{"id":"dd966fad-7318-4649-b342-566e983f3207","arxiv_id":"2411.17894","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A structured catalogue of fairness design patterns, derived from published cases, is proposed with an extended sustainability meta-model and illustrated on two validation cases.","lead":"This paper catalogs reusable fairness design patterns for sustainable information systems, based on an extended sustainability meta-model, and tests them on COVID-19 and child-health monitoring cases. It offers requirements engineers a structured way to elicit and reason about fairness properties in complex sociotechnical systems.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Validation is partly circular: the ONE case used in §4.3 also appears in Table 1 as an input case, so the catalogue's claimed usefulness needs an out-of-sample test.","rationale":"The reader's CONDITIONAL verdict is correct and should stand. My pass adds one sharper point: the ONE case is not merely 'selected by the same authors'; it appears in Table 1, the same inventory from which the patterns were induced, so the second validation risks being in-sample. This does not make the contribution worthless—the catalogue may still be a useful organizing device, the meta-model extension is described, and the COVID case provides a partial out-of-sample illustration—but it means the central claim is currently supported by plausibility and self-assessment rather than by evidence. The paper's own Section 5 limitation statement supports this reading. I therefore keep the verdict at CONDITIONAL: require an independent, out-of-sample evaluation or an explicit reframing of the case studies as illustrative worked examples rather than validation. The reader's weakest assumption captured the self-referential nature broadly; this pass sharpens it to the Table 1 contamination, hence 'partial' agreement.","tokens_in":16870,"tokens_out":7601,"duration_ms":69440,"concrete_test":"Trace the provenance of each pattern instantiated in §4.3 (violation anticipation, transparency, accessibility) to the case inventory in Table 1. If any of those patterns can be traced to the [Chi+15] ONE row—as the text suggests—the second validation is a resubstitution test. To settle usefulness, the authors should then run a blinded study: analysts unfamiliar with the catalogue apply it to a genuinely out-of-sample case, and their coverage of fairness requirements is compared with a checklist baseline. The trace alone settles the circularity; the blinded study settles whether the catalogue adds value.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that the fairness-pattern catalogue supports requirements engineers in identifying fairness requirements. That claim presupposes the patterns transfer to new systems. The second validation case does not test transfer: Section 4.3 uses the ONE early-childhood monitoring system, and Section 1.4 Table 1 row 6 already lists 'Health Early childhood care' with citation [Chi+15]—the same system that §4.3 says was 'published as a representative example in terms of sustainability [Chi+15]'. ONE was therefore part of the input corpus used to discover and document patterns, so applying the catalogue to ONE is a resubstitution exercise, not an independent validation. The COVID case is genuinely absent from Table 1, but it was selected, modelled and interpreted by the same authors who built the catalogue. Section 5 concedes the catalogue 'has only been tested by a small group of researchers close to the authors' and that 'it remains difficult to estimate the level of quality or usefulness of the catalogue'. The subsequent claim that stability of the authors' COVID models 'indicates the high quality of the methodology and the catalogue' is a non-sequitur: model stability is not a measure of transferable usefulness. Thus the load-bearing assumption—that the catalogue, not the authors' prior familiarity with fairness modelling, drives the observed results—is currently unsupported.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a catalogue-based methodology for modelling and classifying fairness concerns in sustainable sociotechnical systems. It builds on an existing sustainability meta-model and extends it with the concepts of Obstacle, Assumption and Fragment (§2.1). From an inventory of sixteen published cases (Table 1), the authors derive six fairness patterns—distributive justice, substantial freedom, rule acceptance, transparency, violation anticipation, and co-evolution—and organise them into a PDCA-style cycle (§3). The claimed contribution is that the resulting library enables a requirements engineer to identify fairness requirements more systematically than with ad hoc analysis. Validation consists of two author-conducted case studies: COVID-19 crisis management (§4.1–4.2), which is not in Table 1, and the ONE early-childhood monitoring system (§4.3), which is listed as row 6 of Table 1.","tokens_in":17112,"tokens_out":4070,"duration_ms":35314,"significance":"If the central claim were established, the paper would make a useful contribution to requirements engineering for sustainability: it offers a concrete modelling vocabulary, a documented set of fairness patterns, and an explicit procedure for instantiating them. The systematic inventory of cases and the explicit extension of the meta-model are valuable, and the paper is honest about several limitations in Section 5. However, the validation is currently self-assessment: the same team discovered, documented, applied and evaluated the patterns, and Section 5 concedes that the catalogue \"has only been tested by a small group of researchers close to the authors\" and that \"it remains difficult to estimate the level of quality or usefulness of the catalogue\". The strongest independent element is the COVID-19 case, which is genuinely absent from the input corpus, but its interpretation is still author-dependent. The paper does not provide a baseline comparison, an inter-rater agreement measure, or a test of whether engineers without prior exposure to the catalogue produce different or better fairness requirements.","major_comments":[{"comment":"The ONE early-childhood monitoring system is not an out-of-sample validation. Table 1 row 6 lists \"Health Early childhood care\" with reference [Chi+15], and §4.3 explicitly states that ONE \"was published as a representative example in terms of sustainability [Chi+15]\". Since the pattern catalogue was derived from the case studies in Table 1, applying the catalogue to ONE is a resubstitution exercise. This does not demonstrate transfer of the catalogue to a new system; it demonstrates that the authors can re-model a system they already know. The paper should either replace this case with a genuinely external one, or clearly label the ONE analysis as a within-sample illustration and move the burden of transfer evidence to independent cases.","section":"§4.3 and Table 1"},{"comment":"The concluding inference from COVID model stability to catalogue quality is a non-sequitur. The text states: \"The initial models proved to be very stable and have been reproduced here virtually unchanged. This indicates the high quality of the methodology and the catalogue.\" Model stability only indicates consistency of the authors' own modelling over time; it does not measure usefulness for third-party analysts and does not support the transferability claim. The paper needs either an external evaluation (e.g., a user study with engineers not involved in pattern creation, or a comparison of the requirements produced with and without the catalogue) or a substantially weaker statement of the claimed contribution.","section":"§5"},{"comment":"The pattern discovery and validation procedures are both performed by the authors without explicit criteria for what counts as a recurrence or a successful application. Section 2.2 lists four steps but gives no operational definition of \"recurring strategy\", and Section 4 uses the authors' own annotations (speech bubbles) to decide where patterns \"seem interesting to apply\". This makes the positive evaluation partly circular and leaves the reader unable to distinguish pattern-driven analysis from the authors' prior domain knowledge. The paper should specify a protocol that separates pattern selection from pattern evaluation, for example by having independent raters apply the catalogue and measuring agreement.","section":"§2.2 and §4"}],"minor_comments":[{"comment":"There are OCR and encoding artifacts: \"Ãl’conomic\" (§3.1), \"stratÃl’gy\" (Table 1 row 6), \"ails at\" (§2.3), and \"prokect\" (Acknowledgement). These should be corrected in a professional copy-edit.","section":"Throughout"},{"comment":"The template fields are inconsistent: \"Dimensions\" in §3.3 is filled with a sentence about community characteristics rather than the sustainability dimensions addressed, and \"Content\" in §3.2 repeats the same sentence. The pattern documentation should use one field discipline.","section":"§3.2 and §3.3"},{"comment":"The meta-model extension in Figure 3 is described with French terminology while Figure 1 uses English; the text should state whether the language switch is intentional and should provide an English caption for Figure 3.","section":"Figure 3"},{"comment":"The definition of the violation anticipation pattern relies on indicators and models, but the archetype diagram does not show how the \"obstacle\" extension from §2.1 is used; adding a pointer to the extended meta-model would help.","section":"§3.5, Figure 11"}],"recommendation":"major_revision","confidential_remarks":"The paper is acknowledged to be an English translation of a previously published French article [Chr24]. The main risk is overclaiming transferability: the ONE case is within-sample, and the Section 5 inference from model stability to catalogue quality is not warranted. A major revision that either adds external validation or reframes the paper as an experience report with a clearly limited claim would make it publishable in the journal."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this is the English translation of the authors' French journal paper [Chr24], built on their own [PNT21]. The core meta-model, pattern catalogue, and COVID-19 case are already there. What's new is a more systematic state-of-the-art review, a doubled input case base, and the ONE early-childhood validation.\n\nThat said, the paper is a decent qualitative contribution. The pattern catalogue is clearly organised around a PDCA cycle with governance, and each pattern is documented with intent, applicability, archetype, examples, and related patterns. The meta-model extensions—obstacle, assumption, fragment—are conservative and genuinely useful for reasoning about barriers to fairness. The authors are frank about their own limitations, which I respect.\n\nThe soft spot is the validation. ONE appears in Table 1 as an input case, so applying the catalogue to it is resubstitution, not an independent test. The COVID case is truly outside the input corpus, but it was selected, modelled, and interpreted by the same team that built the catalogue, with no baseline or external evaluators. Section 5 concedes it's only been tested by a small group close to the authors. The further claim that the stability of the COVID models 'indicates the high quality of the methodology' is a non-sequitur—stability over time is not evidence of transferable usefulness. The load-bearing assumption—that the catalogue, not the authors' own familiarity with fairness modelling, is what drives the results—is not supported.\n\nNone of this kills the paper. The approach is plausible and the catalogue may well be useful to requirements engineers. But the evidence presented is suggestive, not demonstrative. The citation pattern is fine: they explicitly say this is a translation and build on their own prior work, which is normal.\n\nWho should read it: requirements engineers and RE researchers interested in sustainability and fairness patterns. It's a niche audience, but the paper is a reasonable entry point.\n\nMy recommendation: send it to peer review. It deserves referee time. The validation can be improved with an independent, out-of-sample case and at least one external user. With that, it could be a solid contribution. Without it, the transfer claim remains unproven.","headline":"Useful English translation of prior fairness-pattern work; the catalogue is coherent but the validation is partly circular and the transfer claim remains unproven.","tokens_in":17642,"tokens_out":3378,"would_cite":false,"duration_ms":30327,"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":"A reusable library of fairness patterns, supported by an extended sustainability meta-model, enables systematic fairness analysis and information-system requirement discovery in sustainable system design.","keywords":["fairness patterns","sustainable information systems","design patterns","requirements engineering","socio-technical systems","sustainability meta-model","distributive justice","case study"],"falsifier":"Have independent requirements engineers, who did not help build the catalogue, apply it to a new sustainability case in an unfamiliar domain (for example, water quota allocation or public transport planning) and compare the fairness requirements they produce with those from unstructured analysis. If the catalogue-based group does not identify more complete or more consistent fairness requirements, the central claim fails. The paper already notes that wider dissemination and feedback collection are missing.","tokens_in":16681,"feed_emoji":"⚖️","tokens_out":6570,"duration_ms":51908,"temperature":0.7,"pith_summary":"The paper tries to establish that fairness in sociotechnical system design can be treated as a reusable engineering concern rather than an ad hoc afterthought. It builds a catalogue of fairness patterns—such as distributive justice, transparency, rule acceptance, violation anticipation, and co-evolution—and documents them with an extended sustainability meta-model that adds obstacles, assumptions, and modular fragments. The authors claim that this catalogue lets analysts model fairness-related values, spot barriers and assumptions, and instantiate patterns to derive concrete requirements on the information system serving the system. They test the claim on two cases: COVID-19 crisis management (containment and vaccination) and the medico-social monitoring of early childhood. A sympathetic reader would care because, if the catalogue is reusable, fairness requirements could be elicited more systematically across domains.","feed_headline":"Fairness gets a reusable pattern library for sustainable design","feed_subtitle":"Two case studies show the patterns surface fairness requirements and IS features from socio-technical analysis.","key_machinery":"The key machinery is the extended sustainability meta-model and the fairness pattern catalogue built on it. The meta-model, based on the value/dimension/indicator/regulation/activity framework, is extended with three concepts: Obstacle (a condition blocking a value), Assumption (a behaviour that must hold for the value to be guaranteed but is not enforced by the system), and Fragment (a modular model element that can be generalised into a reusable Patron). These extensions let analysts reason about barriers to fairness and modularise complex models. The patterns are organised around a continuous improvement cycle—Plan-Do-Check/Study-Act—with stages for design, adoption, implementation, evolution, and governance, and each pattern's archetype is expressed with the extended notations, linking fairness values to activities, indicators, obstacles, and IS components.","core_discovery":"The central discovery is that fairness patterns can be induced from a set of published case studies and organised into a structured library positioned on a continuous improvement cycle (design, adoption, implementation, evolution, governance). Each pattern is described with a summary, applicability, content, archetype, examples, and related patterns, and is expressed in an extended sustainability meta-model that introduces the concepts of Obstacle, Assumption, and Fragment/Patron. The meta-model extension is what carries the reasoning: obstacles make fairness violations explicit and trigger pattern selection, assumptions make hidden conditions explicit, and fragments turn reusable chunks into generic patterns that can be instantiated in a specific context. Validation on the COVID-19 and ONE cases leads the authors to conclude that the approach identifies fairness strategies and yields IS requirements in a systematic and reusable way.","pith_inferences":["A natural next step, not explored in the paper, would be to convert the obstacle-detection step into a tool that suggests candidate patterns semi-automatically, reducing reliance on analyst expertise.","The catalogue's generalisability could be tested by independent analysts applying it to a new domain, such as water allocation or public transport, and comparing outputs against ad hoc analysis; the paper acknowledges it has only been tested by a small group close to the authors.","The fairness patterns overlap conceptually with algorithmic fairness concerns in AI systems, but the paper does not address bias in machine-learning models; connecting the two would be a testable extension."],"forward_implications":["Requirements engineers can use the catalogue to identify fairness strategies systematically, moving from ad hoc analysis to pattern-based reasoning.","The meta-model extensions (obstacle, assumption, fragment) provide a reusable reasoning vocabulary that can be applied to other sustainability dimensions, such as resource use and environmental footprint.","The pattern catalogue is designed to grow through community enrichment, accumulating knowledge from new cases.","The approach bridges socio-technical fairness analysis and information-system requirements, producing concrete IS obligations from high-level fairness values.","The cyclical structure supports continuous improvement, so fairness can be monitored and restored as systems evolve."],"supporting_citations":[{"why":"Supplies the base sustainability meta-model and graphical syntax that the paper extends with obstacle, assumption, and fragment concepts.","marker":"[PF13]"},{"why":"Provides the fairness reference model that grounds the reasoning framework and meta-model basis.","marker":"[Kie+20]"},{"why":"Provides the textual sustainability pattern template used to document each fairness pattern.","marker":"[RR13]"},{"why":"Offers the modelled equality fragment that feeds the pattern discovery process.","marker":"[HC15]"},{"why":"Presents the oncology care pathway case study used to illustrate pattern discovery and fairness indicators like RDI.","marker":"[PL18]"},{"why":"Provides the earlier analysis of the ONE early childhood monitoring system used as a validation case.","marker":"[Chi+15]"},{"why":"Is the authors' prior work on fairness analysis by pattern, which this article builds on and extends.","marker":"[PNT21]"},{"why":"Supplies the systematic literature review whose WHY/WHAT/HOW axes inform the catalogue's cycle structure.","marker":"[Sav+19]"}],"fun_headline_variants":["Fairness pattern library for sustainable information system design","Reusable fairness patterns for designing sustainable systems","Fairness patterns library supports sustainable IS design","Modelling fairness patterns for sustainable information systems","Fairness pattern library guides sustainable system design"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper's positive conclusion rests on the assumption that two cases, selected and analysed by the same authors who built the catalogue, are representative enough to demonstrate the catalogue's quality and usefulness; the paper itself says it remains difficult to estimate that quality and that the catalogue has only been tested by a small group close to the authors.","fun_headline_variants_meta":{"raw":{"variants":["Fairness pattern library for sustainable information system design","Reusable fairness patterns for designing sustainable systems","Fairness patterns library supports sustainable IS design","Modelling fairness patterns for sustainable information systems","Fairness pattern library guides sustainable system design"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000215,"raw_usage":{"total_tokens":1383,"prompt_tokens":854,"completion_tokens":529,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":470,"completion_tokens_details":{"reasoning_tokens":462}},"tokens_in":470,"tokens_out":529,"duration_ms":4608,"temperature":1.0,"reasoning_tokens":462,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T11:42:57.609041+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Have independent requirements engineers, who did not help build the catalogue, apply it to a new sustainability case in an unfamiliar domain (for example, water quota allocation or public transport planning) and compare the fairness requirements they produce with those from unstructured analysis. If the catalogue-based group does not identify more complete or more consistent fairness requirements, the central claim fails. The paper already notes that wider dissemination and feedback collection are missing.","supporting_citations":[],"review_version":1}