In this episode I investigate the recent release of ‘deep research’ AI models and evaluate their suitability as tools for educators. I comment on the perceived pros and cons of these ‘agentic multi-step research tools’, summarise the outcome of spending a week trawling various articles and posts about the comparions between the models and describe three examples of reports I have created using Perplexity. Can ‘Deep research’ save teachers time? Is it the equivalent of ‘a PhD in your pocket’? Let’s dive in and find out!

00:00 Introduction to Deep Research AI Models

00:25 Evolution of Conversational AI

01:31 The New Wave: Web-Enabled Agentic Reasoning Models

02:05 Practical Applications in Education

03:36 Major Players in Deep Research AI

04:44 Advantages of Deep Research AI Tools

07:14 Criticisms and Limitations

10:27 Personal Experiences and Use Cases

14:54 Recommendations for Educators

17:10 Conclusion and Final Thoughts

In this episode of AI Academia, I explore how artificial intelligence can be used to foster critical thinking in education by employing a modified version of the Socratic method. I discuss the importance of questioning in learning, drawing on resources such as 'Essential Questions' by Jay Mat Teague and Grant Wiggins, and 'Caitlin's Cookbook' by Steve Bertles. I share insights on how AI can be leveraged to engage students in deeper analysis and dialogue, using prompts to guide back-and-forth exchanges of ideas with the AI. The episode includes a detailed example using Shakespeare's 'The Tempest' to illustrate the process. Finally I issue a challenge so you can put this prompt to work with your own learners!

(Essential Questions): https://tinyurl.com/mrytuyur

(Caitlin’s Cookbook): https://cirl.etoncollege.com/wp-content/uploads/sites/4/2024/01/Issue-1.pdf

The Prompt:

‘You are an expert mentor and questioner.

‘I’m studying [subject] at [level] and want to deepen my understanding of [topic].

Use the questioning method outlined below to help me test and refine my understanding.

The objective is to encourage me to engage in critical thinking, to allow me to arrive at a clearer idea of my point of view and to broaden my understanding of the topic under study’

You should only ask one question at a time and then wait for my response before moving onto the next phase of the question workflow.

Here is the workflow which has 5 phases:

#1 Start with the following open-ended question: ‘stem question’

#2 Once I have responded, give some feedback by critically assessing my answer and then ask a probing question which requires me to provide evidence or specifics to support my claims.

#3 Once I have provided plausible evidence or specifics, give some feedback by critically assessing my answer and then ask a question that encourages comparison or contrast with other relevant aspects of the topic.

#4 Once I have responded, give some feedback by critically assessing my answer and then ask a follow-up question that connects the line of enquiry to the wider topic focus.

#5 Once I have responded, give some feedback by critically assessing my answer and then state that you will now adopt the role of Devil’s Advocate and go on to propose a plausible contrarian position to the one I currently hold. Invite me to consider that position and either refute it or modify my response in the light of this alternative viewpoint.

Finally offer a summary of the whole conversation which precisely and concisely captures the material covered in all 5 phases of the discussion and suggest three new lines of enquiry that might complement the topic under investigation.’

00:00 Introduction to AI in Education

00:27 The Importance of Questioning in Learning

01:26 Key Resources for Enhancing Critical Thinking

02:22 Reflecting on Teaching Practices

06:39 Challenges in Modern Education

07:30 Revisiting Effective Questioning Strategies

09:11 Leveraging AI for Socratic Dialogue

10:18 Implementing AI-Driven Questioning Workflow

15:50 Example: Analyzing Prospero in The Tempest

25:48 Adapting AI Prompts for Various Subjects

26:32 Conclusion and Call to Action

In this week’s episode, I’m exploring the recent release of the Chinese large language model, DeepSeek R1. It’s an accessible overview of what makes this model so impressive, and why it has cost one US company 600 billion dollars in a single day! Along the way, I’ll be jargon-busting so you’ll have a better understanding of terms like ‘inference time compute’, and ‘open source models’.

00:00 Introduction to DeepSeek R1 and Its Impact on Education

00:56 Understanding DeepSeek R1: An Analogy

01:52 The Founding and Development of DeepSeek

02:18 What Makes DeepSeek R1 Unique?

03:57 DeepSeek R1 in Action: A Demonstration

06:23 Controversies and Questions Surrounding DeepSeek

09:44 Open Source vs. Closed Models

17:20 Global Implications and Reactions

24:02 Future Prospects and Final Thoughts