Toward Needs-Conscious Design: Co-Designing a Human-Centered Framework for AI-Mediated Communication

Aayushi Dangol; Alexis Hiniker; Jaewon Kim; Robert Wolfe

REVIEW 3 major objections 7 minor 2 references

Toward Needs-Conscious Design: Co-Designing a Human-Centered Framework for AI-Mediated Communication

T0 review · 3 major / 7 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read The paper proposes Needs-Conscious Design, a framework in which AI-mediated communication must preserve three pillars—intentionality, presence, and receptiveness to needs—and warns that AI-generated messages create 'Empathy Fog,' obscuring

desk verdict The three-pillar framework is grounded and useful; Empathy Fog is an overclaimed construct that reads as a theoretical synthesis dressed in participant data. read the letter →

arxiv 2508.11149 v1 pith:LEGDJSVR submitted 2025-08-15 cs.HC

Robert Wolfe , Aayushi Dangol , JaeWon Kim , Alexis Hiniker This is my paper

classification cs.HC

keywords Needs-ConsciousDesignNonviolentCommunicationAI-mediatedEmpathyFogIntentionalityPresenceReceptivenesstoNeedsconsentful

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

Needs-Conscious Design is what the paper argues AI-mediated communication needs: a design orientation centered on human connection rather than fluent output. Drawing on interviews with 14 certified Nonviolent Communication (NVC) trainers and on diary and co-design studies with 13 everyday online communicators, the authors derive three pillars—Intentionality, Presence, and Receptiveness to Needs—that technology should preserve. Their central warning is Empathy Fog: when generative AI drafts or reshapes a message, neither recipient nor sender can tell how much empathy, attention, and effort actually went into it, and that uncertainty erodes the relational value of the exchange. If the framework is right, AI should help people clarify their own feelings and needs rather than produce messages on their behalf.

What carries the argument

The central mechanism is the NVC 'syntax': Observation, Feeling, Need, Request (OFNR), transposed into a design framework. The three pillars—Intentionality, Presence, and Receptiveness to Needs—act as evaluative lenses, and Empathy Fog is the emergent property that appears when AI-generated language breaks the link between a message and the sender's actual investment. Effortful Communication—the idea that discretionary investment and craft make messages meaningful to both recipient and sender—supplies the explanation for why Empathy Fog is harmful.

What would settle it

A controlled messaging experiment: participants receive the same supportive reply, randomly labeled as written by the sender, written with an AI grammar tool, or fully AI-generated. If Empathy Fog is a real mechanism, perceived empathy, closeness, and relationship value should drop in the AI-labeled conditions, and senders who used AI should report lower felt connection after sending. If ratings are unchanged, the framework's central concern is not supported.

Watch

Extended reading notes

Core claim

Drawing on the four-component model of Nonviolent Communication—observing without judgment, naming feelings, tracing them to needs, and making requests—and on co-design sessions with everyday users, the paper defines three pillars of Needs-Conscious Design: Intentionality, Presence, and Receptiveness to Needs. Each pillar carries a design question: does the system maximize the user's agency, does it facilitate human-to-human presence rather than substitute an artificial agent, and does it focus attention on underlying needs rather than blame? The paper's key identified hazard is Empathy Fog: when generative AI drafts or mediates a message, recipients cannot tell how much empathy, attention,

Load-bearing premise

The load-bearing premise is that Nonviolent Communication's four-step model—observe, feel, need, request—is a valid account of how empathetic connection works; the paper itself notes NVC is rarely studied and is practice-based, and if that model is wrong, the three pillars built on it may not transfer to design.

Editorial extensions

If this is right

Features that auto-compose or auto-complete empathetic replies should be avoided, because they hide the sender's effort and trigger Empathy Fog.
Designs should slow interactions down—for example with reflection prompts, calming interruptions, or an empathy flag—so users can act with intentionality.
AI should be positioned as a self-reflection aid or sounding board, not as the voice of one person talking to another, so presence remains human-to-human.
Tools that touch emotional data should offer consent that is voluntary, informed, revertible, specific, and unburdensome, following the paper's guiding questions.
Presence can be signaled in both low-stakes and high-stakes situations, and designs should make such signals visible rather than erase them.

Reading between the lines

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

Empathy Fog likely extends beyond empathy-specific messages: AI-drafted apologies, condolences, congratulations, and even routine replies carry the same ambiguity about sender effort, so the effect could be measured across genres.
The three pillars could be operationalized as a diagnostic audit for existing products: ask whether a feature increases or decreases intentionality, presence, and needs-receptiveness, without requiring users to know NVC.
The framework predicts an asymmetry worth testing: disclosure of AI involvement may harm trust more than message quality itself, making transparency and interaction design as consequential as the underlying model.

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.

Desk Editor's Note

read the letter

First, the useful part: the three-pillar framework — Intentionality, Presence, Receptiveness to Needs — is a genuinely helpful vocabulary for thinking about AI-mediated communication, and the co-design sketches ground it in a way that feels concrete rather than hand-wavy. The study design is honest: 14 NVC trainers plus 13 diary/co-design participants, triangulated across interviews, diaries, and designs, with clear coding procedures. The paper also openly acknowledges that NVC is not an empirically validated theory and that the sample is small and skewed. That transparency earns real credit.

The soft spot is Empathy Fog. The abstract and contribution list say participants identified it, but the Findings never report a participant describing the full phenomenon as defined: a two-sided uncertainty in which the recipient cannot tell how much effort the sender invested and the sender loses their own sense of investment. What you get in the data is T4's doubt that 'it's even you,' T7's warning about 'fooling ourselves,' and P9's sketch of an AI drafting responses — each of these gestures at something nearby, but none of them articulates the mechanism the Discussion builds. The term first appears in Discussion and is explicitly framed through Kelly et al.'s effortful communication theory. That's a legitimate design synthesis, but it is not an inductive finding. The claim 'identified by participants' overstates the evidence. My recommendation would be to either strengthen the empirical grounding or soften the claim to 'participants raised concerns that, along with prior work, led us to define Empathy Fog.'

The rest of the paper holds up. The pillars are consistently derived from trainer and lay-user responses, and the consent questions are a nice addition. The weaknesses are the usual ones for qualitative HCI: small, self-selected, U.S.-centric sample, no deployed system validation, and a framework partly seeded by the researchers' NVC-derived protocol. The paper acknowledges all of this. The citation pattern looks fair; the related work covers trust in AI-written text, AI clones, effortful communication, and the affirmative consent framework.

Who should read it: anyone working on AI-mediated communication or designing supportive communication tools. It is not a paradigm shift, but it gives the field a usable frame and a warning construct that will be cited. It deserves a serious referee; with a revision addressing the Empathy Fog framing, I'd be comfortable seeing it accepted.

Referee Report

3 major / 7 minor

Summary. Wolfe et al. propose Needs-Conscious Design, a framework for AI-mediated communication that centers human connection and is built on Nonviolent Communication (NVC). They conducted interviews with N=14 CNVC-certified trainers and a six-day diary study plus co-design with N=13 lay users, analyzed with a deductive-inductive thematic approach. The paper makes three contributions: (1) a conceptual model with three pillars — Intentionality, Presence, and Receptiveness to Needs — each supported by design concepts and participant sketches; (2) 'Empathy Fog,' an emergent property in which generative AI-mediated messaging obscures the sender's effort, attention, and investment from both recipient and sender; and (3) consentful-design guiding questions applying Im et al.'s affirmative consent framework. The authors argue that AI should support self-reflection rather than substitute for human connection.

Significance. If the findings hold, the paper offers a timely, actionable framework for AI systems entering everyday messaging. Strengths include methodological transparency (detailed protocols, compensation, demographics, and explicit disclosure of NVC's weak empirical base), triangulation of expert and lay participants, and concrete co-design artifacts, plus falsifiable design recommendations (e.g., self-reflection aids should better support relationship outcomes than AI-composed messages). The main risks are evidentiary: the paper's most novel construct, Empathy Fog, is presented as participant-identified but is, on the manuscript's evidence, a researcher synthesis from a few authenticity-related comments; and the NVC-seeded protocol and deductive codes partly prefigure the pillars. Both are addressable.

major comments (3)

[Discussion, 'Empathy Fog in AI-Mediated Communication'; also Abstract/Contribution 2] The abstract calls Empathy Fog a property 'identified by participants'; Contribution 2 says 'we find that empathy fog can also render the sender themselves uncertain.' Yet no Findings subsection documents this, and no quoted participant articulates sender-side loss of felt investment. The three quotes cited (T4 'doubts it's even you'; T7 'fooling ourselves'; P9 'only rely on this technology') concern authenticity and overdependence, not the two-sided uncertainty in the definition. The sentence 'Drawing on this framework [Kelly et al. 2017], we conceptualize...' reveals a theoretical synthesis, not an inductive finding. Since the design recommendation (AI for self-reflection rather than mediation) rests on this construct, either adduce participant evidence for both sides of the uncertainty or reframe Empathy Fog as a researcher-proposed construct with an analytic trail and soften the empi
[Methods (Co-Design paragraph; Data Analysis) and Findings] The pillars are presented as participant-surfaced ('Participants surfaced three common design principles'), but the protocol was seeded with NVC: the co-design asked participants to react to 'thirteen technologies envisioned by the study team based on NVC teaching methods,' the entry/exit instruments probed needs and feelings, and deductive codes 'mirrored our research questions.' The pillars are therefore partly co-constructed by the researchers' framing, not purely emergent. Please add a reflexivity statement distinguishing protocol-elicited from spontaneous themes and soften the 'surfaced' language; this bears directly on the claim that the framework is data-driven.
[Related Work and Limitations] The framework's foundation is NVC, which the authors concede 'is rarely studied in the psychological literature.' If NVC's OFNR model is not a valid account of empathic connection, the three pillars may not transfer to design. This is a correctness-risk concern, not a demand to validate NVC: as a concrete strengthening step, map each pillar onto the empirically grounded dimensions of the frameworks already cited (e.g., Kelly et al.'s effortful communication, Im et al.'s consent) and state future disconfirmation criteria.

minor comments (7)

[Methods, Exit Survey] Typo: 'during the the diary study' has a duplicated article.
[Findings, Pillar 1] 'Needs Consciouness' is a misspelling of 'Needs Consciousness.'
[Findings, 'Design Concepts to Support Presence'] The first sentence is garbled: 'Design Concepts to Support Presence Prioritizes human to human connection proved a chief concern.' Insert a period after 'Presence' and repair the subject–verb agreement.
[Findings, Pillar 2] P7's quote 'after having lost my an uncle' should read 'my uncle' or 'an uncle.'
[Methods, Co-Design] The 'thirteen technologies envisioned by the study team' are never enumerated; an appendix listing them would let readers audit the priming of the co-design.
[Related Work, 'Approaches to Emotional Data'] 'Solove (2023) contend' should be 'contends.'
[Conclusion] 'identifying Empathy Fog as problematic property of generative AI' is missing an article before 'problematic.'

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the framework is induced from an external NVC model and participant data; Empathy Fog is a labeled synthesis, not a fitted/predicted re-statement.

full rationale

The paper's derivation chain is self-contained qualitative work. The three pillars (Intentionality, Presence, Receptiveness to Needs) are built from NVC—an external framework with cited origins in Rosenberg and Chopra—and from thematic coding of CNVC trainer interviews and lay-user diary/co-design data. No parameter is fitted to a subset of data and then renamed as a prediction: the co-designs are used as illustrative examples after the themes were derived, which is standard qualitative practice. Empathy Fog is introduced in the Discussion as a conceptual synthesis, explicitly 'Drawing on this framework, we conceptualize uncertainty induced by AI-mediated communication as Empathy Fog'; although the abstract says it was 'identified by participants,' the paper's own grounding is in participant comments about AI obscuring human attention. Whether those comments sufficiently support the construct is an evidentiary and qualitative-validity concern, not a circularity: the concept is not defined in terms of itself, nor is it the same as the input data by construction. The self-citations in Related Work (e.g., Baughan et al., Fu et al., Lukoff et al.) are contextual prior work and are not load-bearing for the framework's central claims. The paper also transparently limits its premise in Limitations and Future Work, noting that NVC is 'rarely studied in the psychological literature,' which shows the authors are not relying on an unstated self-citation chain. No circular step can be identified by the paper's own equations or definitions.

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

The central claim rests on NVC's validity, the usefulness of the participant sample, and the soundness of the qualitative coding. Each is a domain assumption; no free parameters apply because the framework is qualitative and no quantitative fit is performed.

assumptions (4)

domain assumption The NVC model (Observation-Feeling-Need-Request) is a valid description of empathetic communication and a useful basis for design.
The entire framework builds on NVC principles from Rosenberg and Chopra (2015); the paper does not independently validate NVC's empirical claims.
domain assumption Self-reported experiences of 14 trainers and 13 lay users are representative enough of online communicators to yield general design pillars.
The sample is small and skewed (7 women, 7 aged 25-34); the authors acknowledge limitations but still generalize to a framework.
domain assumption The qualitative coding and thematic analysis reliably capture the participants' meanings.
Two coders used deductive-inductive coding, but no inter-rater reliability or audit trail is reported, so trust in the themes relies on the reported procedure.
domain assumption Design concepts sketched by participants in a one-session co-design reflect their genuine needs and would function as imagined in real systems.
Co-design sketches are hypothetical and not tested; the paper treats them as evidence for design directions.

invented entities (1)

Empathy Fog
purpose: Conceptual construct describing uncertainty about how much empathy, attention, and effort a sender invested in AI-facilitated communication.
Grounded in qualitative participant reports, but not operationally measured or validated outside this study; it is a proposed emergent property, not a tested metric.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Toward Needs-Conscious Design: Co-Designing a Human-Centered Framework for AI-Mediated Communication." pith.science (2026). https://pith.science/paper/LEGDJSVR

@misc{pith2026250811149,
  author       = {Pith},
  title        = {Pith review of: Toward Needs-Conscious Design: Co-Designing a Human-Centered Framework for AI-Mediated Communication},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LEGDJSVR}},
  note         = {Machine review of arXiv:2508.11149}
}

read the original abstract

We introduce Needs-Conscious Design, a human-centered framework for AI-mediated communication that builds on the principles of Nonviolent Communication (NVC). We conducted an interview study with N=14 certified NVC trainers and a diary study and co-design with N=13 lay users of online communication technologies to understand how NVC might inform design that centers human relationships. We define three pillars of Needs-Conscious Design: Intentionality, Presence, and Receptiveness to Needs. Drawing on participant co-designs, we provide design concepts and illustrative examples for each of these pillars. We further describe a problematic emergent property of AI-mediated communication identified by participants, which we call Empathy Fog, and which is characterized by uncertainty over how much empathy, attention, and effort a user has actually invested via an AI-facilitated online interaction. Finally, because even well-intentioned designs may alter user behavior and process emotional data, we provide guiding questions for consentful Needs-Conscious Design, applying an affirmative consent framework used in social media contexts. Needs-Conscious Design offers a foundation for leveraging AI to facilitate human connection, rather than replacing or obscuring it.

Figures

Figures reproduced from arXiv: 2508.11149 by the authors.

**Figure 2.** Needs-Conscious Design requires that users feel they have communicated intentionally. Participants preferred design [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** P12 envisioned an “Empathy Flag” to signal when [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Figures from the paper (5 more)

**Figure 4.** Figure 4: Needs-Conscious Design supports empathetic presence achieved between people, and eschews design that substitutes human connection with artificial companionship, or which fragments attention such that one cannot achieve presence with another person. Top right: Co-desig…

**Figure 5.** Figure 5: A chat assistant envisioned by P2 to help under [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]

**Figure 6.** Figure 6: Needs-Conscious Design emphasizes receptiveness to needs, both of another person and of oneself. Being receptive involves nonjudgmental listening and setting aside grievance to hear what someone else needs, and what one needs oneself. Top right: Co-design with P9, who…

**Figure 7.** Figure 7: Intentionality, Presence, and Receptiveness to [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]

**Figure 8.** Figure 8: Empathy Fog (left), obscuring empathy with AI between sender and recipient; vs. AI for self-reflection (right). • Whether the sender is in fact receptive to their needs, and whether the feelings expressed in the sender’s message in fact characterize the feelings of the…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

2 extracted references · 2 canonical work pages

[2016]

Social work in health care, 55(6): 427–439

Improving interprofessional collaboration: The effect of training in nonviolent communication. Social work in health care, 55(6): 427–439. Nosek, M.; Gifford, E. J.; and Kober, B. 2014. Nonviolent Communication training increases empathy in baccalaureate nursing students: A mixed method study. Office of the U.S. Surgeon General. 2023. Our Epidemic of Lone...

work page 2014
[2023]

From Text to Self: Users' Perceptions of Potential of AI on Interpersonal Communication and Self

The Intricacies of Social Robots: Secondary Analy- sis of Fictional Documentaries to Explore the Benefits and Challenges of Robots in Complex Social Settings. Proceed- ings of the 2023 CHI Conference on Human Factors in Com- puting Systems. Fu, Y .; Foell, S.; Xu, X.; and Hiniker, A. 2023. From Text to Self: Users’ Perceptions of Potential of AI on Interp...

work page Pith review arXiv 2023

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [2016]

Social work in health care, 55(6): 427–439

Improving interprofessional collaboration: The effect of training in nonviolent communication. Social work in health care, 55(6): 427–439. Nosek, M.; Gifford, E. J.; and Kober, B. 2014. Nonviolent Communication training increases empathy in baccalaureate nursing students: A mixed method study. Office of the U.S. Surgeon General. 2023. Our Epidemic of Lone...

work page 2014

[2] [2023]

From Text to Self: Users' Perceptions of Potential of AI on Interpersonal Communication and Self

The Intricacies of Social Robots: Secondary Analy- sis of Fictional Documentaries to Explore the Benefits and Challenges of Robots in Complex Social Settings. Proceed- ings of the 2023 CHI Conference on Human Factors in Com- puting Systems. Fu, Y .; Foell, S.; Xu, X.; and Hiniker, A. 2023. From Text to Self: Users’ Perceptions of Potential of AI on Interp...

work page Pith review arXiv 2023