When we start to think about what conversations look like with a computer, we begin to think about what questions the computer can ask that will solicit useful answers, and will allow the momentum of the conversation to continue. Question and answer volleys will have to be considered and refined carefully. We want students to communicate more than simple recall answers. Designing the scope and style of questions will be a significant part of the design process.
Students will give rich answers to certain types of questions over others. Researchers have found that questions that are open--meaning that they don’t have a single word or short phrase answer--give richer and deeper answers. Questions framed in such a way that students are asked think about what they know invoke metacognition on the part of the students and produce answers that are more likely to make visible that which the student knows. Recognizing the kinds of questions that are most useful Will be important in efforts to create a chatbot that will elicit actionable information from learners. (Harlen, 2013)
Some questions about a phenomenon might include:
One strength of the chatbot construct is the ability to ask questions that will probe deeper or ask students to build on specific answers. If I chatbot is unable to identify or classify a particular idea, it might ask students to restate their concept, which will allow the students further opportunity to clarify and explain their own thinking. This is a built-in redundancy or fail-safe inherent to this kind of assessment. If the computer cannot understand what a student is trying to communicate, it may be that the student is not clear on the idea. On the other hand, when asking the student to clarify, the computer might also produce information from the student that would not have otherwise been given. Part of the conversation can be built such that the chat spot would ask follow-up and helping questions to encourage learners to go deeper.