arxiv: 2604.22998 · v1 · submitted 2026-04-24 · 📊 stat.OT

Recognition: unknown

Perceptions and Utilization of GenAI Tools among Data Science Students and Faculty

Abeer M. Hasan, Sayed A. Mostafa

Pith reviewed 2026-05-08 08:39 UTC · model grok-4.3

classification 📊 stat.OT

keywords generative AIdata science educationstudent perceptionsfaculty perceptionsAI literacysurveyChatGPTHBCU

0 comments

The pith

Survey of data science students and faculty finds heavy generative AI use paired with limited literacy and classroom integration.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper surveys how students and faculty in statistics and data science at one historically Black college or university view and apply generative AI tools. Students report frequent use of ChatGPT mainly for coding assistance and writing, along with positive outlooks on AI for their work and careers, yet they show low confidence in making sense of AI outputs and voice worries about accuracy, reliability, and over-dependence. Faculty members also see value in the tools but rate their own skills low and bring them into classes infrequently. Differences appear more by year in school than by gender. The results point to adoption running ahead of literacy and call for training, validation steps, and institutional rules to guide responsible use in data science education.

Core claim

Students in data science programs use generative AI tools extensively, primarily ChatGPT for coding assistance and writing support, and hold positive perceptions of AI in their field and future careers, yet they demonstrate limited confidence in interpreting AI-generated outputs and raise concerns about accuracy, reliability, and over-reliance. Faculty members also view these tools favorably but report low self-rated proficiency and infrequent use in teaching. These patterns vary more by academic level than by gender, highlighting a disconnect between high adoption rates and insufficient AI literacy that calls for structured training, validation practices, and institutional policies for safe

What carries the argument

Survey responses from 119 students and 14 faculty on familiarity, usage patterns, perceived benefits, awareness of limitations, and instructional support needs, with subgroup comparisons by academic level and gender.

Load-bearing premise

Self-reported answers from one institution capture actual usage, perceptions, and teaching behaviors without major social desirability or non-response bias.

What would settle it

A follow-up that tracks real tool usage logs or tests participants' ability to spot errors in AI-generated code and analysis would contradict the reported confidence levels if actual skills prove lower.

Figures

Figures reproduced from arXiv: 2604.22998 by Abeer M. Hasan, Sayed A. Mostafa.

**Figure 1.** Figure 1: Students’ self-reported usage frequency of AI tools in specific DS tasks. view at source ↗

**Figure 2.** Figure 2: Students’ perspective on AI’s potential to enhance the DS workflow. view at source ↗

**Figure 3.** Figure 3: Students’ awareness of the listed AI limitations in DS workflows. view at source ↗

**Figure 4.** Figure 4: Faculty rating of their ability to perform teaching tasks using AI tools. view at source ↗

**Figure 5.** Figure 5: Faculty self-reported frequency of using AI tools in teaching activities. view at source ↗

**Figure 6.** Figure 6: Faculty agreement on the potential for GenAI tools to improve DS education. view at source ↗

**Figure 7.** Figure 7: Faculty concerns regarding integrating GenAI tools into DS education. view at source ↗

**Figure 1.** Figure 1: Students’ self-reported frequency of using AI tools in their DS coursework. view at source ↗

**Figure 2.** Figure 2: Students’ common use cases for AI tools in DS tasks. view at source ↗

**Figure 3.** Figure 3: Students’ self-reported familiarity levels with AI technologies in DS. view at source ↗

**Figure 4.** Figure 4: AI tools students reported using in DS tasks. view at source ↗

**Figure 5.** Figure 5: Students’ perceptions toward integrating AI tools in DS. view at source ↗

**Figure 6.** Figure 6: Agreement with statements about the future of GenAI in DS. view at source ↗

**Figure 7.** Figure 7: Students’ primary concerns about using AI tools in DS. view at source ↗

**Figure 8.** Figure 8: Challenges faced by students when integrating AI tools into DS workflow. view at source ↗

**Figure 9.** Figure 9: Students’ concerns about AI use in education. view at source ↗

**Figure 10.** Figure 10: Students’ views on whether AI tools should complement or replace traditional meth view at source ↗

**Figure 11.** Figure 11: Faculty self-reported frequency of incorporating AI tools in teaching. view at source ↗

**Figure 12.** Figure 12: Faculty perceptions of the potential of AI tools to improve DS education. view at source ↗

**Figure 13.** Figure 13: Faculty perspectives: AI as complement vs. replacement in education. view at source ↗

**Figure 14.** Figure 14: Reported challenges related to AI use in teaching. view at source ↗

**Figure 15.** Figure 15: Faculty awareness of institutional AI guidelines. view at source ↗

**Figure 16.** Figure 16: Primary ethical concerns in AI use. Over-reliance on AI leading to less human oversight was cited by 84.6% of faculty; 15.4% cited bias in AI-driven evaluations. See Section S1.6 for further discussion. Q19. How important is it to maintain human oversight when using AI for assessments and student feedback? Response options: Very important; Important; Neutral; Not important view at source ↗

**Figure 17.** Figure 17: Faculty perspectives on human oversight in AI applications. view at source ↗

**Figure 18.** Figure 18: AI policy statements in course syllabi/ class. view at source ↗

**Figure 19.** Figure 19: Faculty confidence in using AI tools. Responses were distributed across all levels: 30.8% “Very Confident,” 15.4% “Somewhat Confident,” 30.8% “Neutral,” and 23.1% “Not Confident.” See Section S1.6 for further discussion. Q22. What types of training or support would you need to effectively integrate AI tools into your teaching? Select all that apply. Response options: More AI-related training or professio… view at source ↗

**Figure 20.** Figure 20: Faculty reported training needs to integrate AI into their curricula. view at source ↗

**Figure 21.** Figure 21: Do faculty believe their institution provides sufficient guidance and resources for using view at source ↗

read the original abstract

This study investigates perceptions and use of generative artificial intelligence (GenAI) tools among students and faculty in statistics and data science at a historically Black college or university. Survey data from 119 valid student responses and 14 faculty responses were used to examine familiarity, usage patterns, perceived benefits, awareness of limitations, and instructional support needs. Students reported substantial use of GenAI, with ChatGPT as the dominant tool, primarily for coding assistance and writing support. Although student perceptions of AI in data science workflows and careers were generally positive, confidence in interpreting AI-generated outputs was limited, and concerns about accuracy, reliability, and over-reliance were common. Faculty also viewed GenAI favorably, but self-rated proficiency and the frequency of classroom integration remained limited. Comparisons across student subgroups suggested that familiarity with GenAI and awareness of its limitations varied more by academic level than by gender. These findings highlight a gap between AI adoption and AI literacy and underscore the need for structured training, validation practices, and clearer institutional guidance for responsible AI integration in data science education.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

A small descriptive survey at one HBCU reports high student use of ChatGPT for coding and writing alongside low confidence in outputs, but the evidence for a clear adoption-literacy gap is limited by self-reports and sample size.

read the letter

This paper is a straightforward survey of GenAI perceptions and use in statistics and data science at a single HBCU. It collected 119 student responses and 14 faculty responses, finding that students lean heavily on ChatGPT for coding help and writing, rate their confidence in the results as moderate at best, and flag accuracy and over-reliance as top worries. Faculty see value in the tools but report low integration into classes and limited personal proficiency. Subgroup notes show differences more by year in school than by gender. That specific institutional and demographic slice is the main new piece; prior surveys cover broader student populations but not this HBCU data-science pairing in the same detail. The write-up presents the percentages and patterns clearly enough for a reader who wants baseline numbers from this setting. The soft spots sit mostly in the methods and the reach of the conclusions. The faculty group is only 14 people, which makes any statements about instructors thin. The abstract and available text give no response rate, no description of how the questionnaire was tested or piloted, and no checks for social-desirability or non-response bias. All data are self-reported, so the claimed gap between adoption and literacy rests on students saying they use the tools often but do not trust the outputs much. Without objective tasks or validation steps, that pattern could partly reflect how people answer surveys rather than a true skill deficit. Single-site data also caps how far the call for structured training and institutional guidance can travel. This work is mainly for faculty and administrators at similar colleges who are planning local AI workshops or curriculum tweaks. A reader hunting for general theory on GenAI literacy or large-scale evidence will not find it here. I would send it to peer review if the full methods section adds sampling details and questionnaire information; the core descriptive contribution is modest but real enough to justify referee time with revisions focused on limitations and generalizability.

Referee Report

4 major / 1 minor

Summary. This paper reports findings from a survey of 119 students and 14 faculty in statistics and data science at a single HBCU. It describes high GenAI adoption (ChatGPT dominant for coding/writing), generally positive perceptions of AI in workflows and careers, but limited student confidence in outputs and widespread concerns about accuracy/reliability/over-reliance. Faculty report favorable views yet limited self-rated proficiency and classroom integration. Subgroup patterns suggest greater variation by academic level than gender. The authors conclude there is a gap between adoption and literacy, calling for structured training, validation practices, and institutional guidance.

Significance. If the descriptive patterns hold under improved methodology, the work offers value by documenting AI perceptions in an underrepresented institutional setting (HBCU) and could usefully inform curriculum development in data science education. The explicit focus on an HBCU is a strength that adds diversity to the literature on responsible AI integration.

major comments (4)

[Methods] Methods section: the manuscript provides no response rate, total invitations distributed, or non-response analysis, which directly weakens claims about usage patterns and the inferred adoption-literacy gap (Abstract and Methods).
[Methods] Methods section: no details are given on questionnaire design, pilot testing, or validation (e.g., reliability metrics), so self-rated confidence and limitation-awareness items cannot robustly support the central claim of a literacy deficit.
[Results] Results section: faculty analyses rest on n=14, rendering statements about limited proficiency and classroom integration statistically fragile and limiting the force of recommendations for institutional guidance.
[Results] Results section: subgroup comparisons (academic level vs. gender) are presented descriptively without statistical tests, p-values, or effect sizes, so the assertion that variation is greater by academic level lacks evidential support.

minor comments (1)

[Abstract] Abstract: '119 valid student responses' is stated without defining validity criteria or describing any data-cleaning steps.

Simulated Author's Rebuttal

4 responses · 0 unresolved

We thank the referee for the detailed and constructive review. We address each major comment below and indicate the revisions planned for the next version of the manuscript.

read point-by-point responses

Referee: [Methods] Methods section: the manuscript provides no response rate, total invitations distributed, or non-response analysis, which directly weakens claims about usage patterns and the inferred adoption-literacy gap (Abstract and Methods).

Authors: We agree this information would improve assessment of sample representativeness. The survey was distributed anonymously through departmental email lists to all students and faculty in the relevant programs. Exact invitation counts were not logged, precluding calculation of a response rate or non-response analysis. In revision we will expand the Methods section to detail the recruitment approach, report the 119 valid student and 14 valid faculty responses obtained, and explicitly note the inability to compute a response rate as a limitation, thereby qualifying claims about adoption patterns. revision: partial
Referee: [Methods] Methods section: no details are given on questionnaire design, pilot testing, or validation (e.g., reliability metrics), so self-rated confidence and limitation-awareness items cannot robustly support the central claim of a literacy deficit.

Authors: We accept this critique. Questionnaire items were adapted from prior published surveys on AI and technology adoption in education; the instrument was reviewed by the research team and two additional faculty members for clarity and face validity but was not pilot-tested on a separate sample or subjected to formal reliability analysis. We will add a Methods subsection describing item sources, development process, and internal review, while noting the absence of formal validation as a limitation and discussing its implications for the literacy-deficit interpretation. revision: yes
Referee: [Results] Results section: faculty analyses rest on n=14, rendering statements about limited proficiency and classroom integration statistically fragile and limiting the force of recommendations for institutional guidance.

Authors: We concur that n=14 limits statistical robustness and generalizability. These responses represent a large fraction of the small faculty population in the targeted departments at this single institution. In revision we will frame all faculty results as purely descriptive, remove any language implying broader inference, and strengthen the limitations paragraph to caution against over-interpretation while qualifying institutional-guidance recommendations as preliminary and requiring confirmation in larger samples. revision: yes
Referee: [Results] Results section: subgroup comparisons (academic level vs. gender) are presented descriptively without statistical tests, p-values, or effect sizes, so the assertion that variation is greater by academic level lacks evidential support.

Authors: We agree that the comparative claim would be stronger with statistical support. Subgroup cell sizes are uneven and in some cases small, rendering standard parametric tests inappropriate. In the revised manuscript we will either conduct and report suitable non-parametric tests (e.g., chi-square) where assumptions can be met, or remove the explicit comparative assertion and present the patterns as descriptive observations only, accompanied by appropriate caveats. revision: partial

Circularity Check

0 steps flagged

No circularity in purely descriptive survey study

full rationale

The paper reports survey results on GenAI perceptions and usage among students and faculty with no equations, models, fitted parameters, predictions, or derivation steps of any kind. All claims derive directly from the collected self-reported responses without self-definitional loops, fitted-input predictions, load-bearing self-citations, imported uniqueness theorems, smuggled ansatzes, or renamings of prior results. As a self-contained empirical description with no mathematical chain to inspect, no circularity is present.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard survey assumptions rather than new parameters or entities.

axioms (1)

domain assumption Self-reported survey answers reflect respondents' actual perceptions and behaviors
Invoked implicitly when interpreting usage patterns and perceptions as factual.

pith-pipeline@v0.9.0 · 5480 in / 1117 out tokens · 29766 ms · 2026-05-08T08:39:47.148958+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

3 extracted references

[1]

Bauer E, Richters C, Pickal AJ, Klippert M, Sailer M, Stadler M (2025). Effects of AI-generated adaptive feedback on statistical skills and interest in statistics: A field experiment in higher education.British Journal of Educational Technology, 56: 1735–1757. Brynjolfsson E, Chandar B, Chen R (2025). Canaries in the coal mine? six facts about the recent ...

2025
[2]

Duah JE, McGivern P (2024). How generative artificial intelligence has blurred notions of authorial identity and academic norms in higher education, necessitating clear university usage policies.The International Journal of Information and Learning Technology, 41(2): 180–193. Glickman M, Yan J (2025). ASA members’ perspectives on the use of generative AI....

2024
[3]

Not at All Familiar

Prilop CN, Mah DK, Jacobsen LJ, Hansen RR, Weber KE, Hoya F (2025). Generative AI in teacher education: Educators’ perceptions of transformative potentials and the triadic nature of AI literacy explored through AI-enhanced methods.Computers and Education: Artificial Intelligence, 9: 100471. R Core Team (2024).R: A Language and Environment for Statistical ...

2025