Pith. sign in

REVIEW 3 major objections 3 minor 40 references

Agency Among Agents: Designing with Hypertextual Friction in the Algorithmic Web

T0 review · 3 major / 3 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read The paper argues that reintroducing hypertextual friction—deliberate pauses, visible provenance, and editable structure—into algorithm-driven interfaces can restore user agency.

desk verdict A coherent Blue Skies synthesis that names 'Hypertextual Friction' as a design stance; the central claim is untested but honestly scoped. read the letter →

arxiv 2507.23585 v1 pith:FMTHUTMS submitted 2025-07-31 cs.HC cs.AIcs.MMcs.SI

classification cs.HCcs.AIcs.MMcs.SI
keywords useragencyhypertextgenerativeAIalgorithmicinterfacesinterfacefrictiontraceabilitystructurecreativetools
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 argues that the web's shift from hypertextual trails to algorithmic feeds and generative tools has eroded user agency, and it proposes a design stance called Hypertextual Friction. The stance treats friction, traceability, and structure as interface-level values that can be deliberately designed into hybrid systems. Through comparisons of Wikipedia vs. Instagram Explore and Are.na vs. DALL·E, the authors claim hypertextual systems support provenance, associative thinking, and user-driven meaning-making while algorithmic systems tend to hide process and flatten participation. If the stance is adopted, designers would have a concrete vocabulary for building interfaces that let users navigate, trace, and compose meaning rather than merely receive outputs.

What carries the argument

The central object is Hypertextual Friction, a named design stance with three interface-level values: friction (deliberate pauses, forks, and decision points), traceability (visible sources, trails, and remix histories), and structure (nonlinear, editable scaffolds for composing ideas). The stance is carried by the comparative table contrasting hypertextual (user-driven) and algorithmic (agent-driven) systems on these three axes; the table is what makes the abstract values concrete and actionable for designers. It functions as the paper's contribution: a vocabulary and set of commitments that designers can use to embed agency into hybrid, automation-shaped interfaces.

What would settle it

A controlled study would settle it: take a generative image or feed interface, add visible source trails and decision forks, and compare perceived agency and task satisfaction against the seamless version. If users rate the frictive version lower on agency or abandon it as burdensome, the central claim fails.

Watch

Extended reading notes

Core claim

The central claim is that agency in algorithmically mediated environments can be reclaimed by redesigning the interface itself, not just by explaining algorithms. The paper introduces Hypertextual Friction as a conceptual design stance centered on three values: friction (slowing users down so intentions surface), traceability (making sources and sequencing visible), and structure (giving users scaffolds for composing and relating ideas). These values are drawn from classical hypertext—Bush's trails, Nelson's intertwingularity, and Amisola's claim that linking is authorship—and reoriented toward contemporary AI systems. The paper supports the stance with a comparative analysis: Wikipedia invites decision-laden associative navigation while Instagram Explore delivers an opaque personalized stream; Are.na supports traceable, emergent curation while DALL·E collapses prompt to output, flattening process. The authors argue that deliberately reintroducing hypertextual friction into hybrid systems would make users co-authors rather than subjects of computation.

Load-bearing premise

The argument assumes that users will experience deliberately added friction, traceability, and structure as enabling rather than annoying; the paper asserts this in Section 5 without testing it with users.

Editorial extensions

If this is right

  • Designers of recommendation feeds could add visible forks and source trails so users can choose paths instead of scrolling a single stream.
  • Generative image tools could expose prompt-to-output provenance and remix histories, letting users retrace and repurpose what they make.
  • Hybrid creative tools could embed generated outputs into editable canvases or semantic trails, so meaning evolves through interaction.
  • Agency becomes a property of interface structure rather than a byproduct of transparency or explanation alone.
  • The three values give a common language for comparing systems across domains, from information retrieval to creative tools.

Reading between the lines

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

  • The stance predicts a testable hypothesis: users interacting with a hybrid interface that adds forks and provenance will report higher perceived authorship or agency than users of a seamless baseline, even at some cost to efficiency.
  • The paper's own DALL·E walkthrough hints that hypertextual scaffolding can improve generative output quality as a side effect, which would offer a measurable proxy for agency that does not rely on self-report.
  • A natural extension is to retrofit existing platforms with friction layers, such as browser extensions that add link trails to feeds, and measure whether users adopt them; the paper stops at the stance rather than an implementation.
  • The stance could be applied to agentic assistants in high-stakes domains such as writing, research, and code, where preserving user authorship may matter more than in casual image generation.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 3 minor

Summary. The paper proposes a conceptual design stance called "Hypertextual Friction," which repositions classical hypertext values—friction, traceability, and structure—as actionable commitments for supporting user agency in algorithmically mediated environments. The authors ground the proposal in a comparative analysis of two interface pairings: Wikipedia versus Instagram Explore, and Are.na versus DALL·E. They argue that hypertextual systems foreground provenance, associative thinking, and user-driven meaning-making, while algorithmic systems obscure process and flatten participation. The paper contributes a comparative analysis and a design stance, explicitly positioning itself as a Blue Skies conceptual contribution rather than an empirical study.

Significance. If the central claim were fully supported, this paper would provide a useful vocabulary for designers of hybrid human-AI systems, connecting the hypertext community's historical commitments to contemporary generative AI and recommender systems. The paper is clearly written, situates itself well within prior work on slow technology, algorithmic experience, and reflective creation, and honestly acknowledges that future empirical work is needed. Its strength lies in its synthesis and the clarity of its three values; its weakness is that the evidence for the empirical premise—that deliberate friction, traceability, and structure are experienced by users as enabling agency rather than as obstacles—is anecdotal and indirect. For a short adjunct paper, the conceptual contribution is reasonable, but the stated claim of 'reclaiming agency' goes beyond what the presented evidence can support.

major comments (3)
  1. [Section 4.2] The only illustrative walkthrough in the paper (the DALL·E/Thievery Corporation example) does not actually instantiate Hypertextual Friction inside an algorithmic interface. The friction, traceability, and structure are supplied externally—via Are.na, Pinterest, Wikipedia, and Apple Freeform—while DALL·E itself remains unchanged. The walkthrough therefore shows that manual curation can improve prompt quality, but it cannot demonstrate that an interface designed with Hypertextual Friction enhances user agency. This gap directly affects the central claim stated in Section 5, and the paper would need either a concrete prototype or a careful reframing of the contribution as a speculative stance rather than a demonstrated intervention.
  2. [Section 4.1 and Table 1] The comparative analysis defines the two paradigms in a way that makes the conclusion partly definitional: hypertextual systems are characterized by friction, traceability, and structure, while algorithmic systems are characterized as seamless, opaque, and output-focused. Given this definition, observing that Wikipedia exhibits more traceability than Instagram Explore is an illustration of the taxonomy, not independent evidence for the causal claim that such properties produce agency. The paper would be strengthened by acknowledging that the comparison is analytic rather than empirical, and by deriving specific, testable predictions about user experience from the framework.
  3. [Section 5 and Section 6] The proposed instantiations in Section 5 ('visible forks in content,' 'source trails,' 'editable canvases') are promising, but they are not implemented, prototyped, or evaluated anywhere in the manuscript. Section 6 explicitly defers empirical testing to future work. While this is honest scope-setting, it means the paper's central assertion—that Hypertextual Friction can reclaim agency—is currently an untested hypothesis. To meet the standard implied by the title and abstract, the authors should either provide a low-fidelity design sketch of a hybrid system that embodies the three values, or soften the claim to present the stance as a proposal for design exploration rather than an established pathway to agency.
minor comments (3)
  1. [References [11] and [31]] Reference [11] is a musical album, and [31] is a web dossier from an unspecified organization; both are used as scholarly citations. If these are intended as cultural or artistic references, the text should say so explicitly, and if they are meant to support empirical claims, more conventional sources should be used.
  2. [Section 3.1] The paper defines 'algorithmic systems' narrowly to mean predictive, generative, or adaptive systems, while acknowledging in a footnote that all computational systems use algorithms. The main text might benefit from a brief statement that 'algorithmic' here is an interaction-design label, not a technical claim about the absence of algorithms in hypertext systems.
  3. [Figure 1 caption] The caption states 'Three approaches to visualizing the vibe…' but does not clarify that the left and right images are outputs from the same DALL·E model with different prompts. Adding the exact prompt used for the 'literal' version and the nature of the 'refined prompt' would help readers assess the authors' interpretation of the model's behavior.

Circularity Check

1 steps flagged · score 1.0 of 10

Minor self-definitional restatement in the comparative analysis; the design stance itself is an argued proposal, not an empirically derived prediction.

  1. self definitional [Section 3.2 (Interface Paradigms) vs. Abstract; echoed in Section 4.1]
    "We use the term "hypertextual systems" to describe an interaction paradigm that, in contrast, structures information through visible, intentionally authored links. ... They emphasize manual exploration and associative thinking. Users trace connections, follow authored trails, and construct meaning through linked structure. ... We show that hypertext systems emphasize provenance, associative thinking, and user-driven meaning-making, while algorithmic systems tend to obscure process and flatten participation."

    The abstract's claimed 'show' is a direct restatement of the paper's own stipulative definitions: Section 3.2 defines hypertextual systems as systems whose users 'trace connections... and construct meaning through linked structure,' which is exactly the 'user-driven meaning-making' the abstract claims to demonstrate. Section 3.1 likewise defines algorithmic systems as output-driven and often opaque, so 'obscure process' is built into the taxonomy rather than discovered. The comparative analysis therefore instantiates the definitions rather than providing independent evidence for them.

full rationale

The paper contains no equations, fitted parameters, or numerical predictions, so equation-level circularity is absent. The only self-citation, reference [3], supports the claim that DALL·E's process is opaque and flattens meaning; this is consistent with external critiques and is not load-bearing for the design stance. The Section 4.2 walkthrough is anecdotal and the authors themselves state in Section 6 that 'future empirical work should examine how these values shape user experience,' which is an honest scope limitation rather than a circular move. The one detectable circular trace is definitional: the comparative analysis restates the stipulated meanings of 'hypertextual' and 'algorithmic' systems from Sections 3.1–3.2, so the abstract's summary that hypertext systems emphasize provenance and user-driven meaning-making is entailed by those definitions. Because the central claim is a conceptual proposal rather than a measured outcome, this definitional restatement does not force the result; the normative argument for friction, traceability, and structure retains independent content. Overall, no significant circularity is present.

Assumptions & free parameters 0 free parameters · 4 assumptions · 1 invented entities

No free parameters are fitted. The paper relies on domain assumptions about algorithmic opacity and the value of friction, traceability, and structure, plus the representativeness of its chosen case systems. Hypertextual Friction is introduced as a new conceptual entity with no independent empirical evidence.

assumptions (4)
  • domain assumption Algorithmic systems obscure process and flatten participation.
    Stated in Section 3.1 and used throughout the comparison; not empirically measured here, only asserted.
  • domain assumption Friction, traceability, and structure are values that can restore user agency.
    This is the core commitment of the paper (Section 5); it is assumed rather than demonstrated with user studies.
  • domain assumption Hypertext is a productive lens for analyzing contemporary algorithmic interfaces.
    Section 2 relies on this framing from prior work; the paper extends it without formal justification.
  • domain assumption The four selected systems are representative of the hypertextual and algorithmic paradigms.
    Section 4 compares Wikipedia, Instagram, Are.na, and DALL·E as proxies; the paper acknowledges hybridity in a footnote but treats the dichotomy as valid.
invented entities (1)
  • Hypertextual Friction
    purpose: Conceptual design stance to guide interface design
    Newly named construct; it has no falsifiable handle outside the paper and its utility will only be shown through future design applications.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Agency Among Agents: Designing with Hypertextual Friction in the Algorithmic Web." pith.science (2026). https://pith.science/paper/FMTHUTMS

@misc{pith2026250723585,
  author       = {Pith},
  title        = {Pith review of: Agency Among Agents: Designing with Hypertextual Friction in the Algorithmic Web},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FMTHUTMS}},
  note         = {Machine review of arXiv:2507.23585}
}
read the original abstract

Today's algorithm-driven interfaces, from recommendation feeds to GenAI tools, often prioritize engagement and efficiency at the expense of user agency. As systems take on more decision-making, users have less control over what they see and how meaning or relationships between content are constructed. This paper introduces "Hypertextual Friction," a conceptual design stance that repositions classical hypertext principles--friction, traceability, and structure--as actionable values for reclaiming agency in algorithmically mediated environments. Through a comparative analysis of real-world interfaces--Wikipedia vs. Instagram Explore, and Are.na vs. GenAI image tools--we examine how different systems structure user experience, navigation, and authorship. We show that hypertext systems emphasize provenance, associative thinking, and user-driven meaning-making, while algorithmic systems tend to obscure process and flatten participation. We contribute: (1) a comparative analysis of how interface structures shape agency in user-driven versus agent-driven systems, and (2) a conceptual stance that offers hypertextual values as design commitments for reclaiming agency in an increasingly algorithmic web.

Figures

Figures reproduced from arXiv: 2507.23585 by the authors.

Figure 1
Figure 1. Three approaches to visualizing the "vibe" of Thievery Corporation’s "Lebanese Blonde." Left: A literal interpretation [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 30 canonical work pages

  1. [1]

    Wikipedia: About

    2025. Wikipedia: About. https://en.wikipedia.org/wiki/Wikipedia:About. Ac- cessed: 2025-06-13. Agency Among Agents HT Adjunct 2025, September 15–18, 2025, Chicago, IL, USA

  2. [2]

    Joseph Allemandou. 2018. A new platform to explore statistics about Wiki- media projects. https://wikimediafoundation.org/news/2018/01/02/wikistats-2/. Accessed: 2025-06-12

  3. [3]

    Zamfirescu-Pereira, Kyu Won Kim, Pradeep Mani Rathnam, and Bjoern Hartmann

    Shm Garanganao Almeda, J.D. Zamfirescu-Pereira, Kyu Won Kim, Pradeep Mani Rathnam, and Bjoern Hartmann. 2024. Prompting for Discovery: Flex- ible Sense-Making for AI Art-Making with Dreamsheets. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems (Honolulu, HI, USA) (CHI ’24). Association for Computing Machinery, New York, NY, ...

  4. [4]

    Oscar Alvarado and Annika Waern. 2018. Towards Algorithmic Experience: Initial Efforts for Social Media Contexts. In Proceedings of the 2018 CHI Confer- ence on Human Factors in Computing Systems (Montreal QC, Canada) (CHI ’18). Association for Computing Machinery, New York, NY, USA, 1–12. doi:10.1145/ 3173574.3173860

  5. [5]

    Chia Amisola. 2024. Becoming Hypertext. https://everythingi.love/#performance/ becominghypertext/. Accessed June 12, 2025

  6. [7]

    Apple Inc. 2022. Apple launches Freeform: a powerful new app designed for creative brainstorming and collaboration . https://www.apple.com/newsroom/ 2022/12/apple-launches-freeform-a-powerful-new-app-designed-for-creative- collaboration/ Accessed: 2025-06-13

  7. [8]

    Are.na. 2025. About. https://www.are.na/about. Accessed June 12, 2025

  8. [9]

    Claus Atzenbeck. 2024. Unwinding AI’s Moral Maze: Hypertext’s Ethical Poten- tial. In Proceedings of the 35th ACM Conference on Hypertext and Social Media (Poznan, Poland) (HT ’24). Association for Computing Machinery, New York, NY, USA, 23–28. doi:10.1145/3648188.3678213

Show all 40 references
  1. [10]

    Nürnberg

    Claus Atzenbeck and Peter J. Nürnberg. 2019. Hypertext as Method. InProceedings of the 30th ACM Conference on Hypertext and Social Media (Hof, Germany) (HT ’19). Association for Computing Machinery, New York, NY, USA, 29–38. doi:10. 1145/3342220.3343669

  2. [11]

    Valentina Berthelon and Tobias Freund. 2022. Hypertext. https://recentarts. bandcamp.com/album/hypertext. Audiovisual artwork released March 2022 on Reiten

  3. [12]

    Vannevar Bush. 1996. As we may think. Interactions 3, 2 (March 1996), 35–46. doi:10.1145/227181.227186

  4. [13]

    Yang Deng, An Zhang, Yankai Lin, Xu Chen, Ji-Rong Wen, and Tat-Seng Chua

  5. [15]

    Vera Liao, Samir Passi, Mark O

    Upol Ehsan, Q. Vera Liao, Samir Passi, Mark O. Riedl, and Hal Daumé. 2024. Seamful XAI: Operationalizing Seamful Design in Explainable AI. Proc. ACM Hum.-Comput. Interact. 8, CSCW1, Article 119 (April 2024), 29 pages. doi:10.1145/ 3637396

  6. [16]

    I always assumed that I wasn’t really that close to [her]

    Motahhare Eslami, Aimee Rickman, Kristen Vaccaro, Amirhossein Aleyasen, Andy Vuong, Karrie Karahalios, Kevin Hamilton, and Christian Sandvig. 2015. "I always assumed that I wasn’t really that close to [her]": Reasoning about Invisible Algorithms in News Feeds. In Proceedings o...

  7. [17]

    Aldana Afanador, and Carlos Fernández-Rovira

    Santiago Giraldo-Luque, Paula N. Aldana Afanador, and Carlos Fernández-Rovira

  8. [18]

    GitHub and OpenAI. 2021. GitHub Copilot: Your AI Pair Programmer. https: //github.com/features/copilot. Accessed: 2025-06-13

  9. [19]

    Satvik Golechha, Dylan Cope, and Nandi Schoots. 2024. Training Neural Net- works for Modularity aids Interpretability. arXiv:2409.15747 [cs.LG] https: //arxiv.org/abs/2409.15747

  10. [20]

    Lars Hallnäs and Johan Redström. 2001. Slow Technology – Designing for Reflection. Personal Ubiquitous Comput. 5, 3 (Jan. 2001), 201–212. doi:10.1007/ PL00000019

  11. [21]

    Anne Helmond. 2015. The Platformization of the Web: Making Web Data Platform Ready. Social Media + Society 1 (09 2015). doi:10.1177/2056305115603080

  12. [22]

    Eelco Herder and Jouke Staring. 2024. Digital Junkfood on Social Media: To Each Their Own Poison. In Proceedings of the 35th ACM Conference on Hypertext and Social Media (Poznan, Poland) (HT ’24). Association for Computing Machinery, New York, NY, USA, 126–135. doi:10.1145/364...

  13. [23]

    Mehdi Hosseinzadeh Aghdam. 2019. Context-aware recommender systems using hierarchical hidden Markov model. Physica A: Statistical Mechanics and its Applications 518 (2019), 89–98. doi:10.1016/j.physa.2018.11.037

  14. [24]

    Harmanpreet Kaur, Harsha Nori, Samuel Jenkins, Rich Caruana, Hanna Wallach, and Jennifer Wortman Vaughan. 2020. Interpreting Interpretability: Under- standing Data Scientists’ Use of Interpretability Tools for Machine Learning. In Proceedings of the 2020 CHI Conference on Huma...

  15. [25]

    Taiwo Kolajo, Olawande Daramola, Ayodele Adebiyi, and Aaditeshwar Seth. 2020. A framework for pre-processing of social media feeds based on integrated local knowledge base. Information Processing & Management 57, 6 (2020), 102348. doi:10.1016/j.ipm.2020.102348

  16. [26]

    Max Kreminski and Michael Mateas. 2021. Reflective Creators. In Proceedings of the 12th International Conference on Computational Creativity (ICCC) . Association for Computational Creativity, Virtual. https://computationalcreativity.net/iccc21/ wp-content/uploads/2021/09/ICCC_...

  17. [27]

    Gary Marchionini and Ben Shneiderman. 1988. Finding Facts vs. Browsing Knowledge in Hypertext Systems. Computer 21, 1 (Jan. 1988), 70–80. doi:10.1109/ 2.222119

  18. [28]

    Ted Nelson. 1974. Computer Lib/Dream Machines. Self-published, Chicago, IL

  19. [29]

    Obsidian MD. 2024. Obsidian. https://obsidian.md. A knowledge base that works on local Markdown files

  20. [30]

    William Odom, Richard Banks, Abigail Durrant, David Kirk, and James Pierce

  21. [31]

    Office of Applied Strategy. 2024. Hyper-Optimization. https:// officeofappliedstrategy.com/research#oas-dossier-1-hyper-optimization Ac- cessed February 2025

  22. [32]

    OpenAI. 2023. ChatGPT: Optimizing Language Models for Dialogue. https: //openai.com/blog/chatgpt. Accessed: 2025-06-13

  23. [33]

    Aditya Ramesh, Mikhail Pavlov, Gabriel Goh, Scott Gray, Chelsea Voss, Alec Rad- ford, Mark Chen, and Ilya Sutskever. 2021. Zero-Shot Text-to-Image Generation. CoRR abs/2102.12092 (2021). arXiv:2102.12092 https://arxiv.org/abs/2102.12092

  24. [34]

    Daniel Roßner, Claus Atzenbeck, and Sam Brooker. 2023. SPORE: A Storybreaking Machine. In Proceedings of the 34th ACM Conference on Hypertext and Social Media (Rome, Italy) (HT ’23) . Association for Computing Machinery, New York, NY, USA, Article 1, 6 pages. doi:10.1145/36031...

  25. [35]

    Debasis Roy and Madhurima Dutta. 2022. A Systematic Review and Research Perspective on Recommender Systems. Journal of Big Data 9, 1 (2022), 59. doi:10. 1186/s40537-022-00592-5

  26. [36]

    Cynthia Rudin. 2019. Stop Explaining Black Box Machine Learning Models for High Stakes Decisions and Use Interpretable Models Instead. Nature Machine Intelligence 1, 5 (May 2019), 206–215. doi:10.1038/s42256-019-0048-x

  27. [37]

    Aaron Springer and Steve Whittaker. 2020. Progressive Disclosure: When, Why, and How Do Users Want Algorithmic Transparency Information? ACM Trans. Interact. Intell. Syst. 10, 4, Article 29 (Oct. 2020), 32 pages. doi:10.1145/3374218

  28. [38]

    Weisz, Jessica He, Michael Muller, Gabriela Hoefer, Rachel Miles, and Werner Geyer

    Justin D. Weisz, Jessica He, Michael Muller, Gabriela Hoefer, Rachel Miles, and Werner Geyer. 2024. Design Principles for Generative AI Applications. In Pro- ceedings of the 2024 CHI Conference on Human Factors in Computing Systems (Honolulu, HI, USA) (CHI ’24). Association fo...

  29. [39]

    Yuejiao Zhang. 2006. Wiki means more: hyperreading in Wikipedia. InProceedings of the Seventeenth Conference on Hypertext and Hypermedia (Odense, Denmark) (HT ’06) . Association for Computing Machinery, New York, NY, USA, 23–26. doi:10.1145/1149941.1149946

  30. [2012]

    In Proceedings of the Designing Interactive Systems Conference (Newcastle Upon Tyne, United Kingdom) (DIS ’12)

    Slow technology: critical reflection and future directions. In Proceedings of the Designing Interactive Systems Conference (Newcastle Upon Tyne, United Kingdom) (DIS ’12). Association for Computing Machinery, New York, NY, USA, 816–817. doi:10.1145/2317956.2318088

  31. [2020]

    Healthcare 8, 4 (2020), 497

    The Struggle for Human Attention: Between the Abuse of Social Media and Digital Wellbeing. Healthcare 8, 4 (2020), 497. doi:10.3390/healthcare8040497

  32. [2024]

    In Companion Proceedings of the ACM Web Conference 2024 (Singapore, Singapore) (WWW ’24)

    Large Language Model Powered Agents in the Web. In Companion Proceedings of the ACM Web Conference 2024 (Singapore, Singapore) (WWW ’24). Association for Computing Machinery, New York, NY, USA, 1242–1245. doi:10.1145/3589335.3641240

Pith tools

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