\u201cWhat makes you think that?\u201d This simple question has become a turning point for many of our students over the past two semesters, not because I asked it myself but because we deliberately integrated an AI tool into course assignments that would.<\/p>\n
As educators, we often talk about the importance of developing students\u2019 critical-thinking skills<\/a>. But fostering those skills at scale can be a challenge. Increasingly, instructors are turning to AI tools that engage students in structured, reflective dialogue. These systems are designed not to provide answers but to encourage learners to explain their reasoning, examine underlying assumptions and clarify their thought processes. \u00a0<\/p>\n Our research team at Georgia Tech\u2019s Center for 21st Century Universities has been piloting one such tool, Socratic Mind, which invites students into deeper intellectual engagement. While not developed solely to enhance critical thinking, the tool has helped many students think more deeply and express their ideas with greater clarity.<\/p>\n Centring instruction on \u2018Why\u2019<\/p>\n The path to deeper thinking lies not in more content but in better questions. At the heart of this approach is Socratic questioning, a method rooted in ancient philosophy. Rather than emphasising correct answers, it asks learners to reflect, probe assumptions and clarify their reasoning through dialogue.<\/p>\n When instruction is grounded in this method, it shifts from delivering information to cultivating enquiry. Prompts such as: \u201cCould you explain your thinking?\u201d or: \u201cWhat assumptions are you making?\u201d encourage students to refine their reasoning and revisit responses until they are ready to progress. In an introductory computing course, students analysed a short code. They explained key points, with follow-up questions from the AI encouraging them to consider alternative outcomes. In an applied physiology course, students reflected on the ethical implications of the attention economy and healthcare technologies, with the AI prompting them to challenge preconceived ideas and weigh trade-offs related to consent, privacy and autonomy.<\/p>\n These layered, ongoing prompts helped students move from simply recalling information to reasoning through complex issues. In our pilots, students accessed Socratic Mind through links provided in the course learning management system. Instructors created assignment questions that aligned with course materials and supplied the AI with potential answers, as well as common misconceptions, and embedded rubric criteria to gauge understanding. With this guidance, the tool could respond to students\u2019 inputs and generate follow\u2011up questions that nudged them towards more accurate reasoning within the instructor\u2019s intended framework. In this way, the AI complemented and extended the instructors\u2019 work rather than replaced it.<\/p>\n Using the Socratic method and AI in the classroom<\/p>\n Socratic questioning is a powerful teaching method, but it requires time, skilled facilitation and personalised attention, all of which can be scarce in large or online classes. This is where AI can help.<\/p>\n Rather than requiring custom-built solutions, instructors can now access a range of AI platforms that support this method. Tools such as Google Dialogflow, Botpress and open-source interfaces such as Chatbot UI (paired with models like GPT-4 or Claude) allow educators to create chatbots that prompt students with open-ended, reflective questions.<\/p>\n Through prompt engineering or branching conversation paths, these systems can simulate dialogues that encourage students to justify their reasoning, challenge assumptions and think critically. Even a simple prompt\u00a0such as: \u201cAlways ask a follow-up question that pushes the student to elaborate\u201d can transform a basic chatbot into a powerful thinking partner and help cultivate the habit of enquiry.<\/p>\n AI-implementation tips for university educatorsModel thinking habits, don\u2019t just deliver content<\/p>\n Use AI tools to guide students through their reasoning rather than just asking them to recall information. When prompted to explain their thinking, about 40 per cent of students in our pilot studies began questioning themselves more often and structuring their responses more logically.<\/p>\n Embed enquiry into feedback<\/p>\n Use AI to prompt curiosity rather than simply correct. The AI can offer follow\u2011up questions that encourage students to rethink and justify their ideas. For example, after a student explains a code snippet, it might ask: \u201cWhat would happen if a key element were changed?\u201d \u2013 pushing them to reason beyond a single solution.<\/p>\n Prompt metacognition through self-questioning<\/p>\n Encourage reflection with prompts for students to ask themselves: \u201cWhat am I assuming?\u201d or \u201cIs there a better way to do this?\u201d Over time, students in our pilots began using these questions independently, during and beyond assignments.\u00a0About 20 per cent specifically mentioned that they started questioning themselves more often, even outside the platform.\u00a0<\/p>\n Foster clear and structured articulation<\/p>\n Assign tasks that require students to explain ideas step by step, in writing or aloud. AI can support this with prompts to expand on partial answers, reorganise reasoning or clarify ambiguous points. This helps students connect concepts across contexts and communicate with greater clarity.\u00a0In our pilots, about 10 per cent of students specifically noted improvements in articulating their ideas and reasoning more clearly.<\/p>\n Probe misconceptions, not just correct answers<\/p>\n Design prompts and follow-up questions that guide students to think beyond the obvious and uncover assumptions or oversimplified views. For example, in a physiology course, students explored the ethical implications of emerging healthcare technologies. The AI\u2019s follow-up questions addressed the assumption that new tools were automatically beneficial, leading students to consider consent, privacy and autonomy.<\/p>\n Make reflection part of the process<\/p>\n After each session or assignment, ask students to reflect on their thinking experience. These reflections can be collected through an online form or survey with open\u2011ended or closed questions, giving students a structured space to share insights and suggestions that inform their learning and guide future improvements.<\/p>\n Let AI extend \u2013 not replace \u2013 educator dialogue<\/p>\n AI is a tool to extend, not eliminate, the human touch in teaching. It cannot replicate the depth of an instructor\u2019s feedback, but it can encourage students \u2013 even the quiet ones \u2013 to voice their reasoning, rethink their ideas and engage more deeply with the learning process.<\/p>\n Not every student in our pilot found the tool transformative, though. About a third reported no noticeable change in their thinking compared with courses without it \u2013 but many did. When intentionally designed and paired with structured reflection, AI-powered dialogue can encourage thoughtful engagement and support the internalisation of critical-thinking habits.<\/p>\n That doesn\u2019t mean AI tools are perfect. They can be rigid, repetitive or frustrating without the right framing. Ultimately, it wasn\u2019t about the tool itself. It was about the shift it provoked: from giving answers to justifying them, from following steps to questioning assumptions, from passive reception to active reasoning.<\/p>\n