<aside> 📎

Manus is a powerful autonomous agent. For chemistry educators, this AI can automate content creation, research synthesis and bring it into the curriculum design. Nevertheless, before introducing Manus into the learning environment, we need to conduct a critical assessment of its limitations in education.

</aside>

Manus will experience hallucinations or make mistakes?

Yes! As AI inevitably generates hallucinations and makes mistakes. AI cannot fully encompass all human knowledge and culture during training period. Therefore, when the task exceeds the current representation capacity of AI or the problem is computationally difficult to handle, illusions will occur.

Example：

• Fact errors / fabricated information
• Misinterpretation of User Instructions
• Visual effects are incorrect with the facts
• Markdown Language

1. Dangerous operations in chemical experiment

In one of the Manus generated experimental demonstrations: a person placed a solid fuel block directly beneath a glass flask with an open flame. The personal protective equipment that laboratory personnel must wear, such as protective gloves or goggles, were placed on the table instead of being worn by the experimenters. Although the image appears professional and clean, it actually exposes multiple safety violations in the lab.

Misinterpretation of User Instructions

https://manus.im/share/toiOr1w32QpsonUojH1JVD?replay=1

This example of the visualization experiment highlights a key limitation of Manus' visual generation system: it creates "seemingly reasonable but actually inaccurate" visual effects. This neglects the accuracy of science and will pose teaching risks in the field of chemistry education.

1. Markdown Language

We encountered an issue when opening the file generated by Manus, the content appeared garbled because it was composed using Markdown syntax, which is not supported by the file format provided. This problem is particularly evident when chemical equations are included, as the Markdown formatting causes them to display incorrectly.

How to avoid misuse

When using Manus, you can guide it to generate the desired output by providing prompts that are as clear and detailed as possible. This approach leaves little space for the AI to interpret things by itself which can help reduce the likelihood of hallucinations.

In teaching:

• Set up classroom reflection and discussion sessions to guide students in analyzing the rationality and risks of Manus outputs.
• Require students to compare the key chemical knowledge output by Manus with textbooks, teaching aids, or scientific databases.
• Remind students: When using Manus, they should enter the task with a clear learning objective and knowledge judgment before starting, and maintain a critical mindset. Manus is a "tool to assist your thinking".

In any case, the works generated by AI are drafts. Their output is always based on human knowledge and judgment.

"educational research must prioritize developing learners' Al literacy and evaluative judgement, empowering them to thoughtfully navigate interactions with intelligent agents" (Ba et al., 2025, p. 1681)"

Therefore, educators need to help students build a strong foundation of chemical knowledge first ensure they have the ability to make evaluative judgments of AI. And then guide students to maintain a questioning, verifying and reflective attitude towards these drafts during the process of collaborating with AI. Let learning become a process where humans and technology jointly generate knowledge, rather than passively accepting the results.

Reference:

Yan, L., Pammer‐Schindler, V., Mills, C., Nguyen, A., & Gašević, D. (2025). Beyond efficiency: Empirical insights on generative AI ’s impact on cognition, metacognition and epistemic agency in learning. British Journal of Educational Technology, 56(5), 1675–1685