Pith. sign in

REVIEW 4 major objections 5 minor 26 references

Voice CMS: updating the knowledge base of a digital assistant through conversation

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper aims to show that a voice interface can maintain a digital assistant's knowledge base as well as a graphical CMS does on easy tasks, while a hybrid voice-plus-visual system is the natural next step for harder ones.

desk verdict A small but honest study: the preference result (voice wins on easy tasks, GUI on hard ones) is believable; the untested manual quality scoring makes the 'comparable quality' claim the paper's weak spot. read the letter →

arxiv 2505.22303 v1 pith:NWQPCHAE submitted 2025-05-28 cs.HC cs.AIcs.MA

Grzegorz Wolny , Micha{l} Szczerbak This is my paper
classification cs.HCcs.AIcs.MA
keywords voiceuserinterfaceknowledgemanagementdigitalassistantcontentsystemtaskcomplexitypreferencemulti-agentLLMarchitecturecomparativeusabilitystudy
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

The paper tries to establish that a voice-based content management system can be a practical way for hotel staff to update a digital assistant's knowledge base, competing with a traditional graphical CMS. Working from a multi-agent LLM pipeline, the authors built a Voice CMS and compared it with a GUI across nine tasks of three complexity levels with seven users. They report that the voice interface wins user preference on easy tasks, while users shift to the GUI for medium and complex tasks, and that both interfaces produce content of comparable quality. The authors argue that these results point toward a hybrid voice-plus-visual interface as the next step.

What carries the argument

The load-bearing mechanism is a multi-agent LLM architecture in which named workflow nodes carry the update dialogue: the assistant summarizes incoming information, asks for clarification when details are missing, builds the entry in the background with validity period and category, and only approves it after a full vocal summary is confirmed. The evaluation's measuring instruments are the Single Ease Question for per-task difficulty, the System Usability Scale for overall usability, and a 5-point manual quality rating of the resulting content; inference relies on Bayesian multilevel categorical logistic regression and generalized linear mixed models to account for repeated measures across the seven participants.

What would settle it

Re-score all 126 task outputs from the same transcripts with two or more independent raters using a pre-registered rubric, and look at inter-rater agreement and the mean difference between interfaces; if agreement is poor or the voice mean falls clearly below the GUI mean, the paper's parity claim is not supported.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central finding is that user preference between a Voice CMS and a graphical CMS is not fixed but depends on task complexity, and that this dependence runs opposite to the authors' initial hypothesis. For easy tasks the voice interface is strongly preferred, with the GUI never chosen at that level, whereas for medium and complex tasks the GUI becomes significantly more likely to be preferred. At the same time, objective content quality is nearly equal (GUI mean 4.54 vs Voice CMS mean 4.40 on a 5-point scale), and the GUI needs a clear perceived-ease advantage of about one point on the SEQ scale or about 57 seconds of time savings before it overtakes the voice interface in predicted preference. The paper presents this as evidence that conversational knowledge management is viable for simpler updates and that a hybrid interface should combine voice input with visual feedback.

Load-bearing premise

The paper's claim that voice-entered content is as good as GUI-entered content rests on the authors' own manual 5-point quality ratings, made after the fact without a published scoring rubric or a check that another rater would give the same scores; if those ratings are inconsistent, the quality-parity result collapses.

Editorial extensions

If this is right

  • Voice entry can be offered as the default for short, low-detail updates such as reporting an outage or a same-day event, since users prefer it there and content quality holds.
  • For medium and complex updates, a GUI-style form remains the safer default; forcing voice first for such tasks would run against measured preference.
  • A hybrid interface, voice capture with synchronized visual verification, is the paper's proposed remedy for the main weakness of voice: users' difficulty checking and correcting details.
  • Usability teams should treat a small ease-of-use gap in favor of the GUI as insufficient to flip choice; the voice interface carries a baseline preference advantage worth roughly one SEQ point or about a minute of time.
  • The quality-parity result means voice does not have to be a quality compromise, only a preference compromise, at least in this business setting.

Reading between the lines

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

  • The measured thresholds (about 1 SEQ point, about 57 seconds) are probably specific to this task set and user group; a useful next experiment would vary task framing and input modality to see whether the same crossover points recur.
  • Because the tasks were presented as text, participants may have mentally translated them into form-filling schemas; presenting the same scenarios verbally or through a real hotel workflow could narrow or widen the voice advantage.
  • Quality parity rests on a single manual scorer; an independent, rubric-based rating of the same voice and GUI outputs would be the most direct check of whether voice truly preserves quality.
  • The authors' suggested hybrid could be tested directly by adding a live visual transcript and edit controls to the Voice CMS and measuring whether preference for complex tasks shifts back toward voice.
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.

Referee Report

4 major / 5 minor

Summary. The paper presents Voice CMS, a multi-agent LLM-based voice user interface (VUI) for updating the knowledge base of a hotel-lobby digital assistant, and compares it to a conventional graphical CMS (GUI) in a within-subjects experiment with 7 participants and 9 tasks per interface. The reported findings are that the GUI receives higher overall usability (SUS), perceived ease (SEQ), and faster completion times, but that the Voice CMS is preferred for easy tasks, while user preference shifts significantly toward the GUI for medium and complex tasks, as inferred from a Bayesian multilevel categorical model. The authors also report that content quality is comparable between interfaces (GUI M=4.54, Voice CMS M=4.40 on a 5-point scale) and use qualitative comments to motivate a hybrid voice-visual interface. The paper concludes that voice interfaces are promising for simple knowledge-entry tasks but need visual feedback to handle complex content.

Significance. If the reported results hold, the paper makes a useful empirical contribution to the human-AI interaction literature: it demonstrates that a voice-only content management interface can be preferred for simple data entry, and it provides a concrete system architecture (multi-agent LangGraph with a secure voice mode) that others can build on. The study's honesty in reporting limitations and its detailed interaction metrics (message counts, summary usage) are strengths. However, the significance is materially limited by the small sample (N=7), the reliance on a single unblinded manual quality rating, and the instability of the preference model estimates, so the paper cannot yet support broad claims about when voice CMS systems are viable.

major comments (4)
  1. [Section 4 and Section 5.1] The central claim that content quality via the Voice CMS is 'comparable' to the GUI rests entirely on a single manual 5-point rating performed by the authors in post-processing, with no published rubric, no blinding to interface, no inter-rater reliability check, and no statistical test of equivalence or difference. The paper refers to 'predefined rules' (Section 4) but does not provide them. Since the authors built both interfaces, this is a serious validity threat. Please provide the rating rubric, have at least two independent raters who are blind to interface condition score the 126 outputs, report inter-rater agreement (e.g., Cohen's kappa), and compare the GUI and Voice CMS scores with a paired test or equivalence bound. Without this, the quality-parity claim, which appears prominently in the abstract, is not supported.
  2. [Section 5.2] The Bayesian multilevel model estimating preference yields extreme fixed-effect estimates with very wide credible intervals (e.g., Intercept Log-Odds = -11.92, 95% CrI [-36.27, -3.26]; Medium vs Easy Log-Odds = 11.21, 95% CrI [2.62, 35.58]; Complex vs Easy Log-Odds = 13.69, 95% CrI [4.90, 38.23]). These magnitudes indicate quasi-complete separation, which is unsurprising with 63 observations from 7 participants. The paper's wording that this provides 'strong statistical support' is overstated. Please report prior sensitivity analyses (e.g., a weakly informative regularizing prior on the fixed effects), or reframe these results as exploratory evidence and show descriptive preference counts by participant and difficulty. The current presentation may mislead readers about the strength of the preference shift.
  3. [Section 4 and Section 5.5] All participants performed the same nine tasks in the same fixed order (Section 4: 'Participants completed the same nine tasks in a fixed order ([M, M, E, C, C, E, C, M, E])'; Section 5.5: 'The tasks were executed in a random (but fixed order)'). Because the order is identical across participants, task difficulty is confounded with task position, so learning, fatigue, and interface-adaptation effects cannot be separated from the effect of difficulty on preference. A warm-up task reduces but does not eliminate this confound. Please either report the actual task order and analyze order as a covariate, or discuss explicitly why this confound does not undermine the central preference result. At minimum, the limitations section should acknowledge this as an unresolved alternative explanation.
  4. [Section 6] The paper acknowledges the small sample size (N=7) in its limitations, but several claims in Sections 5.3 and 5.4 (e.g., the 'crossover point' at a SEQ difference >1 and at a time difference of approximately -57 seconds) are presented with a precision that the data cannot support. With 7 participants, these crossover estimates have very wide uncertainty, and the paper does not report confidence intervals for these derived thresholds. Please either supply uncertainty bounds for these crossover values or soften the conclusions to emphasize that they are tentative observations from a small exploratory study.
minor comments (5)
  1. [Section 1] The phrase 'research augmented generation (RAG)' should be 'retrieval-augmented generation (RAG)'.
  2. [Section 5.3] There is a typo: 'GUI was less preffered at seq_diff≈0' should be 'preferred'.
  3. [Section 5.5] The sentence 'The plot in Figure 8 displays them grouped by difficulty but without changing relative order' is confusing because Figure 8 is titled 'Trend of Mean Summary Count over Tasks' and does not clearly relate to the message count discussion. Also, the parenthetical '(plot not provided)' suggests incomplete material. Please clarify which plot is referenced and ensure all figures cited are actually included.
  4. [Section 2] The literature review would benefit from more precise citation details for some works (e.g., page ranges or DOI for the preprint references) to help readers locate the sources.
  5. [Section 5.1] The descriptive statistics would be easier to interpret if the paper reported confidence intervals for the mean SUS, SEQ, and quality scores, or at least the full distributions, given the small N.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's claims are empirical measurements and fitted statistical models, not derivations that reduce to their own inputs.

full rationale

This paper is an empirical comparative usability study, not a derivation chain. The central claims—SUS scores, SEQ ratings, preference probabilities from Bayesian multilevel categorical logistic regression, GLMMs for SEQ difference and time difference, and content-quality scores—are all obtained by measuring participants' behavior and fitting models to those measurements. None of the reported 'findings' is defined in terms of another reported finding in a way that would make the conclusion tautological. The quality-parity claim rests on a manual 5-point post-processing rating described in Section 4, but that is a measurement-validity concern (unblinded scoring, no inter-rater reliability), not a circularity concern: the scores are not constructed from the conclusion they support. The only self-citation, Lesiak et al. [2025], is used to describe the previously built videobot context and does not carry the paper's empirical argument. There is no invoked uniqueness theorem, no fitted parameter renamed as a prediction, and no ansatz smuggled in via citation. Models are fit to the data as standard practice, and the paper's own limitations section appropriately flags small sample size, task-presentation issues, and technical prototype limitations. Therefore, no specific circular step can be quoted, and the appropriate finding is no significant circularity.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The empirical claims rest on hand-made difficulty categories, a manual quality scale, and assumptions about task presentation and participant representativeness. No new physical or conceptual entities are introduced.

free parameters (2)
  • Task difficulty levels = Easy/Medium/Complex
    Authors assigned nine tasks to three difficulty categories by hand based on length and detail; this grouping drives all difficulty-stratified analyses.
  • Quality score rubric = 1-5 manual scale
    Post-hoc rating rules for content quality were defined by the authors; no inter-rater reliability is reported, making the scale an unvalidated hand-set criterion.
assumptions (5)
  • standard math Bayesian multilevel categorical logistic regression and GLMMs are valid for these data
    Standard statistical tools; the paper uses them appropriately for repeated-measures data.
  • domain assumption Textual task presentation did not systematically bias participants toward GUI form-filling
    The authors acknowledge that presenting tasks as text may have favored GUI interaction patterns (Section 6).
  • domain assumption The 7 participants are sufficiently representative of potential hotel staff users
    Small convenience sample, ages 33-51, all from the researchers' environment; generalizability is limited (Section 6).
  • standard math SUS and SEQ are appropriate usability measures
    Validated instruments cited by the authors (Brooke 1996; Sauro and Dumas 2009).
  • domain assumption The VUI and GUI prototypes are comparable implementations of the same functionality
    The VUI uses LLM/ASR and suffers from ASR/NLU errors, while the GUI is a conventional form; this asymmetry likely affected usability scores (Section 6).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Voice CMS: updating the knowledge base of a digital assistant through conversation." pith.science (2026). https://pith.science/paper/NWQPCHAE

@misc{pith2026250522303,
  author       = {Pith},
  title        = {Pith review of: Voice CMS: updating the knowledge base of a digital assistant through conversation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NWQPCHAE}},
  note         = {Machine review of arXiv:2505.22303}
}
read the original abstract

In this study, we propose a solution based on a multi-agent LLM architecture and a voice user interface (VUI) designed to update the knowledge base of a digital assistant. Its usability is evaluated in comparison to a more traditional graphical content management system (CMS), with a focus on understanding the relationship between user preferences and the complexity of the information being provided. The findings demonstrate that, while the overall usability of the VUI is rated lower than the graphical interface, it is already preferred by users for less complex tasks. Furthermore, the quality of content entered through the VUI is comparable to that achieved with the graphical interface, even for highly complex tasks. Obtained qualitative results suggest that a hybrid interface combining the strengths of both approaches could address the key challenges identified during the experiment, such as reducing cognitive load through graphical feedback while maintaining the intuitive nature of voice-based interactions. This work highlights the potential of conversational interfaces as a viable and effective method for knowledge management in specific business contexts.

Figures

Figures reproduced from arXiv: 2505.22303 by the authors.

Figure 1
Figure 1. Digital assistant The solution presented in this study builds upon a digital assistant equipped with a voice user interface (VUI). The assistant, implemented as a 3D-rendered digital character displayed on a large touchscreen in a hotel lobby (see 3 [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Voice CMS architecture 5 [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Preference Distribution by Task Difficulty [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: SEQ Difference (GUI - Voice CMS) by Preference [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Predicted Probability of Preference vs SEQ Difference [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Processing Time Difference (GUI - Voice CMS) by Preference [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: Predicted Probability of Preference vs Processing Time Difference [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]
Figure 8
Figure 8. Figure 8: Trend of Mean Summary Count over Tasks 5.5 Voice CMS Interaction Metrics This subsection details the analysis of user interaction patterns specifically with the Voice CMS interface, focusing on the number of messages exchanged between experiment participants and the di…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

26 extracted references · 19 canonical work pages

  1. [1]

    Exploring chatgpt capabilities and limitations: A survey

    Anis Kouba, Wadii Boulila, Lahouari Ghouti, Ayyub Alzahem, and Shahid Latif. Exploring chatgpt capabilities and limitations: A survey. IEEE Access, 11 0 (194): 0 118698--118721, 2023

    work page 2023

  2. [2]

    Understanding users' dissatisfaction with chatgpt responses: Types, resolving tactics, and the effect of knowledge level

    Yoonsu Kim, Jueon Lee, Seoyoung Kim, Jaehyuk Park, and Yuho Kim. Understanding users' dissatisfaction with chatgpt responses: Types, resolving tactics, and the effect of knowledge level. In Proceedings of the 29th International Conference on Intelligent User Interfaces, pages 385--404, 2024. doi:10.1145/3640543.3645148. URL https://dl.acm.org/doi/pdf/10.1...

    arXiv 2024

  3. [3]

    Evaluating the impact of hallucinations on user trust and satisfaction in llm-based systems

    Richard Oelschlager. Evaluating the impact of hallucinations on user trust and satisfaction in llm-based systems. Master's thesis, Linnaeus University, Sweden, 2024

    work page 2024

  4. [4]

    S.M Towhidul Islam Tonmoy, S. M. Mehedi Zaman, Vinija Jain, Anku Rani, Vipula Rawte, Aman Chadha, and Amitava Das. A comprehensive survey of hallucination mitigation techniques in large language models, 2024

    work page 2024

  5. [5]

    Szczerbak

    Emilia Lesiak, Grzegorz Wolny, Bartosz Przybył, and Michał K. Szczerbak. Digital assistant in a point of sales. In Proceedings of the 20th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applicationss - Volume 1: GRAPP, HUCAPP and IVAPP, HUCAPP '25, pages 439--451, 2025. doi:10.5220/0013042500003912. URL https:/...

    work page doi:10.5220/0013042500003912 2025

  6. [6]

    Knowledge triggering, extraction and storage via human–robot verbal interaction, 2021

    Lucrezia Grassi, Carmine Tommaso Recchiuto, and Antonio Sgorbissa. Knowledge triggering, extraction and storage via human–robot verbal interaction, 2021

    work page 2021

  7. [7]

    Olivares-Rojas, J

    Juan C. Olivares-Rojas, J. Gabriel González-Serna, J. Guadalupe Ramos-Díaz, Noe A. Castro-Sánchez, and Johan W. González-Murueta. From gui to vui: A natural language approach to multimodal medical system. Computación y Sistemas, 29 0 (1): 0 295--309, 2025

    work page 2025

  8. [8]

    It's good to talk: A comparison of using voice versus screen-based interactions for agent-assisted tasks

    Leon Reicherts, Yvonne Rogers, Licia Capra, Ethan Wood, Tu Dinh Duang, and Neil Sebire. It's good to talk: A comparison of using voice versus screen-based interactions for agent-assisted tasks. ACM Transactions on Computer-Human Interaction, 29 0 (3): 0 1--41, 2022

    work page 2022

Show all 26 references
  1. [9]

    Apples and oranges: A framework for the usability evaluation of voice vs graphical user interfaces: A command-event based method proposal

    Fermin Chavez-Sanchez and Lucila Mercado Colin. Apples and oranges: A framework for the usability evaluation of voice vs graphical user interfaces: A command-event based method proposal. In Proceedings of the 2nd Conference on Conversational User Interfaces, pages 1--3, 2020. ...

    2020

  2. [10]

    A comparison of speech and gui input for navigation in complex visualizations on mobile devices

    Rui Zhang, Stephen North, and Eleftherios Koutsofios. A comparison of speech and gui input for navigation in complex visualizations on mobile devices. In Proceedings of the 12th international conference on Human computer interaction with mobile devices and services, pages 357-...

    2010

  3. [11]

    A comparative study of voice and graphical user interfaces with respect to literacy levels

    Priyanka Chandel, Devanuj Kanta Balkrishan, and Pankaj Doke. A comparative study of voice and graphical user interfaces with respect to literacy levels. In Proceedings of the 3rd ACM Symposium on Computing for Development, pages 1--2, 2013. doi:10.1145/2442882.2442921. URL htt...

    2013

  4. [12]

    The msiia experiment: Using speech to enhance human performance on a cognitive task

    Laurie Damianos, Dan Loehr, Carl Burke, Steve Hansen, and Michael Viszmeg. The msiia experiment: Using speech to enhance human performance on a cognitive task. International Journal of Speech Technology, 6 0 (2): 0 133--144, 2003

    2003

  5. [13]

    Context matters: Understanding the effect of usage contexts on users’ modality selection in multimodal systems

    Min Chul Cha and Yong Gu Ji. Context matters: Understanding the effect of usage contexts on users’ modality selection in multimodal systems. International Journal of Human–Computer Interaction, 40 0 (20): 0 6287--6302, 2024

    2024

  6. [14]

    Moore, and David Coyle

    Hannah Limerick, James W. Moore, and David Coyle. Empirical evidence for a diminished sense of agency in speech interfaces. In Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems, pages 3967--3970, 2015. doi:10.1145/2702123.2702379. URL https://...

    2015

  7. [15]

    Effect of modality on collaboration with a dialogue system

    Ludovic Le Bigot, Patrice Terrier, Virginie Amiel, Gérard Poulain, Eric Jamet, and Jean-François Rouet. Effect of modality on collaboration with a dialogue system. International Journal of Human-Computer Studies, 65 0 (12): 0 983--991, 2007

    2007

  8. [16]

    The role of spoken feedback in experiencing multimodal interfaces as human-like

    Pernilla Qvarfordt, Arne Jönsson, and Nils Dahlbäck. The role of spoken feedback in experiencing multimodal interfaces as human-like. In Proceedings of the 5th international conference on Multimodal interfaces, pages 250--257, 2003. doi:10.1145/958432.958478. URL https://dl.ac...

    2003

  9. [17]

    User experience and usability of voice user interfaces: A systematic literature review

    Akshay Madhav Deshmukh and Ricardo Chalmeta. User experience and usability of voice user interfaces: A systematic literature review. Information, 15 0 (9): 0 579, 2024

    2024

  10. [18]

    Faruk Lawal Ibrahim Dutsinma, Debajyoti Pal, Suree Funilkul, and Jonathan H. Chan. A systematic review of voice assistant usability: An iso 9241–11 approach. Computer Science, 3 0 (4): 0 267, 2022

    2022

  11. [19]

    User experience of digital voice assistant: Conceptualization and measurement

    Qian Chen and Yeming Gong. User experience of digital voice assistant: Conceptualization and measurement. ACM Transactions on Computer-Human Interaction, 31 0 (1): 0 1--35, 2024

    2024

  12. [20]

    Improving the usability of voice user interfaces: A new set of ergonomic criteria

    Caroline Nowacki, Anna Gordeeva, and Anne-Hélène Lizé. Improving the usability of voice user interfaces: A new set of ergonomic criteria. In Proceedings of the 9th International Conference: In Design, User Experience, and Usability. Design for Contemporary Interactive Environm...

    2020

  13. [21]

    What’s the talk on vui guidelines? a meta-analysis of guidelines for voice user interface design

    Christine Murad, Heloisa Candello, and Cosmin Munteanu. What’s the talk on vui guidelines? a meta-analysis of guidelines for voice user interface design. In Proceedings of the 5th International Conference on Conversational User Interfaces, pages 1--16, 2023. doi:10.1145/357188...

    2023

  14. [22]

    Cognitive load and productivity implications in human-chatbot interaction

    Johanna Schmidhuber, Stephan Schlögl, and Christian Ploder. Cognitive load and productivity implications in human-chatbot interaction. In 2021 IEEE 2nd International Conference on Human-Machine Systems (ICHMS), pages 1--6, 2021. doi:10.1109/ICHMS53169.2021.9582445

    2021

  15. [23]

    Jeff Sauro and Joseph s. Dumas. Comparison of three one-question, post-task usability questionnaires. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, CHI '09, pages 1599--1608. ACM, 2009. doi:10.1145/1518701.1518946. URL https://dl.acm.org/doi/10...

    2009

  16. [24]

    Sus-a quick and dirty usability scale

    John Brooke. Sus-a quick and dirty usability scale. Usability evaluation in industry, 189 0 (194): 0 4--7, 1996

    1996

  17. [25]

    Determining what individual sus scores mean: adding an adjective rating scale

    Aaron Bangor, Philip Kortum, and James Miller. Determining what individual sus scores mean: adding an adjective rating scale. J. Usability Studies, 4 0 (3): 0 114–123, May 2009

    2009

  18. [26]

    Jeff Sauro and James R. Lewis. Quantifying the User Experience: Practical Statistics for User Research. Morgan Kaufmann Publishers Inc., San Francisco, CA, USA, 1st edition, 2012. ISBN 9780123849687

    2012

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.