CoChat

Evaluating AI Courses: A Valid and Reliable Instrument for Assessing Artificial-Intelligence Learning through Comparative Self-Assessment

Open access

The DOI matches its Crossref record; 35 of the 35 references it cites were found in a registry or an index; 11 of the 26 sources its claims cite were checked for support.

Abstract

A growing number of courses seek to increase the basic artificial-intelligence skills (“AI literacy”) of their participants. At this time, there is no valid and reliable measurement tool that can be used to assess AI-learning gains. However, the existence of such a tool would be important to enable quality assurance and comparability. In this study, a validated AI-literacy-assessment instrument, the “scale for the assessment of non-experts’ AI literacy” (SNAIL) was adapted and used to evaluate an undergraduate AI course. We investigated whether the scale can be used to reliably evaluate AI courses and whether mediator variables, such as attitudes toward AI or participation in other AI courses, had an influence on learning gains. In addition to the traditional mean comparisons (i.e., t-tests), the comparative self-assessment (CSA) gain was calculated, which allowed for a more meaningful assessment of the increase in AI literacy. We found preliminary evidence that the adapted SNAIL questionnaire enables a valid evaluation of AI-learning gains. In particular, distinctions among different subconstructs and the differentiation constructs, such as attitudes toward AI, seem to be possible with the help of the SNAIL questionnaire.

Overview

CoChat Summary

TL;DR

The study adapted the SNAIL AI-literacy scale for evaluating an undergraduate AI course using retrospective/post comparative self-assessment. It found preliminary evidence that the adapted questionnaire validly and reliably assessed AI-learning gains, while attitudes toward AI and prior AI education showed limited or no relationship with learning gains.

Key findings

  • Across 30 AI-literacy items, most showed significant improvement; only three items failed statistical significance: explaining general vs. narrow AI, explaining data-privacy considerations, and identifying ethical issues. Cohen’s d was below 0.5 for five items, while 14 items showed strong effects (d > 0.8).
  • At factor level, significant retrospective-to-post differences were found for technical understanding, t(48) = 4.38, p < 0.001, d = 1.25; critical appraisal, t(48) = 3.47, p < 0.001, d = 1.21; and practical application, t(48) = 3.30, p < 0.001, d = 0.93. Mean CSA gain was 36.9% for technical understanding, 30.1% for critical appraisal, and 37.2% for practical application; the artificial-neural-network item had 56.8% CSA gain, while the ethics item had 15.6%.
  • Reliability was good to excellent: Cronbach’s α for retrospective assessments was 0.90 (technical understanding), 0.92 (critical appraisal), and 0.85 (practical application); post-assessment α was 0.83, 0.90, and 0.88, respectively.
  • Correlations between SNAIL factors and the two “Attitudes Toward Artificial Intelligence” factors (fear and acceptance) did not reach significance. Prior “other courses” and “other AI education” correlated with retrospective/post technical-understanding scores but not with CSA gain; e.g., other courses with TU retrospective ρ = 0.556, p < 0.01 and TU post ρ = 0.402, p = 0.046.

Methods

  • The evaluated course was an interdisciplinary undergraduate AI course with a technical focus on artificial neural networks, approximately 150 h total. Participants were 25 students (62.5% of all attendees), mean age 22.9 years (SD = 2.3), sixth semester on average (M = 6.0, SD = 2.9), with 16 men (64%) and 9 women (36%).
  • The SNAIL questionnaire was translated from English into German using independent forward and back translations plus an expert panel. Each of 31 SNAIL items was presented as a retrospective assessment and a post-assessment; one technical-understanding item was excluded due to a technical error, leaving 30 items (13 instead of 14 TU items). Participants also completed the five-item “Attitudes Toward Artificial Intelligence” scale and sociodemographic/prior-AI-education questions.
  • Analyses used independent t-tests at item and factor level, with Levene tests and a Welch test where homoscedasticity was violated; CSA gain was calculated; Spearman rank correlations were used for metric-ordinal relationships and Pearson correlations for metric variables; internal consistency was assessed with Cronbach’s alpha using Microsoft Excel or IBM SPSS Statistics version 27.

Limitations

  • Stated in the paper: only one course and a single sample were examined; some participants came from computer science backgrounds and were technically inclined, whereas the original SNAIL was aimed at non-experts; the selection of SNAIL was relatively arbitrary; and for evaluation efficiency, other adapted questionnaires could not be tested.
  • Evident from the text: the sample was small (n = 25) and formal enrollment was unavailable, with attendance estimated at about 40; the design relied on self-report and retrospective self-assessment; no retention or test-retest reliability was reported; and the course’s technical focus may limit generalizability to courses emphasizing ethical, legal, or social AI content.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3390/educsci13100978

Cites

35 of 35 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Mean values of retrospective assessment and post-self-assessment of all three factors.
Figure 1 Mean values of retrospective assessment and post-self-assessment of all three factors.From the PDF
Figure 2 Mean CSA gain values, in percentages, for all three factors. Note: CSA gain can reach values from -100% to +100%. Item TU11 was excluded, due to a technical error. Figure 2. Mean CSA gain values, in percentages, for all three factors. Note: CSA gain can reach values from -100% to +100%. Item TU11 was excluded, due to a technical error.
Figure 2 Mean CSA gain values, in percentages, for all three factors. Note: CSA gain can reach values from -100% to +100%. Item TU11 was excluded, due to a technical error. Figure 2. Mean CSA gain values, in percentages, for all three factors. Note: CSA gain can reach values from -100% to +100%. Item TU11 was excluded, due to a technical error.From the PDF

Verification

Claims the paper makes

7 Valid · 3 Not valid · 5 Not checked · 55 Own results and interpretations

27 more claims

Create a free account to see all 70 claims and every source verdict.

See all claims — free account
  • Probably not a substantial claim. 21
  • It may state two separate claims as one. 26

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Statistics recomputed

Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.

  • Consistentt(48) = 4.38reported p < .001 · recomputed p < .001

    “t (48) = 4.38, p < .001”

  • Consistentt(48) = 3.47reported p < .001 · recomputed p = .001

    “t (48) = 3.47, p < .001”

    The paper states a one-tailed test

  • Consistentt(48) = 3.3reported p < .001 · recomputed p = .002

    “t (48) = 3.30, p < .001”

    The paper states a one-tailed test

3 of 3

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

4 of 4

Citations

Cites

Research works printed but not linked: 2

  1. Cambridge Dictionary (2023, September 14). Artificial Intelligence. Available online: https://dictionary.cambridge.org/dictionary/english/artificial-intelligence.
  2. Harkness, J., Pennell, B.E., and Schoua-Glusberg, A. (2004). Methods for Testing and Evaluating Survey Questionnaires, John Wiley & Sons, Inc.

Cited by

See all 49

Metadata

Authors
Laupichler, M. C., Aster, A., Perschewski, J.-O., & Schleiss, J.
Affiliations
  • University Hospital Bonn (DE)
  • Otto-von-Guericke-Universität Magdeburg (DE)
Corresponding author
Laupichler, Matthias Carl
Venue
Education Sciences
Publisher
MDPI AG
Published
2023-09-26
Volume
13
Issue
10
Pages
978
ISSN
2227-7102 (ISSN-L)
OpenAlex
W4387048697
Cited by
48
References
30
Citations by year
2026: 19 · 2025: 18 · 2024: 11
FWCI
11.33
Type
journal article
Language
en
Open access
Locations
  • Education Sciences (publishedVersion, cc-by)
  • bonndoc (University of Bonn) (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Topics
  • Social Sciences › Education › Artificial Intelligence in Education
  • Social Sciences › Education › Digital literacy in education
  • Physical Sciences › Mathematics › Statistics and Probability › Statistics Education and Methodologies
Keywords
AI literacy, AI literacy assessment, learning gains, attitudes toward AI, measurement instrument validation, reliability statistics

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.