Multiple-choice questions are an attractive assessment method for online learning developers, principally because they can be marked automatically. In most professional development scenarios, they need to assess more than factual recall and instead present the learner with a scenario and ask them what they would do. They are used within this learning programme for both formative and summative assessments.
Formative assessments
The purpose of formative assessment is to give learners feedback, so it is necessary not only to create the scenarios, a range of responses and identify the best response but also to explain why. These are presented in this programme as slide presentations with answers and explanations following the scenarios and the possible responses.
Summative assessments
The purpose of summative assessment is to assess learners at the end of the programme, so it is necessary not to create the scenarios, a range of responses and identify the best response and then compare the learners’ responses with the best response. It is necessary to build a question bank from the individual questions and then present the learner with a random selection, to facilitate re-assessments and discourage learners from sharing the answers. This functionality is provided by many learning or content management systems, which may also generate certificates automatically for learners who correctly answer sufficient questions. In this case, you will need to generate the questions as a spreadsheet or comma comma-separated values text file.
Using AI to help you generate your scenarios and multiple-choice questions
AI-based tools can help considerably in generating questions for both summative and formative assessments.
At its most basic, you can write a prompt for your large language model asking to generate x questions about the topic you wish to be assessed.
However, considerable refinement and experimentation is required to create effective, meaningful and in-depth scenarios and questions.
From experience, ChatGPT has two common failings:
- The best response is nearly always the longest. This is a giveaway to human learners. Make sure you include an instruction to make the responses of equal length, but check the output and manually edit as necessary
- It often puts the best response in the same place, often response c, if the responses are labelled a to e. Again, make sure you include an instruction to distribute the best responses equally, but check the output and manually edit as necessary.
My experience is that Google Gemini is slightly better at this task, and if you first upload your content to LM Notebook, and then use the chat function within Notebook, it is easier to get more meaningful scenarios and detailed explanations. You can also specify the output within your prompt, asking for a table of scenarios with column headings tailored to create a comma-separated values file to import directly into a question bank for a summative assessment for example. For formative assessments, I copy and paste the material into a PowerPoint presentation and then export the images to create the slide presentations that you can see throughout this programme.
Again, the best thing to do is to experiment for yourself. Try the following: