Generative artificial intelligence (GenAI) has sharply reduced the cost of producing fluent explanations, syntheses, analyses, code, and recommendations, but it has not reduced the intellectual work required to establish whether those outputs deserve belief or use. This asymmetry creates a plausibility-verifiability gap: an AI-generated artifact can display the linguistic and structural markers of expertise while the evidence, provenance, and limitations needed to warrant reliance remain difficult to inspect. Existing AI literacy and evaluative-judgement frameworks specify broad competencies for critical and responsible engagement, yet students and novice users still need an actionable method for evaluating a particular AI-mediated artifact. To address this gap, this conceptual paper introduces and theoretically grounds the PEARLS framework, an artifact-level verification protocol organized around six interdependent dimensions: Process, Evidence, Access, Reproducibility, Legitimacy, and Source. The framework integrates insights from epistemic cognition, epistemic vigilance, cognitive offloading, calibrated trust, evaluative judgement, and open-science principles to treat AI output as a provisional knowledge claim whose warrant must be assembled and examined. It further advances verification-driven learning as a pedagogical mechanism through which learners develop expertise by iteratively focusing consequential claims, tracing and testing their warrant, judging uncertainty and legitimacy, acting on the results, and reframing subsequent inquiry. Five interdisciplinary cases - psychology theory, educational statistics, computer science, history, and health sciences - demonstrate how the relative emphasis of the six dimensions varies with disciplinary standards and the consequences of error. The paper concludes by deriving implications for assessment design and proposing a research agenda encompassing construct validation, intervention studies, disciplinary calibration, equity, and human-AI interface design. PEARLS is therefore offered as a theoretically informed and practically usable scaffold for preserving human epistemic responsibility as knowledge production becomes increasingly AI-mediated.
The emergence of generative artificial intelligence has introduced a new technological presence into religious life, capable of producing sermons, prayers, hymns, biblical commentary, devotional material, and liturgical text, and of assuming visible roles in worship through avatars and synthetic voices. This article critically examines the place of AI in Christian worship from a Christian and Reformed theological perspective, asking whether such systems can be said to mediate the Divine. It distinguishes between technological mediation, religious mediation and theological mediation, and argues that AI can mediate religious information, facilitate religious experience and assist human participation in worship, but cannot mediate the Divine in the Christological and soteriological sense in which Christian theology understands divine mediation. Drawing on Christian doctrine, Reformed theology of worship, recent scholarship on AI and religion, and empirical evidence from AI-led worship, the article develops a four-level framework for evaluating AI participation in worship: instrumental assistance, ministerial assistance, delegated religious agency and liturgical substitution. It further proposes a five-part theological matrix (agency, authority, accountability, embodiment and mediation) for assessing specific applications, and six principles for a theology of responsible AI-assisted worship. The article concludes that AI may legitimately function as an instrument within worship when its use is transparent, accountable, subordinate to Scripture and ecclesial authority and directed toward human flourishing, but that the Church should resist treating AI as a spiritual subject, an independent religious authority, a sacramental agent or an alternative mediator between God and humanity.
Godswill Ome Ufere· Zenodo (CERN European Organi...· 0 citations
The emergence of generative artificial intelligence has introduced a new technological presence into religious life, capable of producing sermons, prayers, hymns, biblical commentary, devotional material, and liturgical text, and of assuming visible roles in worship through avatars and synthetic voices. This article critically examines the place of AI in Christian worship from a Christian and Reformed theological perspective, asking whether such systems can be said to mediate the Divine. It distinguishes between technological mediation, religious mediation and theological mediation, and argues that AI can mediate religious information, facilitate religious experience and assist human participation in worship, but cannot mediate the Divine in the Christological and soteriological sense in which Christian theology understands divine mediation. Drawing on Christian doctrine, Reformed theology of worship, recent scholarship on AI and religion, and empirical evidence from AI-led worship, the article develops a four-level framework for evaluating AI participation in worship: instrumental assistance, ministerial assistance, delegated religious agency and liturgical substitution. It further proposes a five-part theological matrix (agency, authority, accountability, embodiment and mediation) for assessing specific applications, and six principles for a theology of responsible AI-assisted worship. The article concludes that AI may legitimately function as an instrument within worship when its use is transparent, accountable, subordinate to Scripture and ecclesial authority and directed toward human flourishing, but that the Church should resist treating AI as a spiritual subject, an independent religious authority, a sacramental agent or an alternative mediator between God and humanity.
Godswill Ome Ufere· Zenodo (CERN European Organi...· 0 citations
Universities are having three separate conversations about the same question: one about student devices, another about generative artificial intelligence (GenAI) in assessed work, and a third, unspoken, that leaves instructors to decide both at their own professional risk. The question underneath: does the technology serve the formation of a capacity, or do the work in the student’s place? This paper holds the three as one, at the level where curricular design happens: the programme. On a refrain-then-amplify design, a programme withholds a generative tool while a capacity is forming, then restores it to amplify that capacity once the student can direct it, judge what it returns, and answer for it. Devices are allowed where they support engaged work, excluded where they drain attention. Both sit in a single floor beneath every course, the first governed by a forming-versus-offloading criterion: whether a stretch of work forms a capacity or puts it through the tool. The programme fixes outcomes and integrity, reserving teaching method to the instructor, and places a hard-to-fake checkpoint at each refrain-to-amplify hinge. Each element has published precedent: the withholding, the taught restoration, the test. None of that work owns the movement at programme level, with a verified transition.
Manuel Torres-Sahli, Jorge Blake, Ángela Novoa‐Echaurren et al.· 0 citations
Reach audiences
Advertise in front of researchers, engineers, and readers.
This methodological working paper introduces a rigorous framework of Analogical Structural Isomorphism to map the precise, highly ordered interior geography of the soul’s passive purification as articulated by St. John of the Cross. Grounded in the doctrinal foundations of Thomistic hylomorphism (ST\ I,\ Q.76,\ A.1) and Divine Simplicity (ST\ I,\ Q.3), this study transcends mere poetic metaphor. It positions the laws of the physical universe as lower-resolution pedagogical scaffoldings—the "fingerprints of the Logos"—that mirror the higher-resolution spiritual laws governing the interior life. AI / ASSISTIVE TECHNOLOGY DECLARATIONIn accordance with emerging academic transparency standards and ecclesiastical guidelines for published scholarship, the author discloses the use of generative artificial intelligence (LLM) tools during the preparation of this working paper. AI technology was utilized strictly in an assistive capacity for structural drafting, cross-referencing of canonical texts, comparative stylistic refinement, and formatting optimization. All theological synthesis, canonical interpretations, doctrinal evaluations, and final text selections remain entirely the original work and responsibility of the author.
Fr Joseph Gee· Zenodo (CERN European Organi...· 0 citations
Contemporary pastoral practice within the Catholic Church is increasingly marked by the integration of lay-led spiritual renewal programs, deliverance protocols, and psychological inner-healing curricula. While initially deployed to revive nominal faith, these initiatives frequently institutionalize structural, canonical, and theological distortions. This paper analyses three widespread lay-led phenomena—Life in the Spirit Seminars (LSS), Unbound Ministry, and Little Way Healing Ministries—evaluating their operational frameworks through the combined lenses of Tridentine sacramental dogmatics, Sanjuanist mystical theology, primary operational manuals, and the 1983 Code of Canon Law. It demonstrates how these programs: Subvert the ex opere operato efficacy of the Sacraments of Initiation. Systematically invade the internal forum in violation of Canons 220 and 240 §2. Usurp the cura animarum reserved to sacred orders under Canons 519, 129 §1, and 274 §1. Misdiagnose the mystical Passive Night of the Senses as demonic oppression, psychological trauma, or spiritual pathology. By incorporating the primary directives of the Life in the Spirit Seminars Team Manual and comparing lay interventions to the biblical comforters of Job as analysed by St. John of the Cross, this study argues for a clerical re-assertion of sacramental realism, canonical governance, and the classical Carmelite understanding of the Cross (2 Cor 12:7–9). The critical danger of modern lay-led reductions is not merely that they introduce theological or canonical confusions, but that they structurally operationalize the very adversarial trap St. John of the Cross identified in Living Flame of Love (Stanza 3, §64). By stationing their diagnostic and emotional protocols precisely at the threshold where sense gives way to spirit, these movements intercept souls during the Passive Night of the Senses, feeding the sensory faculties with artificial activity precisely when God is attempting to introduce them into silent, contemplative union. While contemporary revisions of the LSS manual (e.g., Boucher, 2000) have updated their formal text to appease diocesan oversight and excise explicit prohibitions against conditional prayer, pastoral reality reveals a persistent disconnect: in practice, local parish groups continue to operate with de facto autonomy, retaining the operational DNA, performance-based methodologies, and internal-forum invasions established in the original 1979 framework. AI / ASSISTIVE TECHNOLOGY DECLARATIONIn accordance with emerging academic transparency standards and ecclesiastical guidelines for published scholarship, the author discloses the use of generative artificial intelligence (LLM) tools during the preparation of this working paper. AI technology was utilized strictly in an assistive capacity for structural drafting, cross-referencing of canonical texts, comparative stylistic refinement, and formatting optimization. All theological synthesis, canonical interpretations, doctrinal evaluations, and final text selections remain entirely the original work and responsibility of the author.
Fr Joseph Gee· Zenodo (CERN European Organi...· 0 citations
An information brief submitted to Japan’s Ministry of Education, Culture, Sports, Science and Technology (MEXT)24 June 2026Kenji Yamada, Research Coordinator, Shinyoshi Junior High School, Kasaoka City, Okayama Haruka Shibata, Professor, Graduate School of Human and Environmental Studies, Kyoto UniversityGenerative AI is rapidly changing the information environment in which children live. Drawing on classroom data and national statistics, this brief sets out the current situation, the outlook, and a set of options for response.
K. Yamada, Haruka Shibata· Zenodo (CERN European Organi...· 0 citations
This repository contains the supplementary materials and experimental data supporting the research article: "From Semantic Retrieval to Conversational Agent: A Web-Based RAG Architecture for Interactive System Dynamics Modeling". The dataset is divided into two primary components: the source model environment (search space) and the raw experimental benchmarks evaluating retrieval performance across different levels of user expertise and conversational search strategies. 1. Model Corpus, Queries, and Scenarios This section contains the definitions, domain classifications, and configurations used to build the semantic search environment and simulate user interactions. System Dynamics Models: Contains the extracted, curated, and serialized structural definitions of 63 System Dynamics models. These models cover diverse application domains, including Ecology, Macroeconomics, Smart Cities, Agriculture, and Epidemiology. User Queries and Intents: A dataset contrasting authentic broad novice search intents (e.g., "Show me health-related models") with theoretically perfect, expert-formulated structured queries requiring specific domain vocabulary. Benchmark Scenarios: 37 standardized benchmark scenarios engineered to evaluate cross-disciplinary semantic and lexical search performance across the system. Relevance judgments were established a priori by two domain experts, independently of any system output, and comprise 95 scenario–model relevance pairs 2. Experimental Benchmarks The benchmark execution logs provide a quantitative comparative analysis of different retrieval paradigms, running on a local AI ecosystem with direct CPU inference. File: conversational_rag_benchmark_metrics.csv: This file contains the aggregate Information Retrieval metrics (Precision@5, Recall@5, MRR, and nDCG@5) calculated for the 37 test scenarios. Note that the MRR_Mean column is computed over the full retrieval list (L = 10), whereas the paper reports MRR at the evaluation cutoff k = 5; two rows are affected (scenario 18, Method C, 1/7; scenario 37, Method E, 1/6), where the first relevant document falls beyond the top five, and setting both to zero reproduces the Table 5 values exactly. File: inference_latency_logs.csv: Documents the execution timestamps and hardware latency logs for the local ONNX inference engine, tracking the multi-turn conversational delays. File: ablation_study_p_values.csv: Contains the statistical hypothesis testing (paired t-tests) results validating the significance of the agentic retrieval improvements. File: contextless_retrieval_test.csv: Contains the isolated experimental data evaluating the impact of conversational memory (Method F). Evaluated Methodologies (Ablation Study) The benchmark data tests the following six retrieval paths: Method A: Broad Intent (Direct Retrieval Baseline) using standard single-turn semantic search. Method B: Agentic Refinement (Real Multi-Turn Agent Path) representing the complete conversational architecture. Method C: Expert Semantic Baseline (Direct Retrieval), establishing semantic search performance under optimal input conditions. Method D: Apache BM25 (Lexical over Expert Query) testing exact keyword matching. Method E: Expert Query via Agent (Single Agent Turn) to assess system robustness against over-complication. Method F: Contextless User Refinement (Direct Retrieval), submitting the user's raw Turn 2 answer directly to the vector database, thereby bypassing both the conversational history and the generative query rewriting step. Key Finding - Retrieval Accuracy: Replacing the static search baseline (Method A) with the Agentic Orchestrator (Method B) improves mean nDCG@5 from 0.1066 to 0.4422, a rise of over 300%. Expressed as retrieval success, Hit@5 rises from 0.1892 to 0.5946. Key Finding - Lexical vs. Semantic Dynamics: Under optimal conditions with expert queries, exact lexical matching (Method D) outperforms dense retrieval on every reported metric, achieving an MRR@5 of 0.8784 and an nDCG@5 of 0.8053 against 0.6856 and 0.5750 for semantic search (Method C). Key Finding - Computational Latency: The logs document the latency overhead of local CPU processing. A complete multi-turn exploratory session (Method B) averages 59.33 s (SD = 23.02 s), whereas structurally complete expert queries (Method E) execute in 49.81 s (SD = 9.84 s). Each scenario was executed as an independent cold-start process, so these values include ONNX session initialisation and constitute an empirical upper bound rather than steady-state deployment latency.
Pavel Kyurkchiev· Zenodo (CERN European Organi...· 0 citations
This dataset provides the supplementary materials for the article "An Action Research on Generative AI-Based 'Wide-Area Research' for Mitigating the Usage Gap in the Art Field" (in Korean). It contains the controlled-vocabulary codebook (24 concepts), the full article-level coding results for 310 articles collected over 11 weeks (27 April – 12 July 2026), a full URL audit dataset, a cross-model recoding reliability report, and the complete agent prompt specifications. Contents are limited to summaries and metadata; no full text of source articles is reproduced. Documentation is in Korean. 본 자료는 논문 「미술현장의 정보격차 해소를 위한 생성형 AI 기반 '광역 리서치' 실행연구」의 보충자료입니다. 생성형 AI 기반 리서치 워크플로를 11주간(2026.4.27~7.12) 운용하여 산출한 주간 리포트 「Art & AI Weekly Report」의 기사 310건에 대한 코드북, 기사 단위 코딩 결과, URL 전수 검증 데이터셋, 교차 검증 자료, 에이전트 프롬프트 전문을 수록합니다. 제3자가 논문의 집계 결과를 직접 재현하고 코딩 규칙의 타당성을 감사(audit)할 수 있도록 공개합니다. 자료의 범위와 한계본 자료는 요약과 메타데이터에 한정되며, 원 기사의 본문은 포함하지 않습니다. 각 기사의 열람 경로는 원 매체 링크로만 제공됩니다. 코딩은 저자가 확정한 코드북 규칙을 프롬프트로 주입하여 생성형 AI가 자동 부여하였고, 저자가 310건 전체에 대해 제목·요약과 부여 태그를 대조하는 전수 검수를 수행하였습니다. 다만 이는 AI 코더와 동일한 입력을 대상으로 한 규칙 적용의 검수이며, 요약이 원 기사를 정확히 반영하는지에 관한 원문 대조는 전수로 이루어지지 않았습니다. 수록 파일 README.md — 자료 안내, 집계 기준, 재현 절차, 검증 로그 codebook_24concepts.csv — 24개 개념의 부여기준과 클러스터 대응 coding_310_wide.csv — 기사 단위 코딩 결과 (310행) coding_310_long.csv — 기사–개념 쌍 (452행) concept_frequency.csv / cluster_frequency.csv — 개념별·클러스터별 집계 url_audit_310.csv / url_audit_310_README.md — URL 전수 검증 데이터셋과 판정 기준 reliability_crossmodel_report.docx — 표본 62건(seed=42) 교차 검증 보고서 recompute_alpha_per_concept.py — 개념별 Krippendorff's α 재계산 스크립트 agent_prompts_full.docx — 에이전트별 지시문·검증 규칙·오류 대응표 결합 키 — 엔트리ID(주차-카테고리-순번, 예: W05-국내큐레이션-03)는 코딩 결과와 URL 검증 데이터셋을 1:1로 연결합니다. 모든 집계값은 coding_310_long.csv로부터 프로그램적으로 산출·검증되었습니다.*v2 추가 사항 (독립 코더 신뢰도 보충)*심사위원 의견 및 신뢰도 검증 강화를 위해 독립 코더(인간 코더)가 수행한 코딩 원자료 및 일치도 분석 데이터(`intercoder_reliability_supplement.zip`)를 추가 수록하였습니다.
K. Kim· Zenodo (CERN European Organi...· 0 citations
Generative artificial intelligence (GenAI) has intensified pressure on universities to redesign assessment while maintaining integrity, equity, and validity. Frameworks such as the Artificial Intelligence Assessment Scale (AIAS) offer one response, but evidence of how faculty experience their implementation remains limited. This qualitative study examines AIAS implementation as two parallel cases: a private international university in Vietnam, which adopted the first version institution-wide, and a public UK university, which adopted the revised version in a single school. Data from five focus groups with 30 faculty members were analysed using hybrid thematic analysis, with Critical AI Literacy as a sensitising concept. Six themes were developed: recognising and integrating AI, facilitating conditions, building capacity, pathways to adoption, ethics in practice, and reframing pedagogy. Faculty valued the AIAS as a shared language for legitimising GenAI use, clarifying boundaries, and prompting reflection on assessment design. However, they described implementation as shaped by governance, tool access, faculty confidence, workload, integrity concerns, disciplinary context, and alignment with learning outcomes. Their accounts suggest the AIAS could prompt authentic assessment design and student engagement, but may become a compliance layer when disconnected from those conditions. The study contributes evidence on the conditions through which GenAI frameworks reach pedagogical enactment.
Mike Perkins, Darius Postma, Jasper Roe et al.· Assessment & Evaluation in H...· 0 citations
本Rスクリプトは、地方財政分析のために設計された、ggplot2およびpatchworkを活用した自動・高性能なグラフィック生成モジュールである。先行リポジトリ(https://zenodo.org/records/20258441)で構築したデータ抽出パイプラインのシームレスな可視化拡張として機能し、論文・報告書クオリティの個別チャートや複数指標を統合したダッシュボードを自動生成する。処理済みの財政データセットを動的にスキャンし、厳密な数値フォーマット(%表記の小数点第1位固定など)を適用しながら、ファイルの上書き防止機能を備えた高解像度画像を一括エクスポートする。コードの厳密な再現性と即時利用性を検証するため、実行時に直接生成された未加工のサンプルグラフ(豊山町_公債費負担比率.png)を同梱している。 生成AIの利用について 本レポジトリのコード作成には生成AIを利用した。V1.0まではGoogle Gemini、V2.0以降はClaude(Anthropic)を用いて作成した。出力の検証は著者が行い、内容の責任は著者が負う。 Version 2.0 Release: 決算状況調10年パネルへの対応 本バージョン(V2.0)では、原データの系統を追加した。従来の「財政状況資料集」に加え、「市町村別決算状況調」から作成した10年パネル(平成27〜令和6年度、https://doi.org/10.5281/zenodo.22144642)を入力として扱える。作図モジュールは変更しておらず、入力を作る前段(prepare_panel_for_graphics.R)を追加することで対応した。描画対象は13指標で、目的別歳出から2款を含む。豊山町(愛知県)で実行したサンプル画像を同梱している。 関連Rコード 【類似団体検索・列挙ツール】二自治体の財政比較を行う場合に便利な類似団体コードの同一の自治体を検索・列挙するRコード(Shiny)を作成しました。https://doi.org/10.5281/zenodo.20387194 【連動データ抽出・処理モジュール】 本スクリプトでの可視化に必要な複数年の自治体財政指標データを抽出・前処理(「地方自治体財政指標に関する財政状況資料集を用いた10年間集計エクセルデータ作成」)する際は、基盤となるデータ処理モジュール( https://doi.org/10.5281/zenodo.20258441 )を参照のこと。 [English] R Code for Regional Analysis: Automated Visualization and Image Export Using ggplot2 and patchwork This R script provides an automated, high-performance graphic generation module designed for regional fiscal analysis, leveraging the power of ggplot2 and patchwork to produce publication-ready individual charts and integrated multi-indicator dashboards. Operating as a seamless visualization extension to the primary data extraction pipeline established in the preceding repository (available at https://zenodo.org/records/20258441 ), this module dynamically scans processed financial datasets, enforces rigorous decimal formatting (e.g., standardizing percentage representations to ".1f%"), and batch-exports high-resolution assets with native file-overwrite protection. To verify the script's strict reproducibility and out-of-the-box utility, the accompanying sample graph (豊山町_公債費負担比率.png, showing the Debt Service Burden Ratio of Toyoyama Town) is provided as a raw, unaltered file directly generated by the execution of this code. Declaration of Generative AI Use Generative AI was used to write the code in this repository. The code up to version 1.0 was written with Google Gemini; from version 2.0 it has been written with Claude (Anthropic). The author has verified the output and takes responsibility for the content. Version 2.0 Release: Support for the Settlement Survey Ten-Year Panel This version adds a second family of source data. Alongside the Financial Status Documents used since version 1.0, the module now accepts the ten-year panel (FY2015–FY2024) built from the Local Government Financial Settlement Survey (市町村別決算状況調), deposited at https://doi.org/10.5281/zenodo.22144642. The plotting module (compile_municipal_finance_10years_graphics.R) is unchanged. It takes a data frame of one fiscal year per row and decides how to draw each indicator from the column name alone: columns marked as amounts are drawn as bar charts, columns whose name contains the word for ratio as percentage lines, and the remainder as index lines. Because the interface is a data frame rather than a file format, supporting a different source requires only a new front end. The script added here, prepare_panel_for_graphics.R, selects one municipality from the panel by its code and assembles that data frame; the municipality and the expenditure categories to be drawn are set at the top of the script. Thirteen indicators are drawn. Six are amounts (total revenue, total expenditure, real balance, standard financial demand, standard financial revenue, standard fiscal scale), five are ratios or indices (index of financial capability, ratio of current expenditure to current revenue, real balance ratio, real debt service ratio, debt service burden ratio), and two are expenditure categories taken from the expenditure-by-purpose table (social welfare and civil engineering). Plotting expenditure categories over time is new in this version. Sample images produced by running the code on Toyoyama, Aichi Prefecture (municipality code 233421) are deposited as raw, unaltered output. Related R Code [Peer Municipality Search & Listing Tool] Developed an R Shiny application designed to search and list municipalities belonging to the same peer group code. This tool is highly useful for conducting comparative financial analysis between peer municipalities. https://doi.org/10.5281/zenodo.20387194 [Integrated Data Extraction & Processing Module] For extracting and preprocessing multi-year municipal financial indicator data required for visualization in this script (specifically, "Creating a 10-Year Aggregate Excel Dataset Using the Financial Status Documents on Local Government Financial Indicators"), please refer to the underlying data processing module. https://doi.org/10.5281/zenodo.20258441
Yasutoshi Moteki· Zenodo (CERN European Organi...· 0 citations
As generative AI makes it easier to create synthetic visuals, AI-driven visual disinformation isbecoming more common on social media. However, while much research highlights its potentialharm, less is known about how to reduce its potential to mislead. In this study, we thereforeconducted a preregistered online experiment in the Netherlands (N=1,018) to test the effectivenessof various platform interventions: (1) AI labels or “watermarks,” (2) fact-check labels, and (3)community notes. We tested how effective these sources are in lowering credibility of the falsevisual and belief in the false claim it portrays across two polarizing topics: climate change andimmigration. Overall, the interventions showed no significant differences in effectiveness. Thiswas the case when pooling both topics together and for climate-change related disinformation inisolation. However, for visual disinformation about immigration, community notes were mosteffective, especially among participants with strong anti-migrant views. Our findings suggest thatwhile labeling has limited impact overall, its effectiveness varies by context, and no one-size-fits-all solution exists for combating AI-generated visual disinformation.
Teresa Weikmann, Marina Tulin, Michael Hameleers et al.· Digital Journalism· 0 citations
A new method, called CW-Net, translates the reasoning process of an autonomous vehicle’s AI system into understandable concepts that explain its behavior.
