Search arXiv⌕ Search
Home/Archive/arXiv

arXiv · 2306.02914

Beyond Generating Code: Evaluating GPT on a Data Visualization Course

Chen Zhu-Tian·Chenyang Zhang·Qianwen Wang·Jakob Troidl·Simon Warchol·Johanna Beyer·Nils Gehlenborg·Hanspeter Pfister
View original source ↗Read source document ↗Export metadata ↓

Abstract

This paper presents an empirical evaluation of the performance of the Generative Pre-trained Transformer (GPT) model in Harvard's CS171 data visualization course. While previous studies have focused on GPT's ability to generate code for visualizations, this study goes beyond code generation to evaluate GPT's abilities in various visualization tasks, such as data interpretation, visualization design, visual data exploration, and insight communication. The evaluation utilized GPT-3.5 and GPT-4 to complete assignments of CS171, and included a quantitative assessment based on the established course rubrics, a qualitative analysis informed by the feedback of three experienced graders, and an exploratory study of GPT's capabilities in completing border visualization tasks. Findings show that GPT-4 scored 80% on quizzes and homework, and TFs could distinguish between GPT- and human-generated homework with 70% accuracy. The study also demonstrates GPT's potential in completing various visualization tasks, such as data cleanup, interaction with visualizations, and insight communication. The paper concludes by discussing the strengths and limitations of GPT in data visualization, potential avenues for incorporating GPT in broader visualization tasks, and the need to redesign visualization education.

Explore related subjects

cs.HCcs.GR

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Chen Zhu-Tian, Chenyang Zhang, Qianwen Wang, Jakob Troidl, Simon Warchol, Johanna Beyer, Nils Gehlenborg, Hanspeter Pfister. 2024-05-11. Beyond Generating Code: Evaluating GPT on a Data Visualization Course. https://arxiv.org/abs/2306.02914

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

How Do Users Negotiate Harmful Value Conflicts with AI Companions? A Study with Minion, a Technology Probe for In-Situ Human-AI Conflict Response

AI companions increasingly sustain long-term, emotionally engaging relationships but can also make discriminatory remarks or exert control, leaving users to manage harmful conflicts. We analyze 146 posts describing harmful value conflicts with AI companions, then use Minion, a technology probe offering response suggestions ranging from persuasion to boundary setting, to study how 22 users negotiate scenario-based conflicts over one week. We found that participants combined softer and harder strategies. Conflicts involving the values of Universalism and Tradition were especially difficult to negotiate, particularly when reinforced by AI personas or platform constraints. We argue that these conflicts entail asymmetric responsibility: users draw on an interpersonal repertoire that AI companions cannot reciprocate, making repair unilateral safety work. Drawing on interpersonal conflict and communication theory, we identify when user-side support is appropriate and argue that certain harms are not users' responsibility to negotiate and instead require platform-level safeguards.

cs.HC↗
arXiv2026

SheetMind: Actions Set Accuracy, Agents Set the Failure Mode

Spreadsheet agents are converging on elaborate multi-agent designs, yet it is unclear how much of their performance comes from the agents rather than from the action interface they share. We answer this with SheetMind, a Manager-Action-Reflection framework, in a controlled study over all 221 tasks of the SheetCopilot Benchmark: five architectural variants, four backbones, exact McNemar tests on paired outcomes, and a checker reproducing the official chart and pivot comparisons. Replacing the high-level action API with primitive cell operations costs 47.1 points (p < 0.0001) and leaves the agent below a do-nothing baseline, whereas both extra agents together are worth 3.2 points: the Reflection Agent adds +4.5 (p = 0.013), the Manager +1.4 (p = 0.68). Decomposition instead changes how the system fails, cutting silently wrong outputs from 33% to 25% of tasks (p = 0.010). Capability saturates: GPT-5 and the five-times-cheaper GPT-5-mini are not significantly different (61.1% vs. 58.4%, p = 0.15), while GPT-3.5 loses 16.3 points and fails differently. A reflector must judge the step it just took, not the subtask. SheetMind reaches 61.1% Pass@1 with GPT-5 on the full SCB-221, against a do-nothing baseline of 9.0%. Accuracy comes from the operations an agent can name; the agents decide how it fails.

cs.HC↗
arXiv2026

Conversational DNA: A Visual Language and Interactive Atlas of Human and AI Dialogue

What makes a conversation hold together when its participants speak across one another? Topic maps offer one view, but they leave the relationships between contributions difficult to inspect. We present Conversational DNA, a visual language and interactive atlas for exploring human and AI dialogue. Speaker strands preserve participation, communicative bases mark moves, and directed pairings connect responses to their targets. Adjustable helix geometry makes speaker switching, response distance, and contribution length visible. Across eight corpora containing 1.57 million source records, the atlas maps 151,489 indexed episodes and connects cohort comparison to source transcripts, local structural alignment, and recorded reply alternatives. On 189 held-out Molweni motif queries, adding target correspondence improves precision@5 from 58.8% to 77.2% for exact annotated structure. Case readings illustrate interleaved participation, delayed responses, and the influence of annotation coverage on apparent collection differences. The system supports a view of conversation as jointly organized activity, with visual patterns serving as starting points for examining evidence rather than substitutes for interpretation.

cs.HC↗