In recent years, the discourse surrounding artificial intelligence, particularly large language models (LLMs), has become increasingly polarized. On one side, we hear breathless proclamations of a utopian future where AI will solve all our writing and communication challenges. On the other, dire warnings of a dystopian nightmare where critical thinking withers and human creativity is supplanted by soulless machines.
This dichotomous debate – humans versus machines, utopia versus dystopia – often dominates discussions about AI in writing and education. We're bombarded with extreme scenarios: AI tools that promise to make our writing stronger and clearer, or pessimistic forecasts of destroyed cognitive abilities and atrophied critical thinking skills.
But amongst this noise, we risk losing sight of a crucial aspect of engaging with text: the art of close reading. As we grapple with the implications of AI and automation in writing, it's essential to refocus on this fundamental skill and explore how it evolves in the context of LLMs.
Close Reading with LLMs: A New Frontier
While much of the AI discourse focuses on broad, often speculative implications, one of the most significant and immediate applications of LLMs, particularly in education, is their potential to support and enhance the reading of texts. This use case has been developing for some time, but it's now entering the mainstream with tools like NotebookLM.
NotebookLM represents a shift from earlier AI applications. Rather than simply providing pre-trained responses, it leverages the power of language models to interrogate texts in context. This approach marks a significant departure from merely asking a model to regurgitate information from its training data.
With NotebookLM, users can input specific texts into the system's context. The language model then uses its capabilities to analyze and interact with this text, generating prompts and responses based on the provided content. This method allows for a more nuanced and context-aware exploration of the material.
To understand the significance of this development, it's helpful to revisit some of my earlier work on close reading in the digital age. Previous explorations on unlocking AI text processing capabilities and How to Chat with Texts—Close Reading with LLMs offer relevant insights. However, the integration of LLMs into this process adds new dimensions to consider.
The Three-Part Model: A New Paradigm for AI-Assisted Close Reading
As we look deeper into the intersection of close reading and LLMs, it's crucial to understand a new paradigm that's emerging: what I will call the three-part model. This model provides a more comprehensive framework for understanding our interactions with AI in the context of textual analysis.
Let’s breakdown the components:
The User
The first part of this model is the user - the human being who approaches the AI tool with intention and purpose. This user is not a passive recipient of information, but an active participant who:
Has specific intentions and goals
Decides to augment their thinking with the AI tool
Brings their own knowledge, experience, and critical thinking skills to the interaction
Uses the AI as a cognitive tool, much like we've historically used books, notes, or other aids to extend our mental capabilities
The Model
The second component is the AI model itself. This could be any large language model such as GPT-4, Claude, PaLM, or others. Key aspects of the model include:
Its underlying architecture and training
Its capabilities and limitations
The interface through which we access it (e.g., ChatGPT, NotebookLM, etc.)
The Resource
The third and perhaps most crucial part of this model is the resource - typically a text or document that serves as the focus of our close reading. This could be:
A short email or memo
A multi-page academic paper
A book or collection of books
A set of research papers or articles on a specific topic
Visual resources in the case of multimodal models
The Importance of the Three-Part Model
This three-part model represents a significant shift from the traditional model of human-AI interaction. In the that model, we typically focus on the user providing a prompt or question to the AI, and the AI generating a response. This approach has been the subject of much discussion, particularly around issues of accuracy and the potential for hallucinations.
The introduction of the resource as a third component fundamentally changes this dynamic. Now, instead of relying solely on the AI's pre-trained knowledge, we're asking it to engage directly with a specific text or set of texts. This approach:
Grounds the AI's responses in a specific context
Allows for more nuanced and accurate analysis
Enables the user to verify and cross-reference the AI's interpretations
Facilitates a deeper, more interactive form of close reading
In practice, this might involve uploading a research paper to NotebookLM, then asking the AI to analyze specific sections, compare arguments, or highlight key themes. The user can then engage with both the original text and the AI's analysis, leading to a richer understanding of the material.
This three-part model aligns closely with the goals of close reading, encouraging careful examination of the text while leveraging the computational power of AI to enhance our analytical capabilities. It represents a symbiosis of human critical thinking and AI processing power, with the text at the center of this collaborative analysis.
So what?
As we conclude our look of close reading in the age of LLMs, let’s consider the implications for educational leaders. Here are key takeaways and action points:
Embrace the Paradigm Shift: Understand that AI is not replacing critical thinking or close reading skills, but rather augmenting them. Encourage a balanced approach that leverages AI tools while maintaining focus on core analytical skills.
Promote Digital Literacy: Ensure that both educators and students understand how to effectively use LLMs in the context of the three-part model. This includes knowing how to critically evaluate AI-generated analyses and maintaining an awareness of the tool's limitations.
Redesign Curricula: Consider how to integrate AI-assisted close reading into existing curricula. This might involve developing new assignments that require students to use tools like NotebookLM to analyze complex texts, then critically evaluate and expand upon the AI's analysis.
Professional Development: Invest in training programs that help educators understand and effectively use AI tools for close reading. This should include both technical skills and pedagogical strategies for integrating these tools into classroom practice.
Foster Critical Thinking: While AI can enhance close reading, it's crucial to continue emphasizing the development of students' independent critical thinking skills. Design learning experiences that require students to go beyond AI-generated analyses and develop their own insights.
By embracing these strategies, educational leaders can help their institutions navigate the exciting but complex landscape of AI-enhanced learning. The goal is not to replace human intelligence with artificial intelligence, but to create a symbiotic relationship where technology amplifies and extends our cognitive capabilities. In doing so, we can prepare students not just to survive in an AI-driven world, but to thrive as critical thinkers, close readers, and lifelong learners.
Excellent longer post Rod. I really like the 3 part model and how that focuses the learner or user on closer reading.