5.2 Creating hints for your misconceptions

After your agents have established that the user has recently expressed a misconception, they will provide hints to the user in order to correct the misconception. These hints should focus on correcting a specific misconception. The hints are triggered after a misconception text is read by the selected agent. 

A hint is typically a question that will be asked by your selected agent, that aims at guiding the user to the correct answer. Ideally, your hints will guide the user to the correct answer without explicitly stating the answer. This page has some useful information about how to create a good hint.

To create a hint, type your question into both the spoke and display text boxes, located directly below the agent selection menu. If a student is unable to fully answer the expectation after receiving all of the hints you created, the student will be given a set of "prompts". 

You can also provide media when the hint is being asked. For more information about inserting media into your SKO,click here

Hint Answers

Each of your hint questions can have a set of answers, just like how each of your questions can have a set of answers, and each of your expectations can have a set of answers. Creating multiple answers to each hint allows the system to provide more accurate feedback for the student input. 

Name & Keys
  • 1. Name your answer. This is for your use. Name each one of your answers so you can better organize your tutoring pack. 
  • 2. Provide a semantic answer using regular expressions. This is what the AutoTutor Lite system will compare your student's input to. This is very similar to the semantic answers used in both the Self-Reflection assessment type and the Tutoring Assessment type. In the "Keys" box, you should include the essential words that would be needed to fully answer the question. Ideally, you will input these answer keys as regular expressions. The AutoTutor Lite system still takes advantage of LSA cosine similarity when comparing student input to answers, but the use of regular expressions helps make up for some of cosine similarity's shortcomings (i.e, differentiating between positive and negative answers). Because cosine similarity is still used to evaluate student input, the your SKO's semantic engine configuration is still important to consider. 
    • After a question is asked and the user provides input, that input will be compared to each of the keys for each of your hint answers. If the student input has a high overlap with a hint answer the user will leave misconception -> hint -> prompt sequence, and will move on to answer a different expectation. 
  • 3. Select which agent you would like to provide feedback for the student input. You can select either "Tutor" or "Student" as your agent. For hint answers, it is usually better for this hint answer text to be read by a tutor agent. 
  • 4. Select the "answer type". The drop down menu allows you to select an "answer type". The answer types that are available are based on the rules that have been pre-loaded into your ASAT SKO. 
  • 5. Select your answer threshold. This threshold determines what percentage of answer key coverage is necessary for the student input to be considered as answering the question. For example, if we set the threshold for the "Ideal Answer" to .9, then the student input would need to match (or have semantic overlap with) 90% of the answer key. The Tutor or Student agent (depending on who you selected) will then speak what is written in the Text & Speech tab. 
Text & Speech

The Text & Speech tab under "Answers" is where the spoken and written answer should be placed. This will be read by whoever you selected in the Agent drop down menu.