How do you plan to help your students understand, identify, and respond to AI hallucinations—including why they happen, how to reduce them, and how to evaluate AI-generated information critically?
This discussion question is from the Self-Paced Professional Learning for Teaching Problem Solving with AI.
While not a strategy, I would like to discuss the term hallucination. This term gives AI an personification…I think that is the right term at least 
The chatbot is really just trying to answer a question and matching data as close as it can. When it can’t it can get stuck and report false information, it didn’t ‘think’ that data up, it is just answering a question.
This method really shows the importance of double checking everything it reports to us, meaning we do not have to be an expert in the data it is giving us, but we must be a jack-of-all-trades to know a little bit to help call out a AI bluff.
I think more information helps the AI hallucination problem, but I think the larger problem is the students not realizing or understanding that the answers they receive from AI may be wrong. Do any of you give prompts that you know AI will be wrong and expect the students to catch it or research? Some of my students don’t even question calculator answers when we do accounting. They can’t comprehend that they make mistakes, how would they understand AI can be incorrect?
Some additional strategies I’d highlight for my students to help reduce AI hallucinations would be to help them develop neutral writing prompts and templates for common use cases:
-Start with broad research questions
-Request multiple viewpoints
-Ask for contrary evidence
-Verify all statistics independently
I think by providing the documents or links you want the AI to use to formulate a response can help with hallucinations. If you give and AI the whole internet…well…that is trouble
i think make feedback to ai answer and add some source
Well in addition to everything that has been said, it would be recommended to students to know that as well as we have a context to structure so do the machines working with the responses. Machine processing context is limited and very well may become a sort of hallucinations when trying to make up for responses.
Understanding this is a key for students, so that a very good strategy to follow in addition to composing a structured prompt would be to decompose AI inquiries into smaller problems.
Third, but not last, it is recommended to narrow-down inquisitions to certain sources or schemes of thought we provide to AI.
To help students reduce AI hallucinations, I would emphasize several strategies: verifying information with reliable sources, asking the AI to provide evidence or references, using clear and specific prompts, and comparing answers from multiple sources. I would also encourage students to question surprising claims, check dates and facts, and use their own critical thinking instead of accepting AI responses automatically. Finally, I would remind them that AI is a support tool, not an authority, so human judgment is always necessary.
Hola Bendiciones. Considero que siempre debemos advertir sobre el riesgo de las alucinaciones en la IA. Es importante orientar a nuestros estudiantes para que formulen preguntas claras y específicas, y que, al recibir una respuesta, soliciten las fuentes en las que se basa la información. Además, deben verificar que dichas fuentes sean confiables antes de dar por válida cualquier respuesta.