{"id":"88b650e7-c36e-4e14-b3c5-db415a57736a","arxiv_id":"2509.01006","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"REConnect is a participatory requirements engineering framework, drawn from three community projects, that makes trust, co-design, and user empowerment the core of requirements work and assigns humans defined governance roles in AI-assisted RE.","lead":"REConnect is a requirements engineering framework that puts long-term human relationships, co-design, and community empowerment at the center of software projects, built from three community software efforts in Nepal, Canada, and the Arctic. It argues that fast, AI-driven requirements gathering can miss the social and cultural context that decides whether software is adopted and survives.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central claim rests on undocumented retrospective narratives; no described analysis or comparison supports the causal link from REConnect principles to outcomes.","rationale":"The reader's weakest_assumption correctly identifies the retrospective, self-authored narratives as the load-bearing evidence. My stress test sharpens this: the manuscript does not even describe the methodological procedure by which the narratives were turned into the framework, so there is no way to audit the inferential step from observations to principles. This is not an ad hominem; it is a standard concern about qualitative research transparency. The abstract's claim of a 26-project synthesis is unbacked in the body, and the three cases are presented as illustrations rather than as a systematic dataset. The concrete test—independent coding of raw artifacts with a matched comparison—would settle whether the reported causal sequences are actually present in the data or are post hoc reconstructions. This would also address the absence of a baseline. I therefore agree with the reader's assessment and do not recommend changing the verdict: CONDITIONAL is appropriate, with the condition being the provision of such methodological transparency and external validation. The paper's other weaknesses (reference hygiene, positionality) are real but secondary; they do not bear directly on the central claim's validity.","tokens_in":82,"tokens_out":2556,"duration_ms":45180,"concrete_test":"Obtain the INSPIRE program's archived records for BloodSync, Herluma, BridgingRoots, and one matched non-REConnect project (same program, similar community, different engagement approach). Have an independent researcher, blind to the REConnect framework, code the raw meeting notes, workshop transcripts, design docs, and adoption logs using a pre-registered codebook for (a) occurrence of trust-building, co-design, empowerment actions, and (b) evidence of outcomes (adoption, legitimacy, sustainability). Then compare the three projects against the matched control. If the independent coding does not show that REConnect actions precede and covary with the outcomes, or if the control project shows similar outcomes without REConnect actions, the causal claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that trust-building, co-design, and empowerment yield culturally grounded, socially legitimate, sustainable requirements—depends entirely on the three INSPIRE case narratives (Sections 3.1–3.3). The abstract promises a 'reflective synthesis across all 26 projects' and 'in-depth thematic analysis,' but the body never describes that analysis: no coding scheme, no interview/workshop data, no triangulation, no negative cases. Section 4 then 'draws on the three case studies' to induce the framework, making the evidence and the framework co-constructed. Moreover, the counterfactual at the start of Section 3 ('without which... likely leading to a lack of system adoption') asserts causality without any comparison project or baseline. Confounds such as community readiness, prior relationships, funding, and the Hawthorne effect are uncontrolled. Because the authors are program insiders, the narratives' accuracy and causal attributions cannot be independently checked from the manuscript. If the principles are not actually necessary or sufficient for the reported outcomes, the framework reduces to a plausible but unsupported proposal.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper introduces REConnect, a participatory requirements engineering framework whose central thesis is that human connection—manifested as trusting relationships, co-design, and user empowerment—should be the foundation of RE for socially sustainable software. The authors claim that REConnect was derived from qualitative analysis of 26 community-engaged projects in the INSPIRE program and is illustrated through three case studies: BloodSync (rural Nepal), Herluma (Victoria, Canada), and BridgingRoots (Arctic Canada). The framework is operationalized as three principles and nine REActions, and the paper further proposes human governance roles for AI-assisted RE. The contribution is positioned against CrowdRE and automated, LLM-based elicitation, arguing for sustained relational engagement rather than large-scale automated extraction.","tokens_in":21288,"tokens_out":3401,"duration_ms":45409,"significance":"If the framework and its supporting evidence were empirically sound, this paper would make a valuable contribution to the human-centered and participatory RE literature. It articulates a timely critique of scale-driven CrowdRE and GenAI approaches, and it translates a vague commitment to 'human values' into concrete, actionable practices (trust-building actions, co-design actions, empowerment actions) that practitioners could adopt. The proposed human-AI governance roles (guardians of values, curators and contextualisers, inclusion amplifiers, accountability partners, co-reflectors) are a useful conceptual scaffold for future RE tool design. The paper is also honest in not claiming to eliminate AI from RE; it attempts a nuanced integration. However, these strengths are conceptual. The manuscript does not provide a verifiable qualitative analysis, does not report data or analysis procedures, and does not test the causal claims it makes. The current evidence base consists of three retrospective narratives authored by members of the program that ran the projects, which is insufficient to support the paper's central generalizability and effectiveness claims.","major_comments":[{"comment":"The abstract claims REConnect 'was derived through qualitative analysis of 26 community-engaged software projects' and that the authors 'conducted a reflective synthesis across all 26 projects, followed by in-depth thematic analysis of three illustrative projects.' The body never describes this analysis: there is no coding scheme, no corpus description, no analysis procedure, no triangulation, and no account of how the 26 projects were synthesized. Instead, §3 presents only three retrospective narratives and §4 states, 'Drawing on the three case studies... we introduce REConnect.' The 26-project basis is therefore asserted but not evidenced. Either report the qualitative analysis of all 26 projects or revise the derivation claim to reflect the actual basis (three illustrative cases). This is load-bearing because the abstract's method claim is what distinguishes the paper from a purely经验b","section":"Abstract; §3; §4"},{"comment":"There is a circularity risk in the evidence structure. Section 3 introduces the three projects as 'three cases that have inspired our proposal,' and Section 4 then 'draw[s] on the three case studies' to present REConnect's principles and REActions. The principles are abstracted from these same cases and then the cases are cited as supporting evidence for the principles. This does not provide independent validation. The authors should either test the framework on held-out projects (e.g., other INSPIRE projects), include negative cases, or clearly frame the case studies as illustrative rather than confirmatory. As written, the case narratives cannot both inspire and validate the framework without additional evidence.","section":"§3 opening; §3.1–3.3; §4"},{"comment":"The paper makes causal claims about the necessity and effectiveness of human connection. The opening of §3 asserts that without the human-connection practices, 'the subtle elements that mattered... would not have surfaced, likely leading to a lack of system adoption.' This is a counterfactual causal claim with no comparison project, no baseline, and no control for confounds such as community readiness, prior relationships, funding, or the attention effect of the visits themselves. Similar causal language appears in the conclusion ('can yield requirements that are culturally grounded, socially legitimate, and sustainable'). The reported outcomes (adoption, legitimacy, community ownership, youth empowerment) may be attributable to these confounds rather than to the REConnect principles. The authors should either provide comparative evidence from the other 23 INSPIRE projects (e.g., project","section":"§3, first paragraph; §6"},{"comment":"The human-AI governance roles in Table 3 (guardians of values, curators and contextualisers, inclusion amplifiers, accountability partners, co-reflectors) are presented as an extension of REConnect but are not derived from the three case studies nor evaluated in any way. The mapping from REConnect steps to these roles appears plausible but is entirely speculative. This is acceptable for a proposal, but the paper should label these roles as a design proposal rather than an empirically grounded component of the framework. Otherwise, readers may mistake an unvalidated taxonomy for a tested result.","section":"§5.2; Table 3"}],"minor_comments":[{"comment":"The name is inconsistently capitalized: 'REConnect' (title, most sections), 'REconnect' (§4, §5.1), and 'RECOnnect' (Table 3 and §5.2). Please standardize.","section":"Throughout"},{"comment":"Typo: 'each each step' should be 'each step.'","section":"§5.2, Step 5"},{"comment":"The claim that 'GPT-5' represents the 'state-of-the-art' is vague and will date quickly. Cite a specific model version with a date or avoid this phrasing.","section":"§5.2"},{"comment":"Many reference URLs include tracking parameters such as '?utm_source=chatgpt.com.' Please remove these parameters and verify that the cited sources are the canonical versions.","section":"References"},{"comment":"The 'Summary' subsection is useful but reads as a restatement of the narratives rather than a synthesis. Consider renaming it 'Synthesis' and explicitly connecting each narrative element to a REConnect principle, which would also help address the analysis-transparency concern.","section":"§3.4"},{"comment":"The introduction and abstract should be aligned regarding the number of projects: the abstract says 26 projects, while the introduction says 'three case studies.' If the 26-project analysis is not reported, the introduction's framing is more accurate.","section":"§1; Abstract"}],"recommendation":"major_revision","confidential_remarks":"The paper is best viewed as a vision/experience-report contribution rather than an empirical study. The mismatch between the abstract's methodological claims (26-project qualitative analysis) and the body's actual evidence (three retrospective narratives) is significant and should be addressed before publication. If the authors can either supply the missing analysis or honestly reframe the contribution as a proposal grounded in illustrative cases, the paper could be a worthwhile addition. Editorially, I would also flag that the authors are members of the program that ran the projects; independent data or at least a clear statement of insider positioning would strengthen credibility."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this one if you want a clear statement of the participatory-RE-for-social-sustainability position, but do not take the empirical claims at face value. The genuinely new part is the packaged REConnect process—three principles, nine REActions, and especially Table 3's stakeholder roles in human-AI collaboration (guardians of values, curators, inclusion amplifiers, accountability partners, co-reflectors). That is a useful vocabulary for people thinking about governance of AI-assisted RE. The paper also does a good job of situating itself in participatory design, value-based RE, and self-determination theory; the synthesis is honest and readable.\n\nThe soft spots are the ones flagged in our notes, and they are real. The abstract promises a reflective synthesis across 26 INSPIRE projects with in-depth thematic analysis of three. The body never describes that analysis—no coding scheme, no data sources, no triangulation—and Section 4 explicitly draws only on the three cases. That overstates the evidence. The three cases were selected because they embody the principles, so the framework and the evidence are co-constructed. The counterfactual at the start of Section 3 ('without which... likely leading to a lack of system adoption') is an untested causal claim with no comparison or baseline. Confounds like community readiness, prior relationships, funding, and Hawthorne effects are uncontrolled. The authors are program insiders, and the manuscript does not discuss positionality or threats to validity. That matters for a qualitative paper.\n\nThe reference hygiene issue is minor but real: multiple URLs carry utm_source=chatgpt.com parameters, suggesting undisclosed AI-assisted reference curation. Not fatal, but should be cleaned up and disclosed.\n\nNone of this means the paper should be desk-rejected. As a position paper, it is thought through and even interrogates its own premise in Section 5.1. The REActions are actionable, and the AI-governance discussion is the most original and useful part. The right verdict is conditional: revise to align the abstract with what is actually presented, add a limitations/positionality section, and either present the cases as illustrative narratives or do the promised thematic analysis. If the latter, we need the method.\n\nI would send it to a serious referee. It is not a rigorous empirical contribution yet, but it is a valuable conceptual one. For a reading group, it would spark good discussion about what counts as evidence in community-based RE.","headline":"A useful conceptual framework with an overclaimed empirical base—send it to review, but only after the authors align the abstract with the actual evidence.","tokens_in":21896,"tokens_out":2609,"would_cite":false,"duration_ms":29234,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that requirements engineering for socially impactful systems should be re-centered on sustained human connection—trust, co-design, and empowerment—rather than automated extraction of user data.","keywords":["requirements engineering","participatory design","human values","community engagement","social sustainability","generative AI","trust","co-design"],"falsifier":"A controlled comparison of two similar community projects, one following REConnect's REActions and one using conventional or AI-assisted elicitation, observed from the start and matched on prior relationships, funding, and community readiness: if the relational project does not show greater stakeholder legitimacy, adoption, or post-delivery sustainability, the causal claim fails. A shorter test: interview transcripts or project logs showing that reported trust and empowerment narratives were post-hoc rationalizations, or that requirements were actually driven by external constraints rather tha","tokens_in":1767,"feed_emoji":"🤝","tokens_out":1979,"duration_ms":93450,"temperature":0.7,"pith_summary":"This paper argues that requirements engineering for socially impactful software has drifted toward automated, distanced elicitation—mining app reviews, classifying user feedback with AI—and that this drift systematically excludes the communities most affected. It introduces REConnect, a participatory framework whose central claim is that human connection, not extraction, is the foundation for understanding lived experiences. The framework rests on three principles: building trusting relationships, co-designing with and alongside stakeholders, and empowering users as agents of change. Drawing on three long-running community projects—a blood-donation platform in rural Nepal, a shelter navigator for women at risk of homelessness in Canada, and an Indigenous language game in the Arctic—the paper claims these principles yield requirements that are culturally grounded, socially legitimate, and sustainable beyond delivery. It then specifies how AI can assist requirements tasks while humans remain guardians of values and interpreters of AI output.","feed_headline":"Relationship-first requirements beat automated elicitation","feed_subtitle":"A participatory framework built on trust, co-design, and empowerment claims culturally grounded, adopted systems.","key_machinery":"The central object is the REConnect method itself: three principles—building trusting relationships, co-designing with and alongside users, and empowering users as agents of change—each operationalized by a set of nine REConnect Actions (REActions). The REActions are the load-bearing mechanism that turns the principles into practice: Contextual Stakeholder Mapping, Immersive Field Presence, Community-engaging Facilitated Workshops; Iterative Prototyping and Simulation, Community-engaged Requirement Mapping, Co-Design Artifacts within Socio-economic Context; Motivation Mapping, Engaging Empowered Stakeholders, and Capacity Building and Training. The argument works by showing, through the thre","core_discovery":"The paper's discovery is a method, REConnect, derived from a reflective synthesis of 26 community-engaged software projects and a deep thematic analysis of three of them: BloodSync (blood coordination in rural Nepal), Herluma (shelter bed navigation for women at risk of homelessness in Canada), and BridgingRoots (an Inuvialuktun language-learning game in the Canadian Arctic). Across these, the authors claim, requirements did not emerge from one-off elicitation events but were negotiated social artifacts produced through sustained human connection. REConnect operationalizes this through three principles—building trusting relationships, co-designing with and alongside users, and empowering use","pith_inferences":["If the causal story holds, the unit of analysis in requirements engineering should shift from individual requirements to the relationship itself; evaluation would then measure the capacity of a process to build trust, not just the quality of extracted needs.","A testable extension: in a larger sample of community-engaged projects, score adherence to the REActions and test whether it predicts adoption and sustainability beyond project type, funding, and prior relationships.","The framework implies a boundary condition for automation: steps requiring lived presence and motivation mapping cannot be safely delegated to AI, so organizations that skip the groundwork step would lose the context that later AI-assisted analysis depends on.","The principles overlap with community-based participatory research in other fields; REConnect's distinctive contribution is making them operational as requirements-specific actions and as explicit governance roles for human-AI collaboration."],"forward_implications":["Requirements teams working with marginalized communities should budget for sustained field presence and repeated visits, treating trust as a precondition for requirements rather than a side effect.","AI tools in requirements engineering should be positioned as assistants whose outputs are interpreted, challenged, and overridden by stakeholders, with humans retaining final interpretive authority.","Stakeholder roles should expand beyond formal system users to include informal influencers—elders, ward leaders, frontline workers—as co-designers and co-owners of both the requirements and the system.","Requirements quality for socially impactful systems should be judged partly by social legitimacy, community ownership, and post-delivery sustainability, not only by completeness or correctness.","The REActions give practitioners a concrete checklist linking each principle to elicitation, analysis, and validation activities across the project lifecycle."],"supporting_citations":[{"why":"Supplies the value-based requirements engineering tradition and method that REConnect extends, including the elicitation of stakeholder values, motivations, and emotions.","marker":"[78]"},{"why":"Supplies the community-engaging, citizen-centered co-production approach that grounds REConnect's call to expand participation beyond individual users.","marker":"[85]"},{"why":"Defines CrowdRE, the automated and semi-automated crowd-based requirements paradigm that REConnect positions itself against.","marker":"[27]"},{"why":"Provides the automated app-review classification approach that illustrates the scaled, distanced elicitation practices REConnect argues are insufficient.","marker":"[47]"},{"why":"Shows how app store reviews are used as a requirements and bug-report source, anchoring the critique of feedback mining without context.","marker":"[58]"},{"why":"Establishes that trust in requirements elicitation matters and is built over time, supporting the relationship-building principle.","marker":"[14]"},{"why":"Grounds the Indigenous engagement principle in respect for culture, community-participatory action, and the need to move beyond extractive relationships.","marker":"[89]"},{"why":"Supplies evidence that rural health care in Nepal functions through local relationships rather than formal systems, anchoring the BloodSync trust premise.","marker":"[59]"},{"why":"Provides the motivational modelling and homelessness context that informs Herluma's understanding of stigma and lived experience.","marker":"[16]"},{"why":"Supports the claim that generative AI can produce human-like requirements artifacts, motivating the AI-enhanced workflow discussion.","marker":"[66]"}],"fun_headline_variants":["Trust, co-design, and agency: the new RE","REConnect: putting human bonds back in requirements","Community-led requirements for social sustainability","Automation can't replace connection in requirements engineering"],"cache_read_input_tokens":23552,"weakest_assumption_plain":"The framework's evidence is the project team's own retrospective accounts of three projects; the central claim that trust-building, co-design, and empowerment caused the reported adoption and legitimacy depends on those accounts being accurate and on no unseen factor—such as community readiness, funding, or prior relationships—doing the work.","fun_headline_variants_meta":{"raw":{"variants":["Trust, co-design, and agency: the new RE","REConnect: putting human bonds back in requirements","Community-led requirements for social sustainability","Automation can't replace connection in requirements engineering"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000251,"raw_usage":{"total_tokens":1409,"prompt_tokens":777,"completion_tokens":632,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":521,"completion_tokens_details":{"reasoning_tokens":583}},"tokens_in":521,"tokens_out":632,"duration_ms":8154,"temperature":1.0,"reasoning_tokens":583,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T12:59:33.035516+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A controlled comparison of two similar community projects, one following REConnect's REActions and one using conventional or AI-assisted elicitation, observed from the start and matched on prior relationships, funding, and community readiness: if the relational project does not show greater stakeholder legitimacy, adoption, or post-delivery sustainability, the causal claim fails. A shorter test: interview transcripts or project logs showing that reported trust and empowerment narratives were post-hoc rationalizations, or that requirements were actually driven by external constraints rather tha","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the value-based requirements engineering tradition and method that REConnect extends, including the elicitation of stakeholder values, motivations, and emotions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines CrowdRE, the automated and semi-automated crowd-based requirements paradigm that REConnect positions itself against."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the automated app-review classification approach that illustrates the scaled, distanced elicitation practices REConnect argues are insufficient."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows how app store reviews are used as a requirements and bug-report source, anchoring the critique of feedback mining without context."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Grounds the Indigenous engagement principle in respect for culture, community-participatory action, and the need to move beyond extractive relationships."},{"cited_title":"Panday, P","cited_arxiv_id":null,"evidence_quote":"Supplies evidence that rural health care in Nepal functions through local relationships rather than formal systems, anchoring the BloodSync trust premise."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supports the claim that generative AI can produce human-like requirements artifacts, motivating the AI-enhanced workflow discussion."}],"review_version":1}