The learning conversation
Learning takes place within the learning conversation. Broadly, there are two kinds of conversations: social conversations and private conversations.
All learning has a social context. A teacher asking questions and dealing with answers is a learning conversation. So are the conversations in group work. Even reading a book is a social conversation: readers are responding to the author’s voice.
The second learning conversation occurs inside the learners’ heads. It is their private learning conversations. Memorisation of new information and developing connections with previous learning are part of this learning conversation. This is essential if established learning is to take place.
Teachers plan their lessons around social learning conversations. In traditional teacher-centred classrooms, the selection of resources, and the sequencing and pacing of activities are under the teacher’s visible control. The use of AI in such lessons would be as a lesson episode, a component of a lesson.
For student-centred lessons, students are given more autonomy over the flow of the lesson and when and how AI would be used. The teacher’s role would be to facilitate this, ensuring that students use AI responsibly with critical awareness.
AI could have several features that strengthen inquiry-based teaching:
- the value of immediate accurate feedback
- the ability of students to control the pace of their learning
- the ability of students to address ideas that they find difficult by asking questions.
AI’s ability to simplify, translate and generate new questions is potentially significant.
In a Socratic dialogue, questions such as:
- What do you mean by that?
- How do you know?
- Can you give me an example?
- What are the consequences of that?
- What is the counterargument?
Students could be trained to use such questions as the basis for their responses to the suggestions made by AI.
An excellent example of this is the AI in Education post by Craig Donoghue: AI-Assisted Socratic Learning: Refining A-Level Physics Explanations. Here students are encouraged to engage in a dialogue with AI, asking for questions, giving responses and obtaining feedback. The teacher becomes a facilitator for the learning. The lesson shows “how AI can be a powerful ally in personalized learning”.
The article gives a sample prompt:
“I am a year 12 student of A level Physics in the UK. I am revising the topic of X, and am struggling with coherently explaining Phenomenon Y. If I give you the relevant sections from the specification I am studying, can you ask me questions that require explanations only on Phenomenon Y. I will then provide a written explanation myself. I would then like you to evaluate the strengths and weaknesses of my explanations and guide me using a Socratic questioning method towards improving my written responses. Before we begin, ask me any questions you need to fully understand what I’d like us to do.”
One advantage of doing this in a classroom is that teachers can refocus the lesson with some explicit instruction if the group is struggling or extend it further if the lesson is being successful.
The private learning conversation is where students actively engage with the lesson ideas, asking questions, memorising, questioning and testing their understanding. This often takes place after the lesson, in private or with small groups of friends.
It is here that a trained AI has the greatest potential to be of value. If it were stable enough to act as a “thinking companion” it could be of real value to the metacognitive development of students’ learning.
Examples are already starting to appear. Khanmigo offers AI support alongside its resources.
Khanmigo is using AI imaginatively, too. It enables students to “chat” with historical or fictional figures (from Elizabeth Bennet to Zeus!), which might be of value for project activities.
A key insight is that teachers remain responsible for ensuring that AI does not do the thinking for the student, either through lazy plagiarism or through some deeper psychological or emotional dependence.
Controls can be added to AI to prevent students circumventing the thinking by asking for the answer directly. This image from Khanmigo illustrates the type of controls that can be used.
Khanmigo responds by asking a question intended to get the student thinking.
AI can also respond to students’ questions, by:
- simplifying ideas
- organising ideas into mind maps
- generating and marking test questions
- generating quiz cards as retrieval practice
to promote the private learning conversation. Sooner or later, however, the student must actively commit ideas to memory. No amount of technology can take that final step.
|Q. Do you think students could benefit from having a “thinking companion”?
If this companion were to be AI, what characteristics would you like it to have to be a helpful “thinking companion”?
Updated 16/01/24 to include end of article question.