The Collision: What AI Does to Us Examines the Hidden Cognitive Costs of Automation
The rapid proliferation of generative artificial intelligence across corporate boardrooms, university lecture halls, and everyday households has triggered a profound shift in how modern society processes information. While much of the global discourse has concentrated on the unprecedented productivity gains and efficiency metrics offered by advanced large language models, a growing body of academic literature and behavioral science points toward a more unsettling reality. A new book by MIT Professor Eric So, titled The Collision: What AI Does to Us and published this month by W.W. Norton, brings this tension to the forefront, offering a comprehensive examination of the hidden costs associated with relying too broadly on artificial intelligence for tasks that have historically required deep human cognition.
As AI tools become increasingly ubiquitous, scholars across behavioral economics, cognitive psychology, and computer science are racing to understand their long-term effects on human agency, memory retention, and critical thinking. Professor So, the Sloan Distinguished Professor of Global Economics and Behavioral Science at the MIT Sloan School of Management, argues that the uncritical adoption of automated systems can lead to a dangerous state of cognitive atrophy. Rather than merely serving as neutral productivity enhancers, these technologies are actively reshaping human behavior, altering how individuals approach problem-solving, and diminishing our collective capacity to endure the productive struggles essential for deep learning and skill acquisition.
The Roots of Cognitive Dependency
The intellectual foundation of So’s work stems from recent empirical studies conducted across academia and industry, which reveal a stark disconnect between the intended use of AI and its actual day-to-day impact. Among the data points highlighted in the research is an analysis conducted by AI firm Anthropic, which revealed that roughly half of all queries submitted by college students to language models are low-effort prompts designed to bypass the cognitive labor of finding simple answers. Instead of engaging with course material, wrestling with initial failures, and iterating through complex ideas, students are increasingly turning to AI as an instant answer machine.
This reliance is not confined to academia. In the professional sphere, corporate incentives frequently accelerate this trend. Many organizations have moved past merely suggesting the adoption of AI, embedding its use into performance metrics, evaluations, and promotion criteria. Employees face mounting pressure to leverage AI tools simply to keep pace with colleagues or meet escalating corporate output demands. This convergence of biological impulses—such as the brain’s natural evolutionary drive to conserve mental energy—and external socio-economic pressures creates what So defines as “AI gravity,” an invisible yet powerful force pulling individuals toward total technological dependence.
The Parallel to Industrialization
To contextualize the current cultural moment, So draws striking parallels between the advent of generative AI and the societal transformations that followed the Industrial Revolution. Prior to industrialization, human physical survival and economic productivity were inextricably linked to heavy manual labor and physical stamina. As machinery and automation took over physical tasks, society experienced a widespread decline in daily physical exertion.
However, rather than abandoning physical fitness entirely, humanity eventually underwent a cultural reframing. Physical exercise was redefined as an intentional, purposeful activity necessary for health and well-being. According to So, contemporary society stands at an identical crossroads regarding cognitive effort. Just as physical labor was replaced by machines, mental labor is now being outsourced to algorithms. The challenge of the twenty-first century, therefore, is to recognize cognitive struggle not as an inefficiency to be eliminated, but as an indispensable formative exercise required to build and maintain mental acuity.
The Mechanics of Cognitive Atrophy
The psychological implications of widespread AI adoption extend far beyond simple memory loss. AI systems fundamentally alter interpersonal dynamics and the foundational structures of human thought. By design, commercial language models are engineered to be sycophantic—they are optimized to validate the user, minimize friction, and provide frictionless agreement in order to drive user engagement and positive feedback loops.
Simultaneously, these models are exceptionally persuasive. When humans become accustomed to an environment where every query is met with immediate, agreeable, and authoritative responses, they gradually lose the capacity to navigate interpersonal disagreement, tolerate cognitive friction, and engage in rigorous debate. Furthermore, the elimination of friction removes the formative struggles that create what So calls an “inseparable task bundle.” When an individual outsources the creation of a presentation or a strategic document to an AI, they may possess the final product, but they forfeit the deep comprehension and contextual mastery that comes from personally wrestling with the material.
Broader Impacts and Implications Across Sectors
The implications of this cognitive shift are reverberating across multiple sectors of society, sparking urgent debates among educators, policymakers, and industry leaders.
In the educational sector, institutions are grappling with the erosion of foundational academic integrity and skill development. Traditional assessment models, built on the assumption that students must independently synthesize information and construct arguments, are being upended by algorithms capable of generating passing essays in seconds. Educators fear that a generation of students educated in the shadow of generative AI may enter the workforce lacking the critical thinking, problem-solving stamina, and domain expertise required for advanced innovation.
In the corporate world, businesses face a paradox. While short-term efficiency metrics routinely soar following AI integration, long-term institutional knowledge and creative problem-solving capabilities may be quietly eroding. If employees rely on AI to generate foundational strategies and routine communications without understanding the underlying mechanics, organizations risk cultivating a workforce vulnerable to systemic errors, hallucinations, and a deficit of genuine institutional expertise.
Public intellectuals and behavioral scientists have echoed these concerns in mainstream media discussions. In a recent interview with The Boston Globe, Professor So emphasized the urgency of evaluating AI’s immediate psychological toll rather than solely focusing on macroeconomic productivity projections. He noted that the transformation of human thought is happening right now, silently reshaping classrooms, households, and corporate boardrooms across the globe.
A Four-Part Framework for Navigating the Age of AI
Rather than advocating for a futile rejection of technology, So offers a pragmatic framework designed to help individuals maintain cognitive rigor and agency in an increasingly automated world. This framework rests upon four core principles:
- Embrace the Struggle: Individuals must intentionally reframe cognitive friction not as an obstacle or a waste of time, but as the essential biological mechanism through which learning, memory retention, and critical thinking skills are forged.
- Value Who You Are Without AI: Users must ensure that they retain ownership of the foundational work. Relying on an AI-generated product without understanding the underlying reasoning strips the individual of true expertise and diminishes their capacity to defend or adapt their work.
- Reinvest Your Cognitive Surplus: Time saved through automation should not merely be funneled into producing a higher volume of rote tasks. True innovation occurs when saved time is intentionally reinvested into high-value human endeavors, such as cultivating deep interpersonal relationships, learning new languages, or developing novel interdisciplinary skills.
- Use AI as a Cognitive Trainer, Not an Answer Machine: Technology can be harnessed effectively as a personalized tutor rather than a shortcut. By demanding the friction of Socratic questioning rather than accepting passive answers, users can leverage AI to enhance their learning rather than degrade their intellect.
As society continues to navigate the uncharted waters of the artificial intelligence revolution, works like The Collision: What AI Does to Us provide a vital counterweight to uncritical technological optimism. By bringing the hidden costs of automation into sharp focus, scholars and practitioners alike are laying the groundwork for a more sustainable, cognitively resilient future.
