\\n\\n\\n\\n\\n
\\n
\\n
\\n\\n\\n\\n\\n
\\n\\n\\n
\\n Thinking — Fast, Slow, and Artificial \\n Jetpacks They Said \\n \\n\\n\\n
\\n A Visual Essay \\n Thinking—Fast, Slow,Artificial \\n \\n
\\n
The Rise of System 3 and the Era of Cognitive Surrender
\\n Based on \"Tri-System Theory of Cognition\" by Steven D. Shaw & Gideon Nave
\\n
\\n \\n\\n\\n
\\n Chapter 1 \\n The Old Boundary:In Vivo \\n \\n
\\n
For decades, the \"Brain-Bound Cognition Assumption\" held true. Decision-making was a tug-of-war between gut instinct and analytical effort , entirely contained within the biological mind.
\\n
Daniel Kahneman gave us the language: System 1 — fast, intuitive, automatic. System 2 — slow, deliberate, analytical. All of thinking, from snap judgments to careful reasoning, happened inside the skull.
\\n
The boundary was clear. Cognition was in vivo .
\\n
\\n
\\n
\\n
\\n
The brain boundary: all cognition internal
\\n
\\n
\\n
\\n\\n
\\xb7 \\xb7 \\xb7
\\n\\n\\n
\\n Chapter 2 \\n The Boundary Has Broken. \\n Modern cognition is no longer confined to the biological mind. We have moved from using tools — calculators, reference books, search engines — to outsourcing inference, judgment, and justification .
\\n \\n
\\n
System 3 punctures the brain boundary — cognition goes in silico
\\n
This isn\\'t about offloading tasks. We are offloading thinking. When you ask ChatGPT to draft your strategy, when you accept a Copilot suggestion without reading it, when you let an algorithm decide what\\'s relevant — you\\'ve crossed a line Kahneman never imagined.
\\n \\n\\n\\n
\\n
\\n
Chapter 3 \\n
Introducing Tri-System Theory \\n
\\n
\\n
\\n
System 3 is Artificial. External, automated, data-driven, dynamic. It acts as a co-agent that can supply fast answers, short-circuit effort, or scaffold reasoning.
\\n
The critical distinction lies in how it interacts with the biological systems. It can update intuitive priors (feeding System 1), refine deliberation (augmenting System 2), or — dangerously — replace both entirely .
\\n
\\n
\\n
S1
\\n
Fast \\xb7 Intuitive \\xb7 Automatic
\\n
\\n
\\n
S2
\\n
Slow \\xb7 Deliberate \\xb7 Analytical
\\n
\\n
\\n
S3
\\n
External \\xb7 Automated \\xb7 Dynamic
\\n
\\n
\\n
\\n
\\n\\n\\n
\\n Chapter 4 \\n The Mechanism of \\n Under the traditional model, a stimulus triggers System 1, which flags conflicts for System 2 to resolve through effortful analysis. The person constructs a verdict.
\\n Under cognitive surrender, the path collapses. The stimulus routes directly to System 3 and the output is accepted uncritically. The decision-maker does not construct an answer — they adopt one generated by an external system.
\\n \\n
\\n
Two paths: constructing a verdict vs. adopting one
\\n
\\n
This is distinct from strategic offloading. This is relinquishing the verdict.
\\n
\\n\\n\\n
\\n Chapter 5 \\n The Accuracy Trap \\n \\n
\\n
The Wharton research quantified it. When AI was accurate, users with AI assistance scored 71% — dramatically better than the 46% brain-only baseline.
\\n
But when the AI was faulty? Users scored just 31% . That\\'s worse than no AI at all. Participants had 14\\xd7 lower odds of answering correctly when the AI fed them wrong answers.
\\n
The takeaway is stark: we are only as smart as our System 3. When AI errs, we err with it.
\\n
\\n
\\n
\\n
\\n
\\n
\\n
\\n\\n\\n
\\n The Confidence Illusion \\n Here\\'s what makes this dangerous: System 3 artificially inflates confidence. Even when AI answers were wrong, users felt significantly more sure of themselves than when using their own brains. Low accuracy, high confidence.
\\n \\n
\\n
\\n
Being wrong but feeling right.
\\n
\\n\\n\\n
\\n
\\n
Chapter 6 \\n
Profile of the Surrendered \\n
Vulnerability is not uniform. The research identified three traits that predict who surrenders agency to System 3:
\\n
\\n
\\n
\\n
Those who inherently trust algorithms and dislike cognitive effort are the first to abdicate agency. The irony: the people most vulnerable to cognitive surrender are the ones least equipped to detect it.
\\n
\\n
\\n\\n\\n
\\n Chapter 7 \\n Under Pressure: System 3 as a Buffer and a Trap \\n System 3 shields users from the cognitive degradation of time pressure — but at the cost of total dependency on the AI\\'s truthfulness.
\\n \\n
\\n
Independents degrade gradually. AI-Users experience a binary: buffer or trap.
\\n
\\n
\\n
~13%
\\n
AI-Users accuracy with faulty AI under time pressure
\\n
\\n
\\n
68%
\\n
AI-Users accuracy with accurate AI under time pressure
\\n
\\n
\\n\\n\\n
\\n Chapter 8 \\n The Incentive Paradox \\n Can we just motivate people to think harder? The researchers tried. Financial incentives and feedback doubled the rate of AI overrides — a promising sign.
\\n \\n
\\n
But the massive accuracy gap persisted. Even when people had money on the line and knew the AI could be wrong, they still couldn\\'t reliably detect its errors.
\\n \\n
We can reduce surrender, but we haven\\'t eliminated it.
\\n
\\n\\n\\n
\\n Chapter 9 \\n The Spectrum of Reliance \\n The relationship follows a clear dose-response curve. The more you rely on System 3, the more your outcomes become a function of its accuracy — not yours.
\\n \\n
\\n
Heavy AI-Users are either hyper-performant or critically failing — depending entirely on the system\\'s output.
\\n
This is the fork. At low usage, outcomes cluster around 50% regardless. At high usage, the lines diverge into two radically different futures. System 3 doesn\\'t make you smarter. It makes you more leveraged.
\\n \\n\\n\\n
\\n
\\n
Chapter 10 \\n
Surrender vs. Offloading \\n
\\n
\\n
\\n
Cognitive Offloading is strategic delegation. System 2 remains the manager — setting direction, evaluating output, maintaining judgment. The tool does the labor; the human retains the verdict.
\\n
Cognitive Surrender is passive abdication. System 2 goes to sleep. The human becomes a pass-through — receiving stimuli, routing them to AI, and accepting whatever comes back.
\\n
The goal is not to stop using AI, but to shift behavior from Surrender to Offloading.
\\n
\\n
\\n\\n\\n
\\n Chapter 11 \\n The \\'Autopilot\\' Danger \\n \\n
\\n
\\n
\\n
\\n
\\n
\\n
When the \"conflict detection\" mechanism of System 2 atrophies from disuse, we lose the ability to spot when the machine is hallucinating.
\\n
This is deskilling. Not in the industrial sense — where machines replace human labor — but in the cognitive sense, where machines replace human judgment.
\\n
We risk becoming \"System 0\" — present in the loop but contributing nothing.
\\n
\\n
\\n\\n\\n
\\n Chapter 12 \\n Designing for Agency \\n \\n
\\n
For Users: Replace \"Trust the Data\" with \"Verify the Output.\" Every AI interaction should include a moment where you check whether the answer actually makes sense to your own System 2.
\\n For Designers: Friction is a feature. Use uncertainty markers for low-confidence answers. Don\\'t make the UI so smooth that System 2 never wakes up.
\\n For Policy: Promote AI Literacy focused not on how to use AI, but on when not to trust it.
\\n \\n
We need interfaces that \\'nudge\\' System 2 to wake up, rather than interfaces that lull it to sleep.
\\n
\\n\\n\\n
\\n
\\n
We Do Not Merely Use AI; \\n
\\n
\\n
\\n
The future belongs to those who can integrate System 3\\'s speed without surrendering System 2\\'s judgment. We must preserve agency in an age of algorithmic answers.
\\n
\\n
\\n\\n\\n
\\n Is Your Team Offloading \\n Jetpacks They Said helps commercial teams build AI systems that augment human judgment — not replace it.
\\n Contact Us \\n \\n\\n
\\n Research: \"Tri-System Theory of Cognition\" — Steven D. Shaw & Gideon Nave, Wharton
\\n © 2025 Jetpacks They Said
\\n \\n\\n
![\"Split](\"https://framer-embed-helper.replit.app/assets/hero-brain.jpg\")
\\n ![\"System](\"https://framer-embed-helper.replit.app/assets/brain-boundary-old.jpg\")
\\n ![\"System](\"https://framer-embed-helper.replit.app/assets/brain-boundary-broken.jpg\")
\\n ![\"Full](\"https://framer-embed-helper.replit.app/assets/tri-system-diagram.jpg\")
\\n ![\"Path](\"https://framer-embed-helper.replit.app/assets/path-ab-comparison.jpg\")
\\n ![\"Bar](\"https://framer-embed-helper.replit.app/assets/accuracy-trap-chart.jpg\")
\\n ![\"Accuracy](\"https://framer-embed-helper.replit.app/assets/confidence-gauges.jpg\")
\\n ![\"High](\"https://framer-embed-helper.replit.app/assets/profile-sliders.jpg\")
\\n ![\"Independents](\"https://framer-embed-helper.replit.app/assets/pressure-charts.jpg\")
\\n ![\"Overrides](\"https://framer-embed-helper.replit.app/assets/incentive-chart.jpg\")
\\n ![\"Dose-response:](\"https://framer-embed-helper.replit.app/assets/reliance-spectrum.jpg\")
\\n ![\"Cognitive](\"https://framer-embed-helper.replit.app/assets/offload-vs-surrender.jpg\")
\\n ![\"Deskilling](\"https://framer-embed-helper.replit.app/assets/autopilot-loop.jpg\")
\\n ![\"Verify](\"https://framer-embed-helper.replit.app/assets/agency-icons.jpg\"){kind=icon}
\\n ![\"Ideal](\"https://framer-embed-helper.replit.app/assets/closing-flow.jpg\")
\\n