Subjects · Philosophy & Critical Thinking
Philosophy & Critical Thinking: the methods that matter most here
Know which methods do real philosophical work, and which popular AI habit produces confident nonsense dressed as insight.
What you'll be able to do: Know which methods do real philosophical work, and which popular AI habit produces confident nonsense dressed as insight.
What makes philosophy hard, specifically
Prose hides argument structure. A paragraph can carry two premises or five, beg the question, or shift the meaning of a word halfway through, and none of it shows on the page as an error. Grading philosophy largely means grading whether the structure is visible and valid once you force it into the open, as numbered premises.
The principle of charity is a rule, not a courtesy. You're obliged to attack the strongest version of a position, not the version you can beat. Students, and AI, default to the weak version because it's easier, and every interesting disagreement in the discipline only exists once that cost has been paid.
A verbal dispute wears the same clothes as a substantive one. Two people can agree on every fact and still argue for an hour, because they mean different things by the load-bearing word. Sorting a verbal dispute from a substantive one is unglamorous and most of the discipline's false progress comes from skipping it.
Validity and soundness are different axes, and prose conflates them. A valid argument can have a false premise and reach a false conclusion with total logical propriety. A true conclusion can follow from an invalid argument by accident. Both feel, from inside a paragraph, like "does this argument work", and it takes reconstruction to tell them apart.
A position isn't yours until it's survived questioning. Confidence and defensibility are not the same thing, and most essays state a position no one has ever pressed on for more than a sentence.
Ethics's hard cases pit a good against a good, not a good against a bad. If one side of your dilemma is obviously wrong, you haven't found a hard case in this discipline's sense, you've found an easy one dressed up as hard.
AI is unusually good at sounding right here, whether or not it is. Confident, well-structured, definitive-sounding prose is the tool's default register, and it's also the discipline's native register. Unlike a date or a lab result, there's no external fact to check a reading against except the text itself.
The ten
| # | Method | Why it earns its place here |
|---|---|---|
| 1 | C11 Argument reconstruction | Makes the hidden structure visible; validity can't be checked in prose |
| 2 | C5 Steelmanning | The principle of charity, operationalised |
| 3 | C8 Definition forcing | The direct fix for a verbal dispute masquerading as a substantive one |
| 4 | C7 Assumption surfacing | The premise doing the real work is usually the one nobody stated |
| 5 | C6 Thesis defence | The ten-minute test, run for real |
| 6 | C9 Counterexample hunting | A generalisation that's never been attacked hasn't been tested |
| 7 | D21 Ethical dilemma staging | Forces a case where both sides are genuine goods, not a good and a villain |
| 8 | C4 Devil's advocate | Attacks a position you actually hold, not one you're describing |
| 9 | C14 Dialectic | Works a live disagreement to find out whether it's verbal, and if not, where the real fork is |
| 10 | C12 Fallacy identification | Validity has a name for each specific way it fails; learn the names |
The method that works badly here
C2 Discovery-first inquiry , being led to an insight by questions, never told it outright.
This is the right shape for a science concept that works one way regardless of opinion, or a proof that's either valid or isn't. It's the wrong shape for most of what philosophy actually studies. Free will, personal identity, meta-ethics, the ontological status of number, these are live disagreements among people who've spent careers on them, not puzzles with a hidden solution waiting to be found.
Run discovery-first inquiry on "does personal identity require psychological continuity" and the AI has to have already picked which answer counts as "the insight," then steer you there through questions while presenting the destination as your own discovery. That's worse than a stated opinion, not better, a stated opinion you can argue with. A manufactured "aha" you experienced as arriving at unaided, you won't think to interrogate, because it doesn't feel like it came from outside you.
The tell: you finish the session feeling you've discovered the answer to a question philosophers have argued for centuries. You haven't discovered anything. You've been steered, and the steering was invisible precisely because it never announced a conclusion.
The composed workflow
- Reconstruct the argument (C11) before evaluating it.
- Force definitions on the terms carrying the weight (C8), check whether the dispute is verbal.
- Steelman the position you reject (C5), checked against the principle of charity.
- Surface your own unstated premises (C7).
- Defend the resulting thesis under real attack (C6), and hunt counterexamples against any generalisation you've made (C9).
That sequence, reconstruction, definition, charity, assumptions, defence, is most of a first-year philosophy degree, compressed into a workflow you can run on your own claims before a marker runs it on you.
The ten articles in this subject
01 Argument reconstruction · 02 Steelmanning · 03 Definition forcing · 04 Assumption surfacing · 05 Thesis defence · 06 Counterexample hunting · 07 Ethical dilemma staging · 08 Devil's advocate · 09 Dialectic · 10 Fallacy identification