Suggested Practices for Assessments 

 

1. Clarify what is authorised and unauthorised use

   • Have open and transparent discussions with students regarding the acceptable and unacceptable use of AI in assessments. This should include explicit instructions and expectations of what constitutes appropriate use, such as the production of original work. For example, you could allow the use of AI tools, if its use is acknowledged and described, and the text is paraphrased. For some assignments you can specify that the use is not allowed. 

   • Students should be informed that any unacceptable use of generative AI tools will be considered a violation of the SMU Code of Academic Integrity and will be dealt with accordingly. 

2. Detect unauthorised use where possible

   • Use generative AI tools such as ChatGPT or Bing to see examples of the answers it can generate for your assignments. This can help you identify patterns in how the responses are framed.

   • Due to the nature of some generative AI tools' set up, there are signs you can pay attention to in order to make an initial judgement on whether a piece of writing is AI-generated. For example, you can look out for:

     • the absence of personal experiences or emotions

     • inconsistency in writing style compared with in-class assessments

     • lack of in-text citation or wrong citations

   • You may explore AI-detection tools such as GPTZero and Originality.AI. Turnitin also recently introduced AI detection capabilities. It is worth noting these detection tools are work in progress and their results may be unreliable. More effort is also needed to put each student’s work through such scans before grading as automated ways of scanning batches of submissions are currently lacking. 

   • For more information on AI-detection tools, please click HERE.

   • Note: IITS is currently testing different detection tools. Further information and training will be provided in due course.  

3. Shift emphasis to in-class assessments

   • Consider converting some take-home assignments to in-class assessments without access to the internet. This may require breaking up larger assignments into component pieces. It can be suitable when the targeted learning outcomes relate to critical thinking, problem-solving, or written communication based working knowledge of one or more disciplines. For example, as part of a larger group project, you can organize an individual in-class assessment in which students are required to write a 500-word draft of a certain section of the report.

4. Redesign assessments beyond capabilities of AI tools

   • If appropriate, complement and vary different forms of assessment. For example, following up a written assignment with an oral presentation.

   • In assessment questions, include information that is not available on the internet and information that AI engines cannot gather. For example, use questions that require the use of include personal experience, cases/examples discussed in class.
   • Include the use of reflective responses that build in personal insight. AI tools are far less useful in this context. 
   • If appropriate, incorporate diagrams, graphs or pictures as part of the assessment questions (eLearn quizzes can facilitate this).
   • Introduce authentic assessment which involves tasks or activities that closely mirror real life challenges, requiring students to apply knowledge and skills in a practical or problem-solving context. For example, in certain law courses, students can be assessed using mock trials.
   • Try your assignment in generative AI tools so that you can find out how it fails to do well and then add grading criteria that measure exactly those qualities (such as relevance to context, specificity, authenticity).
   • Ask students to provide an annotated bibliography and describe how references were used in their assignment. 

5. Integrate Generative AI Tools into assessments

   • Consider using the generative AI tools as a supplement to other forms of assessment. For example, ask students to critique or build on an AI generated response.

Tips for Different Assessment Types

  

This table below shows how the above practices can be applied to existing assessment types:

Assessment Type	Tips
Take Home Written Assignment (e.g. Essays, Reports, Short Answer Questions)	Focus on knowledge application instead of the final product. Ask for a clear demonstration of the application of core concepts and ideas to a specific audience, or context. Place this inside an authentic problem that requires critical appraisal of data and research (like a detailed case study). For example, rather than asking for a summary of research into a topic, ask students to write a proposal that shows the application of the outcomes of the research to resolve a specific problem. Follow this with required reflections from students detailing how they adapted the research findings from the general to the specific. Focus on the process instead of the final product. For example, getting students to submit an outline and multiple drafts for an assignment and see how they develop their arguments. The meta-data for each draft should show their editing time, changes made, etc. If they lifted content from generative AI tools, the meta-data may show massive changes in a very short editing time. Assessment question design. The assessment question could include information that is not available on the internet, and information that AI engines cannot gather. For example, questions or responses that include personal experience, cases/examples discussed in class. You may also consider getting students to critique draft responses generated by AI tools. Complement and vary different types of assessment. For example, following up a written assignment with an oral presentation where clarifying questions can be asked
Class and Project Presentation	Content Focus the presentation on application of core concepts to an authentic context/problem. Students may be encouraged to relate concepts to local or personal experiences. Delivery Allow significant time after the presentation for a detailed question and answer session to allow the students to demonstrate their understanding of the topic (rather than just their ability to read from a script). Change the grading structure to weigh the answers to the Q&A more heavily.
Coding and Programming	Where appropriate assume that the code has been AI generated, so ask the students to do either or both of the following: Answer detailed questions on the specific choices made (For example, why a particular field would be needed in a database). Find and fix broken code under exam conditions. Explain why it was broken and what they would do to fix it, and why their fix would work.
Quantitative data-related assessment	Ask for demonstration of understanding of the data rather than the ability to generate them. For example, ask students to determine whether they believe that there is likely to be fraudulent behavior in the company as a result of their analysis of the figures.
In class assessments (midterm & final exams, quizzes, short assignments)	Consider making them closed-book and using eLearn quizzes with lock-down browser. If a degree of open book is needed, consider only allowing physical paper or books.

  

Resources:

This article recommends alternative assessments, including inviting students to use chatbots to produce writing they can critique.

• Gleason, Nancy (2022) “ChatGPT and the rise of AI writers: how should higher education respond?” Link to resource

This article shares practical advice on ensuring academic integrity

• Cotton, D., Cotton P. and Shipway, J. (2022) Chatting and Cheating: Ensuring academic integrity in the era of ChatGPT. Link to resource

The paper above contains useful recommendations to integrate ChatGPT into teaching.

• Susnjak, Teo. (2022). ChatGPT: The End of Online Exam Integrity?. Link to resource

A shared document constantly updated with AI-related resources – such as articles, journal papers, etc.Link to resource

Here's an example of it in action using different types of assessment: Essay, Lit Review, Short Answer Questions, Generating Recommendations, Case Studies and Reflection Questions

Suggested Practices for Teaching with Generative AI Tools

 

1. Teach students how to use Generative AI Tools well

   • Clearly communicate expectations and guidelines for using generative AI tools to students. This could include pointers such as the types of tasks the generative AI tool is appropriate for, what the limitations of the tool are, and how the AI-generated responses will be used to help student learning.

   • Ensure that the use of generative AI tools aligns with your overall assessment and learning goals. The use of AI tools should enhance assessment and learning experience, not detract from it.

   • Provide students with opportunities to reflect on their use of generative AI tools and how it relates to their learning. This could include prompts for self-reflection or small-group discussions.

   • Encouraging self-directed learning by allowing students to generate answers to questions they may have and verify the appropriateness of the answers through further research.

   • If appropriate, provide students with opportunities to learn about AI technology and how it works. This will help increase their awareness and appreciation of how AI is likely to shape the world of work they will be joining soon.

2. Use Generative AI Tools to improve your productivity as an instructors

   • Use it as a tool to assist you in creating educational content, such as writing practice exercises, creating interactive lessons, or creating rubrics for assessments you have created.

   • Generative AI tools can be used to generate personalized feedback for student work.

  

If you are keen to share your teaching or assessment practice on the use of ChatGPT, kindly contact cte [at] smu.edu.sg.   

The information here is current as of May 2023. Due to the evolving nature of emerging technologies and their impact, we will be periodically updating this page with new information and tips, so do check back.