.png "kyle_cole"){kind=avatar}
Using an AI model like Ask Your Data is a great way to easily access your data and composing your prompts is the key to that. I know sometimes I wish I could just prompt “You know that thing I want to know about” but unfortunately we aren’t quite there yet with AI. Below I have listed my top four tips to get better results when using Ask Your Data.
Keep in mind what the AI was trained on:
Sometimes I don’t know the exact name of the field I want to use when forming my prompt. Like the other day I wanted to look at rubrics and assignments, but I didn’t know the technical term for when rubrics are attached to an assignment. In Canvas, the rubric is associated with the assignment, not attached, and that improved my response from the AI.I have found that if I review the Canvas namespace documentation I can identify the exact name of the data point I want to use.
Quantify what you are asking:
If I were to ask you what grade is passing, you might have a different answer then I do. You should always quantify what you are looking for. If you were to have a prompt of“ List all students that are passing” the AI would take its best guess at what you mean by failing. To get a much better result you could say “List all students who have a score less than 70”. Another example would be, using the term participation in your prompt. If you were to ask “Show me all students who have not attended in the last 14 days” and your definition of attendance is the student submitting an assignment, it would most likely not pull what you are looking for. You would want to break down the word attendance to your definition of it.
Be Specific:
When forming your prompt you want to be as specific as you can get. For example, if you want to ask about a certain sub-account, you would want to reference that sub-accounts exact name in the prompt. Or if you are asking for a specific time, keep it to the areas of Canvas where you can set a date. If you ask it “All courses in week 2 module…” it will write the query to look for modules that have week 2 in them. Instead, it would be better to say “2 weeks into the course start date”, this gives it the course start date to work off of. If you do need to start with a broad prompt like “show me all…” remember that you can continue the conversation and get specific as long as it remains in context of the original prompt. Which leads us into our next topic.
Chat Mechanic:
Last but not least, chats are contextual. If you were to start a chat talking about student grades then send a prompt about the last time a user sent an inbox message within the same chat, your results for the inbox prompt would be contextual to the first prompt. Just like when you talk to other people, try not to jump around subjects when talking about another subject and if you want to start a new topic start a new chat.
By following these guidelines, you can improve the accuracy and relevance of responses generated by Ask Your Data. If you have any questions or suggestions, please drop them in the comments below!
