{"id":"da15060e-7da8-4249-9474-b79b93dc8f89","arxiv_id":"2505.11684","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper adapts Shneiderman's human-AI framework to civic communication by splitting human control into public leader and community organization control, and offers design considerations for constructive engagement.","lead":"This paper proposes a design framework for using AI in community engagement while ensuring that public leaders and community organizations keep control over decisions. It adapts an existing human-AI collaboration model to civic communication and argues that high automation can coexist with human agency.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The load-bearing assumption is that public-leader control and community-organization control are independent and jointly maximizable; the paper's own Arnstein citation suggests they may conflict.","rationale":"The reader's weakest_assumption correctly identifies the dual-control split as the least secure point in the paper. My reading sharpens this concern: it is not merely that the split lacks empirical evidence, but that the paper's own cited source (Arnstein's ladder of citizen participation, §2.1) suggests a structural tradeoff between community power and leader authority, which directly challenges the independence required for the high-high quadrant. Shneiderman's original framework is designed around a single user-operator relationship; transplanting it to a two-stakeholder civic setting without argument is a logical gap, not just an evidentiary gap. The rest of the paper—the literature review, the catalog of AI benefits and risks, and the five design considerations in §3.4—is reasonable and well-grounded, but those components do not carry the central claim. The conditional verdict remains appropriate because the paper is a conceptual design essay and its central recommendation is plausible; however, until the independence of the two control axes is either defended theoretically or demonstrated empirically, the framework's core prescription should be treated as a hypothesis, not an established design principle. The proposed case analysis would provide a direct check by examining whether real systems that instantiate high-high control are associated with constructive communication outcomes, and whether any observed conflicts between the two controls undermine the framework's feasibility.","tokens_in":13886,"tokens_out":3210,"duration_ms":35701,"concrete_test":"Conduct a structured case analysis of the civic engagement systems cited in §2.1–§2.2 (e.g., Polis, Coalesce, CommunityPulse, BoundarEase), coding each system on public-leader control, community-organization control, and automation level using the paper's definitions, then compare against any reported outcome measures of trust, legitimacy, or agency. If no high-high system exhibits the claimed constructive outcomes, or if high community control is achieved only when leader control is low, the independence assumption fails and the recommendation should be reframed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The framework's central recommendation—design for high automation with high control for both public leaders and community organizations—depends on the assumption, introduced in §3.3, that these two forms of human control are independent dimensions that can both be set high. The paper justifies the split by analogy to Shneiderman's elevator example, which concerns a single human operator, not two stakeholder groups with potentially opposed interests. This independence is not established and is in tension with the paper's own citation of Arnstein in §2.1, which frames community power as requiring a redistribution of power from public leaders—implying a zero-sum relationship. If leader control and community control are negatively correlated in real engagement contexts, the high-high quadrant may be unattainable or ill-defined, and the paper's core design directive loses its basis. The paper also does not operationalize either control dimension or the outcome constructs, so the assertion that high-high yields understanding, trust, respect, legitimacy, and agency is unsupported. This is not a disagreement with consensus; it is a correctness risk in the internal construction of the framework.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper is a conceptual essay that proposes a framework for human-AI collaboration in community engagement processes. It defines two communication pathways—between public leaders and community organizations, and among community organizations facilitated by public leaders—and characterizes constructive communication through outcomes such as understanding, trust, respect, legitimacy, and agency. The paper surveys challenges in community engagement, discusses benefits and risks of AI, adapts Shneiderman's human-centered AI framework to two types of human control (public leader control and community organization control), and concludes that beneficial human-AI collaboration requires high automation alongside high control for both stakeholder groups. It then lists five design considerations: critical engagement, flexibility, transparent boundaries, feedback loops, and reflection. The paper does not present empirical data or formal derivations; its central claim is asserted rather than tested.","tokens_in":14051,"tokens_out":2823,"duration_ms":31598,"significance":"If the framework's central claim is valid, the paper would provide a useful synthesis that connects civic-communication research with human-AI interaction design. The paper draws on a broad literature, articulates concrete design considerations, and grounds the discussion in existing systems including the author's own Coalesce, SenseMate, and BoundarEase. The two-pathway framing and the explicit attention to dual stakeholder control are potentially valuable contributions to a conceptual understanding of civic AI. However, the central claim that high automation plus high control for both public leaders and community organizations reliably produces constructive communication is not empirically or formally supported. The paper is best read as a position piece that generates hypotheses rather than a validated framework; its usefulness depends on subsequent operationalization and testing.","major_comments":[{"comment":"The adaptation of Shneiderman's framework in §3.3 splits human control into public leader control and community organization control, but the relationship between these two dimensions is not established. The paper justifies the split by analogy to elevators, which involve a single operator, not two stakeholder groups with potentially opposed interests. This assumption of independence is in tension with the paper's own citation of Arnstein in §2.1, where community power is described as requiring a redistribution of power from public leaders, implying a zero-sum relationship. If public leader control and community organization control are negatively correlated in real engagement settings, the high-high quadrant may be unattainable or ill-defined, and the central design directive loses its foundation. The authors should either provide evidence or argument for independence, or explicitly address the trade-off and define the conditions under which high-high is achievable.","section":"§3.3, with §2.1"},{"comment":"The central claim in §4—that achieving beneficial human-AI collaboration \"depends on design approaches that maintain a high degree of control in decision-making for both public leaders and community organizations while leveraging automation capabilities\"—is asserted rather than derived or measured. The design considerations in §3.4 are presented as implementing the framework, but no operational definitions are given for \"control\" or for the outcome constructs (understanding, trust, respect, legitimacy, agency), and no evidence is provided that these considerations are sufficient to produce constructive communication. The paper should reframe this as a hypothesis or a provisional framework, or it should provide case studies, empirical evaluations, or a formal argument connecting the design considerations to the claimed outcomes.","section":"§4 and §3.4"},{"comment":"The two-dimensional grid in Fig. 2 appears to treat public leader control and community organization control as components of a single \"Human Control\" axis, but the text in §3.3 says these are \"two distinct types of human control.\" The figure and text should clarify whether the two controls are orthogonal to each other, to automation, or both. Without this clarification, the reader cannot assess whether the high-high recommendation refers to a single point in a three-dimensional space or to a region in a two-dimensional space.","section":"§3.3, Fig. 2"}],"minor_comments":[{"comment":"The x-axis label \"HighLow Computer Automation\" appears garbled; the intended label is likely \"Computer Automation\" with a High/Low scale. The figure caption should also state that \"Community Control\" and \"Public Leader Control\" are introduced as refinements of Shneiderman's original \"Human Control\" dimension.","section":"Fig. 2"},{"comment":"There is a subject-verb agreement error in the sentence beginning \"Hallucinations, which occur when AI generates false information..., poses significant risks\"; \"Hallucinations\" is plural and should take \"pose.\"","section":"§3.2"},{"comment":"The legend in Fig. 1 includes a gray line for communication between community organizations without public leader involvement, but this pathway is not discussed in the paper's main analysis. The authors should either explain why it is excluded or remove it from the diagram to avoid confusion.","section":"§1.2, Fig. 1"},{"comment":"The term \"intentional stance\" is introduced with a citation to Dennett but is not defined for a readership that may not be familiar with the philosophy of mind. A brief definition would improve accessibility.","section":"§3.4"},{"comment":"Reference [45] (Klein) lacks full publication details; please provide the venue or repository information.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The paper is a conceptual essay with no empirical validation. If the journal's scope emphasizes empirical contributions, the fit may be questionable; however, as a design-oriented position paper it could still be acceptable after substantial revision. The main technical concern is the unsupported independence assumption between public leader control and community organization control, which the authors should either defend or weaken. The authors should also be encouraged to operationalize the core constructs and to present the central claim as a hypothesis, supported by a research agenda, rather than as an established result."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a well-written conceptual essay that adapts Shneiderman's human-AI collaboration framework to civic communication. The genuinely new bit is splitting 'human control' into two separate dimensions—public leader control and community organization control—and arguing that the target quadrant is high automation with high control for both. If you work in civic HCI, it's worth a read as a map of the design space; just don't expect evidence that the map is correct.\n\nWhat it does well: the paper is clear and honest about being a design essay rather than an empirical study. The literature synthesis is broad and mostly accurate, and the two communication pathways (leader↔community, community↔community via leader) are a useful way to organize the messy landscape of public participation. The design considerations in §3.4—critical engagement, flexibility, transparent boundaries, feedback loops, reflection—are sensible and well-cited. And the idea that AI should serve two distinct stakeholders rather than a single 'user' is a real improvement on the usual one-dimensional human-control framing.\n\nWhere it's soft: the central recommendation in §3.3 and the conclusion assumes that leader control and community control can both be set high without friction. The paper's own Arnstein citation in §2.1 suggests otherwise: meaningful community power often comes at the expense of leader control, at least in the short term. That tension isn't mentioned. The framework never operationalizes either control dimension, so 'high-high' is more of an aspiration than a measurable design target. And since there is no system built or evaluated, the claim that high-high produces constructive communication (understanding, trust, etc.) is asserted rather than supported. For a conceptual piece that's acceptable, but it means the framework is a hypothesis, not a finding.\n\nMinor: Figure 2 is hard to parse—it tries to show two control dimensions plus automation on a 2D grid. Consider a clearer representation.\n\nWho it's for: graduate students and researchers looking for a structured lens on civic AI design, or anyone teaching human-AI collaboration with a civic angle. It could also serve as a position piece to motivate empirical work. I'd be happy to see it in the peer-reviewed literature after the control-tension issue is addressed, either by defending independence or by reframing the framework around trade-offs and negotiation between the two stakeholders. It's not a desk reject; it deserves a serious referee.","headline":"A useful but untested design framework for civic AI; the dual-control split is interesting, but the paper should confront the power tension between leaders and communities.","tokens_in":14561,"tokens_out":3150,"would_cite":false,"duration_ms":31597,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"AI in community engagement should be designed to keep both public leaders and community organizations in high control while using high automation, or it will fail to produce constructive civic communication.","keywords":["community engagement","human-AI collaboration","civic communication","constructive communication","human-centered AI","public participation","design framework","democratic governance"],"falsifier":"A field experiment comparing an AI-assisted engagement tool set to high automation with dual high control against the same tool with, say, high public-leader control but low community-organization control; if community members report no greater agency, trust, or legitimacy under the dual-control condition, or if public leaders find that condition unworkable, the central claim is contradicted.","tokens_in":13631,"feed_emoji":"🗳️","tokens_out":5478,"duration_ms":53146,"temperature":0.7,"pith_summary":"This paper argues that the value of AI in community engagement depends not on automation alone but on design that keeps two distinct human stakeholders, public leaders and community organizations, in control of decisions while the AI handles labor-intensive work. It maps these two forms of control onto a two-dimensional framework whose target quadrant combines high automation with high control for both groups. The paper also identifies two communication pathways whose health determines whether engagement succeeds: dialogue between leaders and organizations, and dialogue among organizations facilitated by leaders. It synthesizes known benefits and risks of AI in civic contexts and proposes five design considerations, critical engagement, flexibility, transparent boundaries, feedback loops, and reflection, that would make the target quadrant realizable. A sympathetic reader would care because this gives civic technology designers a concrete, evaluable target: use AI's capacity while preventing either side from losing agency.","feed_headline":"Best civic AI design pairs high automation with dual human control","feed_subtitle":"A framework argues civic engagement AI should let public leaders and community organizations both keep decision-making control, not just…","key_machinery":"The central object is a two-dimensional human-AI collaboration framework adapted to civic communication. In the original framing, human control and computer automation are independent axes, producing a grid whose 'reliable, safe, and trustworthy' quadrant is high automation combined with high human control; the paper's adaptation splits human control into two axes, public leader control and community organization control, so the design target becomes the quadrant with high automation, high public leader control, and high community organization control. This object carries the argument because every design consideration in the paper, critical engagement, flexibility, transparent boundaries, feedback loops, and reflection and deliberation, is presented as a way to stay inside that quadrant.","core_discovery":"Stated in the paper's own terms, the central claim is that achieving beneficial human-AI collaboration in community engagement depends on design approaches that maintain a high degree of control in decision-making for both public leaders and community organizations while leveraging automation capabilities. The paper grounds this in an adapted two-dimensional framework in which human control and computer automation are separate axes: automation can be high while human control remains high, and in civic settings human control must be split into public leader control and community organization control. The recommended target is the quadrant with high automation plus high control on both stakeholder axes. Within that quadrant, the paper expects community engagement to exhibit the hallmarks of constructive communication: understanding, trust, respect, legitimacy, and agency. The conclusion follows from the paper's synthesis of the civic-communication literature, the documented failure modes of engagement, and the identified risks of unconstrained AI, including hallucination, opacity, bias, and over- or under-reliance.","pith_inferences":["If the framework is right, the right metric for civic AI is not adoption or efficiency but changes in the five relationship outcomes, which the paper does not itself operationalize; building validated scales for those outcomes is a natural next step.","The two-control assumption could be tested directly with a platform that exposes separate control interfaces to each stakeholder group, then varies which controls are active; the paper does not describe such an experiment.","The same dual-control logic may apply to other two-party civic processes, such as participatory budgeting or school-district consultation, though the paper focuses on municipal engagement."],"forward_implications":["Civic AI tools should not be built with a single human-in-the-loop; they need distinct control surfaces for public leaders and community organizations.","Automation that saves time but erodes community agency is a design failure, because the target quadrant requires both stakeholder controls to stay high.","The five design considerations give implementers a checklist: build in critical engagement with AI output, flexible interaction, transparent role boundaries, feedback loops, and reflective deliberation.","Transparency about when and how AI is used is not optional, because opaque AI threatens the perceived legitimacy that constructive civic communication requires."],"supporting_citations":[{"why":"Supplies the two-dimensional human control / computer automation framework that the paper adapts for civic communication.","marker":"[76]"},{"why":"Establishes the ladder of citizen participation and the argument that participation without power redistribution is futile, motivating the demand for community control.","marker":"[4]"},{"why":"Defines community engagement practices and the challenges of municipality-to-citizen interaction that the paper targets.","marker":"[19]"},{"why":"Documents the experience of community feedback going 'into a void,' the core problem that transparent responses and dual control are meant to solve.","marker":"[39]"},{"why":"Documents the civic data deluge and sensemaking challenges that motivate the need for high automation.","marker":"[56]"},{"why":"Provides guidelines for human-AI interaction, including user control and feedback, that inform the design considerations.","marker":"[3]"}],"fun_headline_variants":["Civic AI sweet spot: high automation, dual human control","For civic AI, automate heavily but keep both sides in control","Framework: high automation with twin human controls for civic AI","Civic engagement AI: balance automation with leader and community control","High automation plus dual control: civic AI's constructive path"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that public leader control and community organization control are separable and that maximizing both together with automation produces more constructive civic communication than any other arrangement.","fun_headline_variants_meta":{"raw":{"variants":["Civic AI sweet spot: high automation, dual human control","For civic AI, automate heavily but keep both sides in control","Framework: high automation with twin human controls for civic AI","Civic engagement AI: balance automation with leader and community control","High automation plus dual control: civic AI's constructive path"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000247,"raw_usage":{"total_tokens":1524,"prompt_tokens":910,"completion_tokens":614,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":526,"completion_tokens_details":{"reasoning_tokens":528}},"tokens_in":526,"tokens_out":614,"duration_ms":6115,"temperature":1.0,"reasoning_tokens":528,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T20:49:09.717099+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A field experiment comparing an AI-assisted engagement tool set to high automation with dual high control against the same tool with, say, high public-leader control but low community-organization control; if community members report no greater agency, trust, or legitimacy under the dual-control condition, or if public leaders find that condition unworkable, the central claim is contradicted.","supporting_citations":[],"review_version":1}