THE SIGNAL IN ONE SENTENCE
Researchers surveyed 523 Indonesian university students about using generative AI for academic writing. Students generally reported high engagement and saw benefits for creativity and efficiency. The authors say 64.2 percent fell into a high category for AI-assisted creativity. Most also described their critical evaluation as moderate to high, but the item about recognizing misleading or incorrect AI output was a particular weak point. There is a catch large enough to write on the whiteboard: this was a questionnaire, not an exam in which students had to catch planted errors. The study tells us how students perceive their habits and abilities. It does not prove that AI made them more creative, damaged critical thinking, or caused any change in writing quality. Its most useful lesson is practical. A university can teach verification as a visible step in writing instead of hoping students perform it silently.
01
WHAT ACTUALLY CHANGED
SN Social Sciences published the peer-reviewed paper on September 12. Herri Mulyono, Gunawan Suryoputro, Ummu Salma Al Azizah, and Zain Adinul Falah report a quantitative cross-sectional survey of 523 Indonesian university students using a five-point questionnaire about attitudes, critical evaluation, and AI-assisted creativity in academic writing.
The paper reports high engagement with generative AI and perceived benefits for idea generation, creativity, and writing efficiency. It places 64.2 percent of participants in a high AI-assisted-creativity category. That number describes responses to a constructed survey scale. It is not a blind assessment showing that 64.2 percent produced more original or better writing.
Students generally reported moderate-to-high critical evaluation, yet item-level analysis found particular difficulty with recognizing potentially misleading or incorrect AI-generated information. The supplementary questionnaire shows what that means: students rated statements about checking accuracy, comparing output with other sources, questioning reasoning, correcting errors, and refusing to paste output before confirming it and adding their own thinking.
The researchers found that critical evaluation and creativity were statistically significant predictors of positive attitudes toward AI. In plain terms, students who reported stronger evaluation and creative-use habits also tended to feel more positively about the tools under the authors' model. A regression in a one-time survey does not establish that one of those factors caused the other.
The study was funded under a 2024 fundamental-research scheme from Indonesia's Ministry of Education, Culture, Research and Technology. The authors report ethics approval dated June 24, 2024, informed-consent planning, no competing interests, and no public release of the underlying data because of funder confidentiality requirements.
02
WHY THIS MATTERS
Fluent writing can hide a verification problem. Generative AI is good at producing orderly paragraphs, plausible transitions, and citations that look finished from across the room. A student may use the tool productively for ideas and structure while still missing an invented source, distorted statistic, or confident explanation. The better the surface, the easier it is to forget that every factual claim still needs a receipt.
A ban-only policy misses the behavior universities actually need to teach. Students will meet generative systems in workplaces, public services, research tools, and everyday software. Telling them never to touch the machine does not teach when to question it. Permission without practice is not enough either. The useful middle is supervised use with visible claim checking, source comparison, disclosure, and human revision.
The survey also exposes a measurement trap. Asking students whether they can recognize misinformation measures confidence and self-perception. It does not show whether they can detect a subtle fabricated citation under time pressure. Universities should pair attitude surveys with behavioral exercises containing known errors, ambiguous evidence, persuasive nonsense, and sources of different quality.
The Indonesian context matters beyond one country. Many students write academically in English as an additional language while navigating local subjects, regional sources, and global publication norms. AI can help lower the friction of drafting, but it may also flatten local voice or privilege easily retrieved English-language material. Verification therefore includes checking whether the system ignored relevant Indonesian evidence, names, law, language, or context.
Creativity should be measured as a process, not a cheerful output. AI can increase the number of options a student sees, help connect ideas, or make a topic easier to enter. The student still has to choose, reject, synthesize, and explain. An education system should reward the trail of judgment, including what the learner discarded and why, rather than treating the final polished page as the whole intellectual event.
03
WHERE IT COULD HELP
- Give students an AI claim ledger that lists each factual statement, the supporting direct source, the source date, the check performed, and the final revision
- Run short error hunts using AI passages with planted false citations, reversed findings, unsupported confidence, outdated facts, and locally wrong context
- Require a source swap in which students replace an AI-provided citation with the original paper, official record, dataset, interview, or other direct evidence
- Grade the reasoning trail alongside the final prose, including prompts used, options rejected, claims corrected, sources checked, and ideas the student contributed
- Teach a traffic-light review: green for supported claims, yellow for interpretation that needs context, and red for facts that remain unverified or should be removed
KEEP A HAND ON THE WHEEL
This was a cross-sectional self-report survey, not a randomized intervention, longitudinal study, writing-quality assessment, or behavioral misinformation test. It cannot establish that generative AI caused greater creativity, lower authenticity, weaker critical thinking, or more positive attitudes. The 64.2 percent figure comes from the authors' AI-assisted-creativity scale and should not be translated into 64.2 percent of students producing objectively creative work. The reported difficulty recognizing misleading or incorrect output is an item-level questionnaire result, not a measured error-catching rate. The abstract does not provide enough public detail to judge every sampling choice, response pattern, model assumption, or subgroup. The underlying participant data are not public because of funding-agency confidentiality requirements, limiting independent reanalysis. The questionnaire names ChatGPT as an example, but the findings should not be treated as a head-to-head evaluation of specific tools or model versions. The study was conducted by researchers at one Indonesian university and funded through an Indonesian government research scheme. Its findings are useful evidence about reported experience, not a verdict on every Indonesian student, university, discipline, language, or classroom.
04
TERMS WORTH KEEPING
OPEN GLOSSARY CARD
AI literacy
The ability to understand what AI systems do, question their outputs, use them responsibly, and recognize their limits and effects.
OPEN GLOSSARY CARD
Content provenance
Information attached to or embedded in media that helps identify where it came from and how it was created.
OPEN GLOSSARY CARD
Benchmark
A fixed test used to compare how systems perform on the same tasks.
SOURCES AND VERIFICATION STATUS
This article was written from the materials below. Product claims and dates were checked against those sources on September 13, 2026.
PUBLICATION RECEIPT: Revision 1. Published September 13, 2026.
THE PUBLICATION ENGINE
WANT A SIGNAL OF YOUR OWN?
We build source-grounded publications, private briefings, and editorial systems for organizations with something useful to say.
WORK WITH US