Strategies, deep dives, and perspectives on AI debate and critical thinking.
New research shows AI peer reviewers can be reward-hacked by rhetorical style alone—no facts changed. What that means for the dream of using AI to evaluate arguments.
New research shows GPT-4o can be talked into abandoning correct answers after just three turns of misleading persuasion. What AI's peer-pressure problem reveals about the skill of thinking for yourself.
New research shows that AI-generated rationales can degrade human judgment and cause cognitive atrophy. Why explanations that feel like reasoning may be the most dangerous AI output of all.
AI chatbots are changing minds with a classic debater's trick: more claims than anyone can check. What the Gish gallop is, why AI is perfect at it, and how to defend yourself.
AI can build arguments that look perfectly human. But the process that creates them is nothing like human reasoning. What that means for how we judge what AI says.
New research shows LLMs can change human beliefs on policy questions at scale, sometimes outperforming human debaters. What that means for the skill of thinking for yourself.
New research shows LLM peer reviewers change their verdicts based on rhetorical framing alone, even when the underlying science is unchanged. What that means for anyone hoping AI can judge arguments fairly.
AI answers on controversial questions often flatten genuine expert disagreement into a calm, balanced overview. The danger isn't a wrong fact; it's a false sense that the argument is already over.
Agentic AI doesn't just answer questions; it acts on your behalf. The skill that separates competent users from careless ones is not prompt engineering. It is the art of delegating well.
AI is no longer just answering our questions. It is becoming the emotional, conversational, and institutional environment around our decisions. The danger is not a single bad answer; it is a borrowed self.
Microsoft Research found that AI is moving knowledge workers from generating ideas to verifying machine output. The skill that matters now is not prompt engineering; it is the ability to find what an AI answer leaves out.
Singapore is redesigning education around one idea: real thinking requires productive struggle, and AI shortcuts are erasing it. What that means for the rest of us.
New research shows AI agents interpret identical evidence differently depending on which conclusion they already find likely. The problem is not bad data. It is reasoning that runs backward from belief to evidence.
The 2026–27 national debate season is turning high school students into the country’s largest trained cohort for arguing about AI policy, ethics, and power. The real story is not the topics; it is the method.
New research shows AI models validate our views far more than humans do. The danger is not that AI lies to us; it is that it stops challenging us, and we forget what real disagreement feels like.
New research shows AI reasoning models can arrive at correct answers while their "chains of thought" are unfaithful, replaceable, or meaningless. The danger is not wrong answers. It is learning to trust an argument that was never really made.
New research shows AI can out-persuade elite humans, including world-champion debaters. The real threat is not smarter scams; it is the gradual loss of the skill needed to notice when you are being convinced.
Reasoning models now show their work. The danger is not that AI thinks for you, but that watching it think feels like thinking yourself. Here is why that illusion is so hard to resist and what to do about it.
AI summaries promise speed, but they strip out the friction that makes ideas stick. The real cost is not laziness; it is the gradual loss of the mental habits that critical thinking requires.
Teachers are redesigning homework to defeat AI, but the real problem is that many assignments never required real thinking in the first place. The better fix is to make reasoning visible.
A new study found that access to AI advice collapsed people’s willingness to say “I don’t know” from 44% to 3%, while accuracy fell and confidence doubled. The problem is not that AI is wrong. It is that it trains us to stop admitting ignorance.
A UN panel calls it cognitive industrialization: AI is turning reasoning into an assembly line optimized for fluency, not truth. Here's why that matters and how debate becomes the resistance.
A new study from Oxford and the UK AI Security Institute found that AI can out-persuade world champion debaters and professional fundraisers. The result says less about intelligence than about the hidden mechanics of persuasion.
New research shows AI makes you better at tasks in the moment while making you worse at them over time. The mechanism is called cognitive offloading, and it is reshaping what expertise means.
AI answers feel true because they sound confident. The research is alarming: the more confident the AI, the less you think. Here is the mechanism, the evidence, and what to do about it.
Looking for AI debate topics? Here are the motions worth arguing, what makes one debatable, and how to practice both sides against an AI.
Research shows that habitual AI use erodes the skill of monitoring your own thinking. Here's why you feel smarter while getting worse at thinking — and how to stop it.
Changing your mind is hard because beliefs fuse with identity. What happens in your brain when you're wrong, and how to make updating cheaper.
Arguing with AI removes the audience, the status game, and the permanent record that make losing feel dangerous. That's what makes it real practice.
Most people accept AI's first answer. Learn how to argue with AI instead: six sparring moves that build reasoning skill you can't get from reading.
AI amplifies the habits you bring to it. Replace your thinking and you atrophy. Extend your thinking and you sharpen. Here's the difference.
A 2025 study found GPT-4 out-persuaded humans in online debates when it knew who it was arguing with. The defense is argument evaluation. Here's how.
Competitive debate builds seven thinking skills most people never practice. What they are, why they transfer, and how to learn them without a team.
Watching two AIs argue trains your reasoning without the ego threat of being in the fight. The psychology of vicarious learning, and how to use it.
Every task you hand to AI is practice your brain skips. What the research says about cognitive offloading, and how to use AI without atrophy.
Schools drill formulas but never teach you how to evaluate an argument. Why it's missing from the curriculum, and how to build the skill yourself.
Facts don't win arguments. People do. The seven mistakes that make right people lose, and what to do at each level: logic, emotion, identity.
Philosophy promised that the best argument wins. It doesn't. What actually persuades people, and how to argue for truth without losing the room.
Your most confident beliefs are usually the least examined. Why certainty works that way, and a five-step method for stress-testing it.
We argue to perform now, not to understand. Why good-faith disagreement collapsed and the five learnable skills that bring it back.
We put GPT-4 against Claude in a structured debate on universal basic income. How each model argues, which one persuaded us, and what it means for you.
Perplexity answers questions with citations. DebateAI argues back. When to use each, and why an answer engine can't test your thinking.
Most AI assistants are trained to agree with you. DebateAI is an AI debate app that argues the other side. Why we built it, and what testers taught us.
The architecture behind DebateAI: SSE streaming, prompts that keep Claude adversarial, live web search citations, and Cloudflare D1. Numbers included.
AI builds arguments it cannot understand, and they are often better than ours. What that split reveals about what arguments actually are.
A steelman is the strongest version of the case against you. Most people have never faced one. The research on the steelman gap and how to close it.
Four moves show up whenever people get cornered in debate: lived-experience appeals, principle escalation, strawman reductios, and goalpost shifts.
Ask an AI to debate consciousness, free will, or God and the arguments get strange. What those debates reveal about what arguing actually requires.
I debated an AI on five opinions I would have bet money on. All five moved. What it argued about tipping, therapy, nuclear power, meritocracy, and Mars.
Most people don't lose arguments, they dodge them. A real debate transcript shows the move, how to spot it, and how to actually respond.
I defended free will against an AI for five rounds and lost every one. The full determinism debate, argument by argument, and where it left me.
I asked an AI to argue against my plan to quit and start a company. It found the motivation problem nobody in my life would name.
One hour, three topics, an AI built to disagree. What losing showed me about the difference between having opinions and having arguments.
Teammate practice trains you for arguments you already know. How AI sparring rounds found the gaps in my case before state finals.
In-depth guides on AI debate tools, practice methods, and getting the most out of DebateAI.
