Skip to main content

30 April 2024

AI to enhance assessment and integrity. AI TC Meeting

Becci Peters profile image
Written by

Becci Peters

This is a summary of a TC meeting from Jonathan O'Donnell, AI Lead and Hazel Hatch, Deputy Lead for Assessment, both from the Harris Federation of schools.

Using AI Tools Thoughtfully in the Classroom: Opportunities and Cautions for Teachers

AI tools like ChatGPT and Copilot can be valuable assistants for teachers, but they require careful implementation to ensure high-quality, accurate outputs. The key is to use these tools thoughtfully, with appropriate safeguards and oversight, rather than relying on them blindly.

Prompt Engineering is Essential

Jon emphasised the importance of "prompt engineering" - providing the AI system with very detailed instructions, context, and specifications to get the desired output. This includes:

  • Ensuring the AI does not use outdated or inaccurate information, and thoroughly checking the outputs
  • Being cautious about using AI to generate lesson plans or assessments without proper oversight, as the results may contain errors
  • Providing the AI with clear instructions, roles/personas, reference materials, and step-by-step prompts to improve the quality of the outputs
  • Avoiding uploading copyrighted or student-specific data into the AI systems
  • Using the AI as an assistive tool, not a replacement, and being encouraging in feedback to students

Leveraging AI for Identifying Misconceptions

Jon notes that AI tools can be useful for surfacing common student misconceptions, such as confusing physical network layouts with logical topologies. This can help teachers better understand and address knowledge gaps.

Limitations of Generic AI Tools

While "off-the-shelf" generic AI question and lesson planning tools are easy to use, they have significant limitations:

  • They often use outdated language models prone to hallucinations and inaccuracies
  • They lack the specificity required for accurate question generation and marking based on exam board criteria and student age/level
  • They do not effectively teach teachers how to use AI tools thoughtfully and mitigate the risks

Subject-Specific AI Tools Show Promise

More specialised AI tools like Smart Revise, trained on subject-specific data, can save teachers time on marking and identifying student understanding. However, they still have limitations, especially for math and coding questions.

Key Considerations for Teachers

When using any AI tools in the classroom, teachers should:

  • Be very specific with prompts to get the best results
  • Train the AI models with high-quality examples beforehand
  • Treat the AI as an assistant, not a replacement, and work to refine the outputs
  • Carefully review AI-generated content for accuracy, hallucinations, and inappropriate language
  • Adhere to terms of service and age restrictions for student use
  • Avoid uploading copyrighted or student-specific data into the systems

See the slides

Watch the recording

Join the AI community