{"id":"607fe802-3d48-432f-b8ab-e874e70d722b","arxiv_id":"2502.04940","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A qualitative study of 220 artworks and 12 artists produces a design taxonomy for artistic data visualization, showing that it centers on expression and concept rather than analytic efficiency.","lead":"This paper analyzes 220 data artworks and interviews 12 artists to build a taxonomy of how artistic data visualizations are designed and what they aim to do. It matters because it gives the visualization community a structured vocabulary for art-driven design and a set of research directions, while highlighting how art discourse shapes the field.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Sampling frame drives the central conclusion: 195/220 works come from IEEE VISAP, where acceptance and artist statements already presuppose an artistic framing; the claimed distinctive characteristics may be artifacts of that venue rather than of data art generally.","rationale":"I read the paper as a qualitative taxonomy study rather than a formal proof, and I credit it for assembling a substantial public corpus, providing data and code links, and triangulating the taxonomy with interviews. The internal coding process is described, and the authors do disclose the main sampling limitation in Section 7. My concern is not about internal inconsistency or authorial intent; it is about the inferential leap from a VISAP-heavy, explicitly tagged sample to an unqualified characterization of artistic data visualization as a whole. The reader's weakest-assumption statement identifies the same issue, and I agree with it. The concrete test would settle whether the concern lands: if a matched art-world sample produces similar taxonomy distributions, the central claim is substantially strengthened; if not, the conclusion must be restricted to the studied academic-art community. Given that the corpus skew is acknowledged but the headline claim is not correspondingly qualified, a conditional posture remains appropriate, so I would leave the reader's verdict unchanged rather than escalate or downgrade it.","tokens_in":21061,"tokens_out":5158,"duration_ms":53216,"concrete_test":"Construct a comparison corpus of 80–120 data artworks from non-academic art-world sources (e.g., Ars Electronica, Rhizome, major gallery exhibitions, and artist portfolios) using the same explicit 'data art' or 'artistic data visualization' tagging criterion. Have two independent coders apply the published codebook blind to venue, then compare the distributions of design intents and technique families against the VISAP-derived frequencies. If the VISAP and art-world distributions differ by more than roughly 15 percentage points on any headline category (e.g., experiment, physicality, criticize), the field-level claim should be narrowed to academic-venue data art.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 4.1 reports that the corpus consists of 195 works scraped from IEEE VISAP plus 25 snowballed from sources explicitly tagged with terms like 'data art'. Section 7 concedes that the collection 'primarily features works from or closely associated with the visualization academic community' and is not exhaustive. The abstract and conclusion, however, make an unqualified field-level claim: artistic data visualization is 'deeply rooted in art discourse' with distinctive 'inner pursuits and outer presentations'. The evidence for that claim is the taxonomy's frequency structure—e.g., experiment as the most frequent design intent and the prominence of sensation and physicality techniques—computed over this venue-defined corpus. VISAP is not a neutral sample: its review process already selects for works that can be framed as artistic data visualization, and the artist statements used for coding are written for that academic-art audience. Thus, the distinctive characteristics may be partly constituted by the inclusion rule rather than discovered in data art at large. The acknowledged limitation qualifies the conclusion, but the central assertion remains worded as a general property of the domain, and that is the load-bearing weakness.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents an empirical study of artistic data visualization. The authors analyze 220 data artworks (195 from IEEE VISAP and 25 from snowballed sources), code them for design paradigms, intents, and techniques, and interview twelve data artists. They propose a three-part design taxonomy and seven future research directions, and they make their corpus, raw datasets, and code publicly available. The central claim, stated in the abstract and conclusion, is that artistic data visualization is deeply rooted in art discourse and has distinctive characteristics in both inner pursuits and outer presentations.","tokens_in":21246,"tokens_out":8069,"duration_ms":75047,"significance":"This is a timely and useful synthesis of an under-theorized area. The corpus is large for a qualitative design analysis, the five newly identified design intents and the four technique categories are plausible and well illustrated, and the interviews add a practitioner perspective that cross-checks the taxonomy. The authors are transparent about corpus limitations and provide reproducible artifacts. If the central claim is substantiated after revision, the taxonomy and the seven lessons would be valuable to both visualization researchers and data artists. The main risk is that the strength of the conclusion currently exceeds the venue-defined evidence on which it rests.","major_comments":[{"comment":"The abstract and conclusion make a field-level claim that artistic data visualization is 'deeply rooted in art discourse' with 'distinctive characteristics in both inner pursuits and outer presentations.' However, Section 4.1 reports that 195 of the 220 corpus works come from IEEE VISAP, and the remaining 25 were snowballed from sources explicitly tagged as data art; Section 7 concedes that the corpus primarily features works from or closely associated with the visualization academic community and is not exhaustive. Because VISAP's review process and artist statements already presuppose an artistic framing, the frequency distributions (e.g., experiment as the top intent, the prominence of sensation and physicality) may be partly constituted by the inclusion rule rather than discovered in data art generally. Please either restrict the central claim to the academic/venue ecosystem or provide additional evidence—such as a broader non-VISAP sample or a sensitivity analysis—that the patterns generalize.","section":"Abstract, Section 4.1, Section 7"},{"comment":"The coding process is reported as reaching '100% agreement' after four rounds of meetings and discussion, but the paper does not report independent pre-reconciliation agreement, per-code reliability, or a metric such as Cohen's kappa. Since the frequency distributions of the taxonomy are the paper's main empirical result, this makes it difficult to assess coding robustness. Please report the initial agreement level and how disagreements were resolved; the same applies to the thematic analysis of the interviews in Section 5.1.","section":"Section 4.2.1, Section 5.1"},{"comment":"The 'design paradigm' dimension is coded only from 37 explicit mentions across the 220-work corpus, meaning that for the large majority of works no paradigm was identified. The paper nevertheless presents design paradigm as one of the three main taxonomy dimensions and later states that artistic data visualization is 'fundamentally shaped by high-level design paradigms.' This claim goes beyond what the data support for the works without such explicit mentions. Either code paradigms for the full corpus or present this material as a subtheme rather than as a full taxonomy dimension.","section":"Section 4.2.2, Section 4.3"},{"comment":"The text says that all ten prior intents from Lan et al. are present and that five new intents were identified, implying fifteen intent categories total. However, the enumeration that follows lists only fourteen categories with frequencies: experiment, inform, engage, re-present, provoke, criticize, equip, analyze, advocate, socialize, witness, archive, commemorate, and empower. Please reconcile this discrepancy by either adding the missing intent and its count or revising the description of how the prior and new intents combine.","section":"Section 4.2.3"},{"comment":"The claim that artistic data visualization has 'different distributions' of intents and 'distinct features' in techniques is made by comparing frequency observations with previously published taxonomies, but no direct comparison is reported in which the same codebook is applied to a comparable non-artistic visualization corpus. Without such a baseline, the distinctiveness claim is not fully supported. Please add a comparison coding exercise or soften the conclusion to describe the observed patterns within this corpus rather than differences from other domains.","section":"Section 4.3"}],"minor_comments":[{"comment":"The term 'aministic design' appears to be a typo; it should read 'animistic design.'","section":"Section 4.2.2"},{"comment":"The author affiliations contain stray spaces in 'Harbin Institute of T echnology' and 'Sun Y at-sen University'; these should be corrected.","section":"Author affiliations"},{"comment":"The quoted sentence from Willers reads 'the artistic approach is unlikely be appreciated'; it should be 'unlikely to be appreciated.'","section":"Section 2.2"},{"comment":"The notation 'P11*3 times (3 different works)' is unclear; please explain the convention for multiple works by the same participant or present the workflow counts in a small table.","section":"Section 5.2.1"},{"comment":"Reference [22] is cited in the text as 'Gates' but the reference is to Gates-Stuart et al.; the in-text citation should match the reference entry.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The paper is a solid qualitative contribution and a good fit for the journal's scope. The core issue is a scope mismatch: the field-level conclusion is stronger than the venue-defined corpus can support, and the coding reliability reporting is thin. Both are fixable through rephrasing or added evidence. I would not reject the paper; after the authors address the major comments, it could be a valuable reference for the artistic visualization community."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nWhat you should know: this is a qualitative HCI study that builds a design taxonomy for artistic data visualization from 220 artworks and 12 artist interviews. The genuinely new things are five design intents (re-present, criticize, equip, analyze, witness) beyond the authors' own prior affective visualization taxonomy, a four-category technique space (sensation, interaction, narrative, physicality), and a paradigm dimension. The interview findings—especially the four motivation modes and the output-driven vs input-driven workflow distinction—are a useful addition, and they cross-check the taxonomy reasonably well.\n\nThe paper does several things right. The corpus is large, the method is appropriate for the research question, and the authors are transparent about the main weakness: 195 of 220 works come from IEEE VISAP, and the rest were snowballed from explicitly tagged \"data art\" sources. They also flag the scarcity of non-Western and early works. The links to the corpus and code are real. The writing is clear.\n\nThe soft spots are the ones you'd expect. First, the coding reliability is reported as \"100% agreement\" achieved after four rounds of joint discussion, with no inter-rater metric. That's not a fatal flaw in a qualitative study, but it means the frequency distributions should be read as the authors' curated synthesis, not an independently verifiable measurement. Second, the abstract and conclusion make a field-level claim—\"artistic data visualization is deeply rooted in art discourse with its own distinctive characteristics\"—while the limitations say the corpus primarily features works from or closely associated with the visualization academic community. The stress-test note is right that the sampling frame may constitute part of the finding: VISAP acceptance already selects for works that can be framed as artistic data visualization, so the \"distinctive characteristics\" may be partly an artifact of the venue. The authors' own acknowledgment softens this, but it doesn't fully close the gap between the evidence and the unqualified phrasing.\n\nOn self-citation: the paper builds directly on Lan et al.'s affective visualization taxonomy and uses it as a starting codebook. That's legitimate, and the five new intents are drawn from the data, not forced.\n\nWho is this for? Visualization researchers interested in aesthetics, data physicalization, or the art-science boundary. It's a niche paper but a serious one. It deserves a proper peer review; the reviewer should push on the generalizing language and ask for a more explicit separation of corpus-derived claims from interview-derived claims.\n\nRecommendation: engage with it. Send it out. A good referee can help the authors align their claims with their evidence.","headline":"A mostly sound taxonomy-building study whose central generalizing claim is undercut by the VISAP-heavy corpus, but the limitations are acknowledged and the new constructs are real.","tokens_in":21823,"tokens_out":2238,"would_cite":false,"duration_ms":19763,"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":"Artistic data visualization is a distinct design practice rooted in contemporary art, not a decorative offshoot of analytics.","keywords":["artistic visualization","data art","visualization design","design taxonomy","design intents","design techniques","aesthetics","qualitative study"],"falsifier":"Build a new corpus from non-academic, non-Western, and pre-2010 data artworks, code it with the same three-dimension taxonomy, and compare the frequency distributions. If the dominant intent is no longer experiment and the four technique categories fail to recover the codes, the field-level claim would be refuted.","tokens_in":20805,"feed_emoji":"🎨","tokens_out":7757,"duration_ms":67729,"temperature":0.7,"pith_summary":"This paper tries to establish that artistic data visualization is a distinct design practice in its own right, deeply rooted in contemporary art discourse, and not merely a decorative or low-stakes version of analytics. The authors support this by analyzing 220 data artworks through open coding and by interviewing twelve practicing data artists, producing a design taxonomy with three dimensions: design paradigms, design intents, and design techniques. Their finding is that data artists consistently privilege concepts, subjective expression, and sensory richness over the readability and efficiency that dominate conventional visualization work. If the taxonomy is accepted, it supplies the visualization community with a shared vocabulary for studying art-driven data work and with seven concrete directions for future research.","feed_headline":"Data art has its own design logic, taxonomy of 220 works shows","feed_subtitle":"Data art prizes concepts and senses, not just readability.","key_machinery":"The design taxonomy is the central machinery of the paper. It organizes 220 artworks along three dimensions: design paradigms (high-level creative lenses such as generative art and speculative art), design intents (what the artist is trying to do, including experiment, re-present, and witness), and 32 design techniques grouped into sensation, interaction, narrative, and physicality. The taxonomy does the argumentative work by showing that art-driven visualization shares some techniques with narrative and affective visualization yet differs in its reliance on high-level paradigms, its openness to meaning, and its use of sound, smell, taste, and physical materials as encoding channels.","core_discovery":"The paper's central claim is that artistic data visualization has its own distinctive characteristics in both inner pursuits and outer presentations, such that it belongs more to art discourse than to the analytics-centered vocabulary of information visualization. Concretely, the corpus analysis finds that the most common design intent is experiment, challenging or reimagining conventional representation, followed by inform, engage, re-present, provoke, and criticize, and that the thirty-two design techniques cluster into sensation, interaction, narrative, and physicality. The interviews show artists often work in an output-driven mode, deciding the aesthetic form and concept first and then using data as a medium, and they justify lower precision by appealing to an alternative mechanism of communication that operates through immersion and open interpretation.","pith_inferences":["If the taxonomy is representative, the recurring boundary dispute between data art and data visualization could be recast as an empirical question about where works fall on the intent and technique dimensions rather than a matter of definition.","Because most of the corpus comes from one academic arts-program venue, the frequency rankings may describe that community better than commercial, non-Western, or early data art; a more diverse sample could test whether experiment and physicality remain dominant.","The witness and re-present intents suggest artistic data visualization can function as public testimony about data; a testable extension would compare long-term reflective engagement between witness-style and conventional data stories.","The output-driven workflow reported by artists, if formalized, could become a design pattern language for expressive visualization that inverts the usual data-first pipeline."],"forward_implications":["Aesthetics in visualization should be treated as broader than beauty, encompassing critical, deconstructive, and anti-beautiful stances that provoke reflection.","Evaluation metrics should move beyond immediate hedonic ratings toward context-rich, long-term measures such as interviews, diaries, and ethnographic observation.","Data artists' output-driven workflows suggest that tools built for them should support concept-first ideation and expressive experimentation, not just analytic pipelines.","Cross-modal encoding using sound, smell, taste, and physical materials offers concrete pathways for accessible and everyday visualization.","Artists can adopt structured design-study and rigor criteria from visualization research, while researchers can borrow artists' participatory, community-engaged deployment methods."],"supporting_citations":[{"why":"supplies the founding definition of artistic data visualization that the paper extends into a full design taxonomy.","marker":"[80]"},{"why":"provides the prior taxonomy of design intents against which the corpus was coded and extended with five new intents.","marker":"[43]"},{"why":"positions artistic and pragmatic visualization at opposite ends of a spectrum, motivating the need for a dedicated account of artistic design.","marker":"[39]"},{"why":"supplies the art-historical framework of mimesis, expression, and conceptual art that interprets the artists' stated intentions.","marker":"[57]"},{"why":"supplies the thematic-analysis method used to code the twelve artist interviews.","marker":"[8]"},{"why":"provides the history of contemporary art used to locate the identified design paradigms within art discourse.","marker":"[29]"},{"why":"offers a prior narrative-visualization design space that the corpus analysis compares against and differentiates from.","marker":"[71]"}],"fun_headline_variants":["Data art: experiment and sensation outrank readability","Artists use data as medium, not just to inform","220 artworks show data art has its own design logic","In data art, concept and sensation trump readability","Data art challenges norms, not just communicates data"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The taxonomy is meant to describe artistic data visualization at large, but most of the 220 works were drawn from a single academic venue's arts program, and the authors themselves note the corpus is not exhaustive and skews toward recent, Western, and explicitly tagged works.","fun_headline_variants_meta":{"raw":{"variants":["Data art: experiment and sensation outrank readability","Artists use data as medium, not just to inform","220 artworks show data art has its own design logic","In data art, concept and sensation trump readability","Data art challenges norms, not just communicates data"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000307,"raw_usage":{"total_tokens":1703,"prompt_tokens":837,"completion_tokens":866,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":453,"completion_tokens_details":{"reasoning_tokens":792}},"tokens_in":453,"tokens_out":866,"duration_ms":8922,"temperature":1.0,"reasoning_tokens":792,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T20:52:54.297704+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Build a new corpus from non-academic, non-Western, and pre-2010 data artworks, code it with the same three-dimension taxonomy, and compare the frequency distributions. If the dominant intent is no longer experiment and the four technique categories fail to recover the codes, the field-level claim would be refuted.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the founding definition of artistic data visualization that the paper extends into a full design taxonomy."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"positions artistic and pragmatic visualization at opposite ends of a spectrum, motivating the need for a dedicated account of artistic design."},{"cited_title":"Pooke and D","cited_arxiv_id":null,"evidence_quote":"supplies the art-historical framework of mimesis, expression, and conceptual art that interprets the artists' stated intentions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"provides the history of contemporary art used to locate the identified design paradigms within art discourse."}],"review_version":1}