You are here:

Spotting AI-Generated Scientific Abstracts

School of Science

About this use case

In this group-based assessment activity, students examine a selection of scientific abstracts. Some are genuine abstracts written by researchers and drawn from published journal articles, while others are generated using artificial intelligence on similar scientific topics.

Students determine which abstracts are human-written and which are AI-generated, then explain the reasoning behind their decisions. They present their conclusions to the educator and receive feedback on the accuracy and quality of their analysis.

The activity is designed to help students recognise potential misinformation, hallucination and other limitations in AI-generated scientific communication. Students do not use generative AI themselves during the activity; the educator prepares the AI-generated examples in advance.

Students compare a retail store operating in Malaysia with its counterpart in the retailer’s home country, such as Daiso in Malaysia and Japan or Lotus’s in Malaysia and Thailand. They investigate market entry, growth and competitive positioning strategies; examine cultural influences on marketing and consumer behaviour; and communicate their findings through a poster and video presentation. Together, these outputs contribute 30% of the total course grade.

Learning outcomes:

Adapt scientific practices and communication to respond to the rise of misinformation in a changing information landscape.

The activity was developed to help students become more critical of AI-generated scientific content, particularly where misinformation or hallucinated information may appear in scientific communication.
1. Select several genuine scientific abstracts from published journal articles.
2. Generate comparable abstracts on similar topics using a generative AI tool.
3. Ensure that the human-written and AI-generated examples are sufficiently similar to support meaningful comparison.
4. Prepare prompts that ask students to identify whether each abstract is human-written or AI-generated and explain their reasoning.
5. Plan feedback that highlights useful indicators, limitations and misconceptions.
- Students identify features that may distinguish AI-generated scientific writing from human-authored research abstracts.
- Students critically evaluate the credibility and quality of scientific content.
- Students explain and justify their decisions using evidence from the text.
- Students become more aware of misinformation and hallucination risks in AI-generated scientific communication.
- Students adapt their communication practices in response to a changing information environment.

Try it out

This use case requires an easy level of effort to implement.

- Before class, prepare a set of scientific abstracts that includes both genuine published examples and AI-generated examples on similar topics.
- During the activity, place students into groups and ask them to review each abstract. Students decide whether they believe it was written by a human or generated by AI and discuss the evidence supporting their decision.
- Each group then presents its conclusions to the educator. The educator provides feedback, explains relevant indicators and addresses any misconceptions.
- The activity can also be adapted into an interactive quiz. For example, show each abstract through a tool such as Slido and ask participants to vote yes or no on whether it was AI-generated. Discuss the result and reasoning after each response.
- The activity has worked well in both classroom and professional-development settings.
- Participants responded positively to the challenge of distinguishing between human-written and AI-generated content.
- A quiz format can make the activity easier to run with larger groups.
- Presenting one abstract at a time, collecting responses and discussing the answer immediately can support engagement and focused discussion.
Direct observation
This assessment uses Direct Observation paired with Verbal Explanations to evaluate learning in real time. Working in groups, students analyze pre-selected abstracts and explain their rationale to the educator before the correct answers are disclosed. Observing this deliberation live allows the educator to assess analytical reasoning as it happens, while keeping tool access restricted during the session ensures student engagement remains authentic.

Supporting resources

Here are some additional resources that you can browse to help you from TeachHQ:

Learn about AI at your own pace? Enrol in online courses via myDevelopment. Alternatively, you can join ongoing chats about AI and webinars.

Unit: SCI2010	Condition: Group
Class size: 250	Typology: Artefact
AI approach: Evaluating AI outputs	AI statement: Guided use
Visibility of Learning: Direct observation

Education Excellence