{"id":"7c5a6338-81e9-40c2-a4d8-33d38400e906","arxiv_id":"2412.12107","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A literature-based proposal of twelve competencies for generative AI literacy, intended as a framework for education, policy, and assessment.","lead":"This paper proposes twelve competencies defining what it takes to be literate with generative AI, from basic AI knowledge to prompt engineering, ethics, and law. It provides a framework for educators, policymakers, and trainers designing curricula and assessments for responsible use of generative AI.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 'defining' claim rests on a non-systematic review and a six-record search that cannot establish completeness or non-redundancy of the twelve competencies.","rationale":"I read the paper in good faith. It is a clearly written conceptual contribution that assembles a plausible set of twelve competencies for generative AI literacy, with useful examples and discussion of implications. The authors are transparent about the limits of their method: Section 2 states the review is non-systematic and the implications are speculative, and Section 6.1 acknowledges the framework may need revision as technology evolves. Those admissions are to their credit. However, the title and the opening claim use the word 'defining', which asserts that these twelve competencies are the canonical, comprehensive set. That assertion is load-bearing because it motivates the framework's use in assessment, curriculum design, and policy. The method section does not provide a reproducible path from literature to list: there is no explicit inclusion/exclusion protocol for the broader literature, no coding of extracted competencies, no test of mutual exclusivity, and no external validation such as expert agreement or empirical assessment. The six-record database search supports a novelty claim only for the exact phrase searched, not for the broader conceptual space of generative AI competence frameworks. The reader's conditional verdict is appropriate: the paper is useful as a starting proposal, but the strongest claim should be tempered or supported by a more rigorous derivation. My concrete test is a systematic review plus expert mapping; it would settle whether additional competencies or redundant entries actually exist. Because my concern reinforces rather than changes the reader's conditional verdict, I recommend no change in the verdict.","tokens_in":21639,"tokens_out":2682,"duration_ms":27218,"concrete_test":"Run a preregistered systematic review of generative AI literacy and competence frameworks, using expanded queries ('generative AI literacy', 'GenAI literacy', 'language model literacy', 'generative AI competencies', 'AI literacy' AND 'generative') across Scopus, Web of Science, ACM Digital Library, IEEExplore, ERIC, and arXiv, with backward and forward citation chaining up to the paper's acceptance date. Extract all proposed competencies from each retrieved framework and map them onto the twelve. If any framework contains a competency not mappable to the twelve, or if independent coders judge two or more of the twelve as redundant, the 'defining' status fails and the paper should be repositioned as one proposed framework pending empirical validation.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's strongest claim is that the twelve competencies are the defining set for generative AI literacy. That claim requires the set to be both complete and non-redundant. The evidence offered is a December 2023 database search limited to the exact phrases 'generative AI literacy' and 'generative artificial intelligence literacy' across five databases, yielding six records (Section 3.1), plus an explicitly non-systematic literature review (Section 2). No selection criteria, coding scheme, inter-competency independence check, or external validation is reported for the list in Table 3. The Discussion (Section 6.1) concedes that the framework reflects the current state of the field and may need revision. A broader or more recent search could surface prior or parallel frameworks with additional competencies, and expert judgment could identify overlaps within the twelve, for example Competency 12 is transversal to the other eleven and Competency 1 overlaps with Competency 2. The central claim is therefore not supported by the method actually used; this is a correctness risk for the strongest claim, not merely a wording preference.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper proposes a competency-based model of generative AI literacy, identifying twelve competencies ranging from foundational AI literacy through prompt engineering and programming to ethical, legal, and continuous-learning abilities. The authors argue that these competencies delineate a general definition of generative AI literacy and can serve as a framework for assessment, curriculum development, and policy. The method combines a non-systematic literature review with a narrow database search (six records, Section 3.1), and the potential implications are explicitly labeled as speculative in Sections 2 and 4.","tokens_in":21792,"tokens_out":4744,"duration_ms":38019,"significance":"If the twelve competencies were validated as complete and non-redundant, the framework could provide a useful common vocabulary for educators, policymakers, and researchers, complementing existing AI literacy frameworks such as Long and Magerko (2020). The paper's strengths are its transparency about the non-systematic method and speculative implications, and its clear, exemplified descriptions of each competency in Section 5. However, the evidence presented does not support the 'defining' status of the set, so the contribution is best read as a proposal or starting point rather than an established definition.","major_comments":[{"comment":"The paper's central claim, stated in the title and abstract, is that the twelve competencies are 'defining' for generative AI literacy. That claim requires both completeness and non-redundancy of the set. The supporting evidence is an explicitly non-systematic literature review (Section 2) and a database search restricted to two exact phrases across five databases, yielding six records (Section 3.1). No selection criteria, coding scheme, inter-rater agreement, or external validation is reported for the list in Table 3, and Section 6.1 concedes that the framework reflects the current state of the field and may require revision. As written, the evidence cannot distinguish the twelve competencies from the authors' judgment, so the 'defining' claim is not supported. I recommend either softening the claim to 'proposed' or 'initial' competencies, or substantially strengthening the method with a systematic review and a validation step.","section":"Sections 2, 3.1, Table 3"},{"comment":"The Introduction states that 'no set of competencies specific to the use of generative AI had been previously proposed.' This novelty claim is contradicted by the paper's own Table 1, which lists Dadhich and Bhaumik (2023) proposing a 'Comprehensive Model' of generative AI literacy and Noh and Han (2023) implementing a 'generative AI literacy education program.' The authors dismiss these works as conflating generative AI literacy with general AI literacy, but this dismissal is asserted rather than demonstrated. Because the novelty of the contribution rests on this gap, the authors should either provide a more systematic analysis of the six records, including any competency lists they contain, or qualify the gap claim.","section":"Section 1, Table 1"},{"comment":"The twelve competencies are not shown to be non-redundant, which is required if they are 'defining.' Competency 12 ('Ability to continuously learn') is explicitly described as 'transversal to all the other capabilities' in Section 4 and as applying 'to each of the other ones' in Section 5.12. Similarly, Competency 1 ('Basic AI literacy') and Competency 2 ('Knowledge of generative AI models') overlap in that the latter presupposes the former. If the model is intended as a set of independent competencies, the authors should justify the boundaries; alternatively, the model should be presented as a structured learning progression with a transversal meta-competency, which would be a different and more defensible claim.","section":"Table 3, Sections 4 and 5.12"}],"minor_comments":[{"comment":"The phrase 'a long-standing review approach' is unclear; presumably 'a systematic review approach' was intended, or the sentence should be reworded for clarity.","section":"Section 2"},{"comment":"The sentence beginning 'Alongside, this paper defines...' uses an awkward transition; consider 'In addition, this paper defines...' or 'The paper also defines...'.","section":"Section 4"},{"comment":"The phrase 'to which extent' should be 'to what extent.'","section":"Section 5.9"},{"comment":"The database search was conducted in December 2023, but the manuscript was received in January 2024 and contains a March 2024 regulatory reference (the European AI Act approval) elsewhere; the authors should acknowledge that the 'state of the art' on generative AI literacy may have advanced since the search date.","section":"Section 3.1 and Table 1"},{"comment":"The positive and negative implications in Table 3 are described as mostly speculative in Section 4, but the table itself does not display this caveat; consider adding a footnote to the table so the caveat is visible without reading the main text.","section":"Table 3"}],"recommendation":"major_revision","confidential_remarks":"The paper is transparent about its limitations, which is a genuine strength, but the title and abstract overstate what the method can support. In my view the contribution is a reasonable initial proposal for a generative AI literacy framework, and the authors should calibrate the claims accordingly (e.g., 'proposed competencies' rather than 'defining competencies'). The six-record database search is a weak base for a completeness claim, and the internal overlaps among competencies (e.g., Competency 12 as transversal) make the 'defining' wording especially problematic. If the authors are unwilling to soften this claim, rejection may be the appropriate outcome; if they are, major revision is feasible."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear [Colleague],\n\nYou should know about arXiv:2412.12107. It is a clearly written conceptual paper that integrates existing AI literacy, prompt engineering, and ethics literature into twelve competencies for generative AI literacy. The authors are transparent about their method: they call the literature review non-systematic, report that the database search in Section 3.1 returned only six records, and label their implications speculative. That honesty is a genuine strength.\n\nWhat is new is the integration and recontextualization. No prior framework specifically for generative AI literacy appears in the cited literature, and the twelve competencies do cover the relevant terrain. The examples are practical, and the public-sector discussion is plausible. As a starting point for curriculum or policy conversations, it has real value.\n\nThe soft spots are proportional. The title says 'defining,' but the method cannot support completeness or non-redundancy. Six records and a non-systematic narrative review are too thin to establish that no other framework exists or that these twelve are the definitive set. There are also internal overlaps: Competency 12 is transversal rather than a peer of the others, and Competency 1 overlaps with Competency 2. The authors acknowledge in Section 6.1 that revisions may be needed as the field evolves, so this is not a hidden defect.\n\nI do not see a load-bearing flaw. The central argument that generative AI literacy deserves its own competency set holds up, and the paper is honest about its evidential basis. It is a useful conceptual contribution with clear limitations.\n\nThis paper is for educators, curriculum designers, and digital-government staff who need a practical checklist to build training or assessments. It deserves a serious referee. The right outcome is a revision that tempers the 'defining' language, perhaps to 'a proposed framework,' and engages with the possibility of parallel frameworks.\n\nI would bring it to a reading group as a useful case study in how conceptual papers handle the tension between ambition and evidence. Send it to peer review.\n\nBest,\n[Your name]","headline":"A transparent, useful competency framework that overstates its 'defining' claim but clearly deserves constructive peer review.","tokens_in":22303,"tokens_out":3108,"would_cite":false,"duration_ms":26841,"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 proposes that generative AI literacy consists of twelve defined competencies, from basic AI knowledge to prompt engineering and legal awareness, and argues this set fills a gap left by generic AI literacy frameworks.","keywords":["generative AI literacy","AI literacy","competency model","prompt engineering","AI ethics","AI legal aspects","continuous learning","AI competencies"],"falsifier":"A systematic search that locates a published generative AI literacy framework existing before or alongside this one, or an empirical study showing that a person can be fully competent with generative AI while lacking one of the twelve competencies, would overturn the claim that the set is defining.","tokens_in":21431,"feed_emoji":"🤖","tokens_out":6447,"duration_ms":52674,"temperature":0.7,"pith_summary":"This paper proposes a competency-based model of generative AI literacy built from twelve competencies, ranging from basic AI awareness through tool use, prompt engineering, and programming to contextual, ethical, legal, and continuous-learning skills. The authors argue that such a specific framework is needed because existing AI literacy frameworks are too generic and do not distinguish generative from predictive AI, even though generative tools raise distinct questions of authorship, hallucination, and authenticity. The model is put forward as a roadmap for individuals, educators, policymakers, and government bodies, and as a foundation for future assessments, curricula, and regulations. A sympathetic reader would take the paper's core contribution to be a shared vocabulary for naming and organizing what it means to be literate with generative AI.","feed_headline":"Twelve competencies define generative AI literacy","feed_subtitle":"An ordered model that educators, policymakers, and public bodies can use to teach and assess generative AI skills.","key_machinery":"The central object is the twelve-item competency model itself, an ordered list of knowledge, skills, abilities, and behaviors arranged as a learning path. The ordering is purposeful: foundational awareness comes first, then tool use and output assessment, then prompting and programming, then context, ethics, law, and finally continuous learning, which cuts across all the others. The model works as a checklist and roadmap: it lets individuals, educators, and institutions locate themselves on the spectrum from consumer to developer.","core_discovery":"The paper's central claim is that generative AI literacy is a distinct concept from general AI literacy and is fully delineated by the twelve competencies listed here. Existing frameworks, it argues, treat literacy as generic and fail to separate generative from predictive AI, whereas these competencies cover what users need to understand, assess, interact with, and create generative AI. The list is presented as definitive: these twelve, in this order, are the defining competencies.","pith_inferences":["The paper does not itself build the assessment instruments it calls for; a direct next step would be to turn each competency into testable tasks and rubrics.","The twelve-competency set could be treated as a layered extension of existing digital and data literacy frameworks, rather than a standalone construct, since many competencies (continuous learning, context, ethics) are not unique to generative AI.","The claimed logical progression is testable: one could compare learning outcomes for groups trained in the proposed order versus a different order to see whether the sequence matters."],"forward_implications":["Educators can use the twelve competencies as the backbone of curricula that move learners from passive consumers to interpreters and creators of generative AI.","Public institutions can benchmark employee skills against the model and use it to design training for generative AI adoption.","Assessment designers can derive tasks and metrics from the competencies, supporting certification, proficiency levels, and educational placement.","Policymakers can ground literacy guidelines and regulation in the competencies, including awareness of legal frameworks such as the AI Act."],"supporting_citations":[{"why":"Supplies the baseline definition of AI literacy and the 16-competency framework this model extends and differentiates.","marker":"[61]"},{"why":"Provides a systematic review of AI literacy that anchors the paper's background and gap argument.","marker":"[67]"},{"why":"Defines competencies as knowledge, skills, abilities, and behaviors, the structure the model adopts.","marker":"[70]"},{"why":"The AI Act appears as the legal and regulatory context that motivates literacy for compliance.","marker":"[32]"},{"why":"Gives the prompt-programming concept that competency 7 (prompt engineering) builds on.","marker":"[79]"},{"why":"Provides prompt-engineering principles and techniques that the paper treats as part of the literacy set.","marker":"[34]"},{"why":"Broad review of ChatGPT covering applications, limitations, ethics, and future scope, used to justify the need for specific literacy.","marker":"[78]"}],"fun_headline_variants":["Generative AI literacy unpacked: twelve defining competencies","The twelve skills that make you generative AI literate","A twelve-competency model for responsible generative AI use","Twelve ordered competencies for generative AI literacy","Twelve competencies: the roadmap to generative AI literacy"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that a small database search and a non-systematic review of the literature are enough to prove that no prior generative AI literacy framework exists and that these twelve competencies are the complete, non-redundant set.","fun_headline_variants_meta":{"raw":{"variants":["Generative AI literacy unpacked: twelve defining competencies","The twelve skills that make you generative AI literate","A twelve-competency model for responsible generative AI use","Twelve ordered competencies for generative AI literacy","Twelve competencies: the roadmap to generative AI literacy"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00084,"raw_usage":{"total_tokens":3552,"prompt_tokens":730,"completion_tokens":2822,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":346,"completion_tokens_details":{"reasoning_tokens":2748}},"tokens_in":346,"tokens_out":2822,"duration_ms":19642,"temperature":1.0,"reasoning_tokens":2748,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T05:50:46.064797+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A systematic search that locates a published generative AI literacy framework existing before or alongside this one, or an empirical study showing that a person can be fully competent with generative AI while lacking one of the twelve competencies, would overturn the claim that the set is defining.","supporting_citations":[],"review_version":1}