In this episode, we’re going to explore how artificial intelligence can be used to enhance critical thinking in the classroom through a modified version of the Socratic method. The AIcademia podcast is a weekly podcast helping educators like you leverage AI in your everyday practice. I’m your host, Andy Fisher. Thanks for joining me. 
As educators, we understand all too well that questioning should be at the heart of learning. The word “educare” itself means, of course, “to draw out,” and this is precisely what the best educators aspire to do —  we use carefully crafted enquiries to draw out knowledge, reasoning, and insight from our students. This kind of critical thinking, primed by good questions, is an essential competence, particularly in our increasingly AI-augmented world. While foundational subject knowledge still has its place, a curriculum which is dominated by the regurgitation of facts under timed conditions is rapidly becoming anachronistic. I think in order to prepare them for the future, we should be less concerned with what our pupils know and more interested in how they think and how they find out what they don’t yet know. To frame today’s discussion, I’ll be drawing on two key resources: The first is a book entitled ‘Essential Questions’ and is subtitled ‘Opening Doors to Student Understanding’. It was written by Jay McTighe and Grant Wiggins and was published by ASCD Books in 2013. The second resource is a downloadable PDF called Caitlin’s Cookbook which was published by Steve Birtles on behalf of Eton College in September 2023. It’s a guide on how to integrate conversational AI into teaching and learning and it includes a prompt for Socratic dialogue which helped me shape my questioning workflow which I’ll be sharing with you today. I’ve linked both resources in the show notes for those who want to take a look in more depth. One of the things I love about working on a podcast release each week, is that the process of researching and reflecting on the topics I’m covering invites me to take a good hard look at my own teaching practice and consider whether I am walking my talk. I know the importance of good questioning and I’m aware that well-framed questions don’t just happen but need to be planned, but am I doing it? If I’m honest, not as often as perhaps I should. As a practitioner well into my 3rd decade in the classroom, it is easy at times to lean a little too complacently on my experience instead. If I have a well-annotated copy of the text I’m teaching, some decent student resources, some individual, small and whole group activities in mind and a good sense of my students’ current subject knowledge, I can usually do a pretty decent job of progressing their understanding of the topic in hand. BUT, if I took just ten minutes or so and came up with some pre-planned open-ended questions which I could deploy at salient points in the lesson too, I suspect I would have an even better set of outcomes. And why? Because questions are the lifeblood of learning. As the great physicist Richard Feynman quipped: ‘I would rather have questions that can't be answered than answers that can't be questioned.’ or how about this from author Leo Babauta: ‘At the end of the day, the questions we ask of ourselves determine the type of people that we will become.’ Or this from Rapper Lil Wayne ‘ ‘I’m a gangster and gangsters don’t ask questions’ Well shame on you Lil Wayne - shame on you! Questions foster intellectual humility as we come to appreciate how little we actually know and how much more there is yet to learn. Questions help us recognise that there may not always be a single “right” answer. In fact the most interesting questions have many right answers, just as the most interesting mountains have many paths to their summit. Well shaped questions remind us that real learning takes effort—it is not a spectator sport. If we want to learn, we have to put in the mental reps and sets, wrestle with those questions and be comfortable with uncertainty while working towards an answer. Questions also invite us to develop resilience because we are invariably used to test our ideas in the crucible of public opinion - and if our opinions or hypotheses are found wanting, we will need to learn that such a critique need not be personal - it is our ideas that are being challenged - not our worth as people. When someone uses a well framed question to irrigate our ignorance, it might feel uncomfortable but growth often is - my goodness our world needs so desperately to re-engage with the practice of benevolent and spirited disagreement, don’t you think? And finally, questions immunise us from the nefarious intentions of those who would rather we didn’t think at all but instead just accepted their claims as truth. Politicians, advertisers, cult leaders and con artists are all threatened by students armed with sharp-tipped questions and an enquiring mind. There’s a reason Socrates was condemned to death for encouraging the youth of Athens to engage in dialectic discourse - thinking, questioning citizens are far more dangerous than placid regurgitators of rote learning. When we model good questioning, when we teach how to ask questions and when we place dialectic discourse at the heart of education, we empower our learners not just in our lessons but for life beyond the school gates. But here’s the challenge: we teachers are often stretched thin, we are managing students with varying levels of need and engagement. We are time poor and we’re tired. And so we find ourselves ‘covering the curriculum’ rather than uncovering the nuances of the topic under investigation. We find ourselves teaching to the exam, didactically ploughing through the content and then wondering why our pupils become passive and disengaged. Or perhaps we DO use enquiry based learning but the percentage of our class that engage with the process is limited. The same few hands shoot up when we raise a question and those keen beans, god bless them, carry the majority of their peers who are happy to go along for the ride. Recently, in an effort to address my neglect of planned questioning, I revisited some of the other questioning strategies I’ve found effective in the past. I’m personally not a fan of the hands down pause and pounce approach and so I’m generally shy away from that one. I accept that it is effective in engaging more of the class because no one knows if they might be next, but in my experience it can also induce anxiety and can humiliate rather than empower pupils. My go-to approach remains the think-pair-share strategy which works well and ticks many of the boxes - but even this can be gamed by pupils who seem deep in discussion but when I attune my teacher radar ears to their debate, the more animated pairs are rarely on task if they are given more than a couple of minutes and there’s no record of their conversation. I even dusted off my set of mini white boards and sourced 26 non-permanent pens to see if the current cohort might be less inclined to draw willies or florals doodles but no - this seemingly genetic predisposition is still alive and well! Now I know I’m laying it on a bit thick and most of my classes are actually pretty good at going along with sir’s strange ways but I all this enquiring about enquiry led me to look into ways in which AI might provide another tool in our questioning toolbox. This is somewhat against the tide of current use cases for conversational AI. Most people tend to use it as a Google substitute and in fact providers like Perplexity and Open AI actively promote their tools as a new way of tapping into the vast swathes of information held in their training data. And it’s fine to use it in this way but today, I’m proposing that we turn this approach on its head. Rather than ask a question and have a Large Language Model provide us with an answer, what if we task AI with asking our students questions and then have it evaluate their responses to provide a back and forth exchange of ideas. In other words, what if AI could serve as a 21st century Socrates? With this goal in mind, I started playing with prompt designs - I wanted something that was a little more directed than the one provided by Caitlin’s cookbook and after several iterations and tweaks I arrived at a 5 phase outline. I’ll provide the full prompt in the show notes but I’d like to talk you through the phases and then give an example of how this works in practice so you can decide if this framework would work for your subject and needs. You might begin the lesson by setting up the objectives on the board and have pairs of students log into a device and access the prompt which could be emailed to them ahead of time or posted on a shared online space such as Microsoft Teams or Google Classroom. They would then cut and paste the prompt into your Large Language model of choice and you would set a timer going to ensure that the activity is bounded. The prompt starts by giving the AI model some context - it identifies the topic under study, the academic level of the learner and the overview of the use case. It ensures that the questions will be asked one at a time and will incorporate the user input into its next exchange. Then we get into the 5 phases. Phase 1 gets the ball rolling with an open ended question which you have specified in the prompt. This ensures that the direction of travel is aligned with your lesson objectives. For example if studying Shakespeare’s play ‘The Tempest’ you might ask something quite focused like ‘is Prospero a good person?’ or you could lean more towards an essential question like ‘Does Jacobean theatre have anything to say to a 21st Century audience?’ The student pairs will collaborate and discuss their answer before inputting a thoughtful response. I like to specify an expected length for this answer - typically a short paragraph which they have planned in their exercise books. In response the AI model will provide a brief critique of the students’ answer and will then transition to phase 2, asking for some evidence or specific examples, in this case from the play, which supports their claims. The pair then use their copies of the text to find some quotations or moments from the text that can be used to justify their position and compose a second short paragraph which they provide to the model. The AI absorbs this new input, responds to it and then moves to phase 3 which involves asking the users to then compare or contrast their response to other aspects of the topic under study. Again the students will prepare and enter their ideas and the Large Language Model will provide a brief critique before moving to phase 4 which will involve a question to connect their ideas to the broader topic issues so they are invited to move from the specific to the wider implications. Finally, once their response has been digested by the AI, it will offer a commentary and then signpost that it will now adopt the role of devil’s advocate. I think this final phase is really important because it encourages divergent thinking and intellectual flexibility - even if the learners are convinced that their position is ‘right’ by adopting an opposing position it can provide fresh insights and a more nuanced understanding of the topic being studied. After the students’ enter their final response, the prompt triggers the AI to provide a succinct and useful summary of the discussion and the pupils can then either copy the entire conversation or just this summarise statement and email it to themselves and cc you as the teacher so that you have a record of the interactions. While all this is happening, the teacher is freed up to wander between pairs and oversee the process, commenting on the interactions and redirecting any pupils who are not on task. Once the time allocated to the task has expired, the teacher can then bring the group together for a plenary in which the class can discuss some of the most useful insights gained from the process. I also like to have a reliability check when we will identify any examples of hallucinations that came up - even if there were none this step reminds students that we need to remain vigilant when using AI tools and check the content for accuracy. 

If you’re using an AI model which is web-enabled, then you can also ask for citations and have the students go back and look at those sources. Alternatively, if you have the necessary AI literacy, you could set up a RAG system so that the AI model draws from a predetermined bank of materials in generating its responses and this can assure the veracity of the outputs. I think this questioning workflow and lesson outline has a number of advantages which I’m sure you’ve identified as I’ve walked you through the 5 phases. It is more likely that a higher percentage of your pupils will be actively engaged in critical thinking than if you were to use more traditional whole class questioning frameworks. Next, a record of the interactions is generated along the way which can be used as a starting point for an extended writing activity or it can be printed and pasted into the pupils’ workbooks. And the prompt can be modified so that it mines down into the minutae of the topic or it can be broadened so that the work being done is metacognitive - with a little imagination it can prove a very flexible learning tool. Let’s now look at a specific example of this workflow in practice. I’ve taken the example of a student studying ‘The Tempest’ at A Level who is investigating the character of Prospero and so my stem question which I pasted into Chat GPT 40 is the one mentioned earlier - ‘Is Prospero a good person?’ I typed this initial response into the model: ‘No I don’t think he is. While he might seem to be a figure who has endured exile and is doing his best to protect his daughter and survive on the island until he can make his way back to the mainland, I think he is a cruel and vindictive despot.’ Here is the LLMs response to my phase 1 answer: ‘Your perspective on Prospero as a cruel and vindictive despot is compelling. To deepen our analysis, could you provide specific examples from the text that illustrate his despotic behavior?’ This neatly moves us into phase 2 and I responded: ‘He inflicts unnecessary torture on Caliban, even though the island is rightfully his home and Prospero is a colonist. He imprisons Ariel to do his bidding and he treats his daughter as if she were an object to be traded. He wields his magic to serve his own ends and is perhaps even a narcissist. When he boasts of his ‘potent art’ it reminds me of 20th century dictators rather than a benevolent magician’. Chat GPT took these ideas and replied: ‘Your analysis highlights several instances where Prospero's actions can be interpreted as despotic and self-serving. For example, his enslavement and dehumanization of Caliban, whom he refers to as a "poisonous slave, got by the devil himself", and his control over Ariel, whom he threatens with imprisonment if he does not comply with his orders. Additionally, Prospero's orchestration of Miranda's relationship with Ferdinand can be seen as manipulative, as he uses their union to further his own political ambitions. As part of its response, the model provided two citations and did you notice how it also models the use of quotations to reinforce good practice? To transition to phase 3 it then states: ‘Considering these actions, how do you interpret Prospero's decision to renounce his magic at the end of the play? Does this act suggest a transformation in his character, or does it serve another purpose?’

This question invites a comparison and implies that there may be a contrast between Prospero’s behaviour at the start and end of the play. My answer was as follows: ‘Well, he might claim he will do so - he says he will break his staff and drown his magical books in the sea but we don’t actually see him do this. It’s easy for him to seem to be relinquishing his power but he is really just exchanging one kind of power for another now that he has secured a political marriage for his daughter and has re established himself as duke. I think Shakespeare may be using irony here to suggest that Prospero has not really changed at all - he can be seen as a Machiavellian figure.’ It responds: ‘Your interpretation of Prospero as a Machiavellian figure is insightful. Critics have noted that Prospero's strategic orchestration of events, such as manipulating Miranda and Ferdinand's relationship to secure political alliances, aligns with Machiavellian principles of power and statecraft. His treatment of Ariel and Caliban also may reflect his authoritative control over the island’s inhabitants. Prospero’s manipulation of Ferdinand and his plans to punish the shipwrecked visitors further illustrate his complex character. These actions can be interpreted as self-serving, highlighting his desire to reclaim his dukedom and secure a favorable future for Miranda. Considering Prospero’s multifaceted nature now, how do you interpret his final act of forgiveness towards his usurping brother Antonio and the others? Does this suggest a genuine transformation in his character, or could it be viewed as another strategic move to restore his position? So this question guides the learner into phase 4 - connecting us to the wider issue of Prospero’s character. Here’s my answer: ‘No I don’t think he’s changed - I think it is easy to be magnanimous when you’ve already got everything you want out of the interaction. We also know that Prospero will not always speak the truth or be transparent in his intentions..’ So you can see that I’m doubling down on my position that Prospero is a despot which I’ve found to be quite typical of many learners who are resistant to nuance and ambiguity and so gravitate to fixed readings of a text. So let’s see how the prompt first affirms my contribution and then encourages me to broaden my perspective and engage in genuinely critical thinking. This is why the final fifth phase is a useful shift in direction. The chatbot replies: ‘Your skepticism regarding Prospero’s apparent magnanimity is well-founded. Critics have debated the sincerity of his forgiveness, noting that he extends pardon only after regaining his dukedom and securing his objectives. For instance, in Act 5, Scene 1, Prospero confronts his brother Antonio, stating, “For you, most wicked sir, whom to call brother / Would even infect my mouth, I do forgive / Thy rankest fault; all of them; and require / My dukedom of thee, which perforce, I know, / Thou must restore.” This interaction suggests that Prospero’s forgiveness is conditional, tied to the restoration of his title. Moreover, some scholars argue that Prospero’s forgiveness is more coercive than genuine, as he orchestrates events to elicit repentance from his adversaries, thereby reinforcing his authority. Given these considerations, it’s plausible to interpret Prospero’s actions as self-serving, casting doubt on the authenticity of his transformation. Now, adopting the role of Devil’s Advocate, let’s consider an alternative perspective: Prospero’s decision to forgive his enemies and renounce his magical powers could be viewed as a genuine act of redemption and personal growth. By choosing virtue over vengeance, he may be demonstrating a sincere change of heart, embracing his humanity and the importance of mercy. How does this perspective influence your assessment of Prospero’s character? Here I play a receptive pupil who recognises that having an awareness of different readings of the text is one of the Assessment Objectives and therefore I reply: ‘Well given the play is a tragi-comedy, this more redemptive ending might make sense. After all, it is widely considered Shakespeare’s last play and a swan song to his career in theatre. Some even suggest that Prospero could represent Shakespeare himself as a kind of stage manager and director of the play’s events. With this in mind, I suppose celebrating forebearance and mercy could be a fair reading.’ The chatbot ends the interaction with the following commentary and then suggests some avenues for future reading and research: ‘Your observation aligns with a prevalent interpretation among scholars. The Tempest is often considered Shakespeare’s valedictory work, with Prospero’s character serving as a surrogate for the playwright himself. Prospero’s renunciation of magic in Act V, where he speaks of breaking his staff and drowning his book, is frequently seen as symbolic of Shakespeare’s own farewell to the theatre. This perspective is supported by the thematic elements of the play, which emphasize forgiveness and reconciliation, suggesting a contemplative and redemptive conclusion to Shakespeare’s literary career. Furthermore, the epilogue delivered by Prospero, in which he directly addresses the audience and seeks their applause to set him free, can be interpreted as Shakespeare’s personal request for acknowledgment and release from his life’s work. This metatheatrical moment blurs the lines between character and creator, reinforcing the notion of Prospero as Shakespeare’s self-representation. (Notice how here again the AI is modelling how to articulate complex ideas using concise and precise expression so as well as producing an engaged back and forth dialogue, this activity has some secondary benefits that weren’t prompted for but arise simply because these Large Language Models will employ a syntax and vocabulary suitable to the target learner. The response continues:) Considering this interpretation, Prospero’s actions throughout the play, including his ultimate choice to forgive his adversaries, can be viewed as reflective of Shakespeare’s own contemplation on themes of power, control, and the human capacity for mercy. This lends a nuanced complexity to Prospero’s character, suggesting that his journey mirrors a personal evolution towards empathy and understanding. To further enrich your analysis of Prospero’s character, you might consider exploring the following lines of inquiry: The Role of Colonialism: Examine how Prospero’s dominion over the island and its inhabitants reflects the dynamics of colonialism, and how this influences interpretations of his character. The Use of Power and Authority: Analyze how Prospero’s use of magic as a tool of control parallels political power structures, and what this reveals about his leadership and morality. The Theme of Illusion vs. Reality: Investigate how Prospero’s manipulation of events and characters through theatrical illusions serves to comment on the nature of reality and performance within the play. I know it took a little while to run through that example and if your not an English specialist, I apologise but perhaps you learned something about the Tempest along the way - but can you see how easily this prompt could be adapted to your own subject? You could start with a stem question like ‘what distinguishes music from noise?’, ‘what are the qualities of a real friend?’, ‘What is Science?’ or ‘Are we doomed to repeat history?’ 
I think this way of leveraging the power of AI could be a keystone to the revivification of schools. It offers the best of both worlds - we as teachers can craft and direct the shape of the dialogue but then the pupils will individually be expected to put in the intellectual labour and actively engage with the ideas presented. But don’t take my word for it - here’s the challenge. takeaway the prompt from the show notes, choose one class that you would like to field test it with and modify the prompt accordingly. Then run the activity and see for yourself whether it deserves a place in your classroom or your home if you are offering an alternative approach to schooling. 
Thanks for listening! I hope you’ve found some useful takeaways from today’s discussion. If you enjoyed the episode, please spread the word, and check out the AIcademia YouTube channel for practical tutorials that complement our podcast topics. Have a great week, and I look forward to catching up again soon!