillance and control in mobile checking-in , journal =. 2014 , pages =

    2014

  36. [36]

    International Journal of Human-Computer Interaction , year =

    Shneiderman, Ben , title =. International Journal of Human-Computer Interaction , year =

  37. [37]

    Environment and Planning B: Urban Analytics and City Science , year =

    Sieber, Renee and Brandusescu, Ana and Sangiambut, Suthee and Adu-Daako, Anthony , title =. Environment and Planning B: Urban Analytics and City Science , year =

  38. [38]

    PLOS ONE , year =

    Skatova, Anya and Goulding, James , title =. PLOS ONE , year =

  39. [39]

    IEEE Data Engineering Bulletin , year =

    Stoyanovich, Julia and Howe, Bill , title =. IEEE Data Engineering Bulletin , year =

  40. [40]

    and Kim, Jang Hyun , title =

    Sun, Seungjong and Lee, Eungu and Nan, Dongyan and Zhao, Xiangying and Lee, Wonbyung and Jansen, Bernard J. and Kim, Jang Hyun , title =. 2024 , howpublished =

    2024

  41. [41]

    Federal Act on Data Protection (FADP) of 25 September 2020, as revised 1 September 2023 (SR 235.1) , year =

    2020

  42. [42]

    , title =

    Thomson, Ian and Boutilier, Robert G. , title =. SME Mining Engineering Handbook , edition =

  43. [43]

    Proceedings of the 11th International Conference on Communities and Technologies (C&T '23) , year =

    Drobotowicz, Karolina and Truong, Nghiep Lucy and Ylipulli, Johanna and Gonzalez Torres, Ana Paula and Sawhney, Nitin , title =. Proceedings of the 11th International Conference on Communities and Technologies (C&T '23) , year =