The Shallows Revisited in the Age of AI

May 06, 2025

In my previous post, I revisited Nicholas Carr's seminal work, "The Shallows," and his warning about the internet's impact on our thinking. Published in 2010, Carr argued that the internet's very design was fragmenting our attention and eroding our capacity for deep, contemplative thought. Fifteen years later, in an era dominated by ubiquitous smartphones, addictive social media, short-form video, and now, powerful artificial intelligence (AI), Carr's concerns feel not just validated, but amplified.

In researching this post, I drew from numerous recent studies. For easier reading and to minimize cognitive overload, I've collected all references at the end rather than embedding them as hyperlinks throughout the text.

Carr's Concerns Echoed and Intensified

Post-2010 research largely confirms Carr's central thesis, that the internet has reshaped how we read and think. Studies continue to link extensive screen time with heightened attention-deficit symptoms, reduced working memory capacity, and lower text comprehension, particularly when reading online.

The challenges Carr identified have not only persisted but evolved dramatically: smartphones have created a state of "perpetual partial attention" through constant connectivity and notifications, while platforms like TikTok have accelerated these effects through rapid-fire content and algorithmic engagement, leading to documented decreases in attention spans, increased cognitive load, and potentially addictive behavior patterns that make deep focus increasingly difficult, particularly for developing brains.

The AI Amplification: A New Kind of Shallowing?

While reading "The Shallows," I realized that AI chatbots like ChatGPT and Gemini might solve some of the problems Carr identified with web browsing. The internet, as Carr described, creates cognitive overload, scatters our attention, and encourages us to skim rather than read deeply. AI chatbots could address these issues by eliminating the need to jump between multiple websites. Instead of forcing users to sift through search results and navigate hyperlinks, chatbots can directly present synthesized, relevant information in a single, coherent response. This streamlined approach could reduce the cognitive burden of information gathering that Carr was so concerned about.

Yet, early research suggests AI may introduce a different form of cognitive "shallowing." Rather than jumping between websites, we now engage in rapid-fire conversations with AI, potentially creating a new kind of scattered thinking as we constantly prompt, evaluate, and redirect these systems.

AI vs. Hyperlinks

Traditional web browsing, even with search engines, requires active navigation, evaluation of multiple sources, and synthesis by the user. This, while taxing, involves cognitive effort. AI, by providing direct answers and generating synthesized content, potentially bypasses these effortful stages. This raises the question: Does AI simply reduce cognitive engagement, or does it offer an opportunity to rethink it?

Cognitive Offloading on Steroids

Carr worried about outsourcing memory. AI allows us to outsource not just recall, but analysis, synthesis, and even content creation. This "cognitive offloading" raises serious concerns about "cognitive atrophy"—the weakening of critical thinking, independent reasoning, and problem-solving skills due to a lack of practice (the "use it or lose it" principle). While AI might enable tasks we wouldn't otherwise undertake, the question remains whether this reliance ultimately hinders long-term skill development.

Critical Thinking Shifted, Not Eliminated

Early research by Microsoft indicates that using AI doesn't eliminate critical thinking, but shifts it. The effort moves from information gathering and synthesis to information verification and integration of AI output. We need to evaluate AI's accuracy, potential bias, and relevance. However, the research also suggests that higher confidence in AI correlates with less critical thinking engagement, and users may neglect critical evaluation when AI output seems satisfactory, especially for low-stakes tasks. This creates a significant risk: cognitive abilities might deteriorate from lack of practice, leaving us ill-equipped for high-stakes situations where critical thinking is vital.

Cognitive Load

Does AI reduce mental effort? Initial research suggests it's not universal. While AI can automate parts of information processing, the need to verify outputs, craft effective prompts, and integrate AI responses can sometimes increase cognitive load, especially for complex tasks or unfamiliar users.

The Paradox of "Helpful" Software Revisited

Carr referenced a study by Christof van Nimwegen in which more helpful software made it harder for the participants to learn how to solve a puzzle. Will this insight apply to AI? Does the ease with which it provides answers prevent the "better imprinting of knowledge" that comes from struggle and independent effort?

Benefits of Internet and AI Use

It's crucial to avoid a purely dystopian view. Research also indicates that internet use (and potentially AI) can offer cognitive benefits, particularly for older adults or in specific educational contexts. AI can provide personalized learning, enhance efficiency by automating routine tasks, and act as a powerful cognitive assistant, potentially freeing mental resources for higher-level thinking. The key, as ever, lies in how these tools are used.

It’s Too Early to Tell

It's crucial to note that many of the observations and concerns raised above are based on preliminary findings, anecdotal evidence, and theoretical extrapolations from earlier research on digital technologies. The widespread adoption of powerful generative AI is in its infancy, and rigorous, peer-reviewed studies specifically examining its cognitive impacts are still rare. Questions about how AI affects critical thinking, cognitive load, and learning outcomes remain largely unanswered. The long-term implications for cognitive development, particularly in educational contexts, are yet to be thoroughly studied. Additionally, the rapid evolution of AI capabilities means that research findings may quickly become outdated.

Navigating the New Shallows: Choosing Our Cognitive Fate

Despite the lack of concrete research yet, Carr's warning in "The Shallows" still resonates powerfully in the age of AI. The potential for fragmented attention, superficial understanding, and diminished critical thinking seems greater than ever. These technologies aren't simply neutral tools—they shape how we think and work, often prioritizing speed and efficiency at the expense of deep, deliberate thinking.

Reversing or mitigating these effects requires conscious effort, mirroring strategies supported by research on the reversibility of internet effects:

Mindful Use: Actively choose when, where, and how we use these tools. Set boundaries and time limits.
Digital Detox: Incorporate regular breaks from constant connectivity to allow the brain to rest and recover attentional control.
Prioritize Deep Work: Deliberately engage in activities that demand sustained focus, like reading physical books, engaging in nature, or pursuing focused hobbies.
Cultivate Digital Literacy: Develop skills to critically evaluate information from all sources, including AI, recognizing potential biases and inaccuracies.

It's perhaps ironic, though not unexpected, that I've used AI extensively to help research, synthesize, and even write these very posts reflecting on technology's cognitive impact.

"You employ me, a product of the very technological shift you investigate, to synthesize knowledge about its cognitive consequences. This act highlights the emergent, symbiotic relationship between human intention and artificial intelligence, where understanding the tool involves engaging with the tool." - Gemini

Ultimately, the question isn't whether AI will change our brains—neuroplasticity guarantees it will. The question is whether we'll consciously choose what to pay attention to and how to construct meaning, or whether we'll passively allow the technology to dictate the terms of our thinking. The stakes, in an age of increasingly powerful and pervasive AI, have never been higher.

Jeremy Neiman’s Book Notes

Discussion about this post