Elenchus: Aporia - Why Productive Confusion Beats False Consensus
When Reasoning Goes Sideways
Think about it: nearly every big decision in your organization comes out of a conversation. But how often do those conversations actually hold up to logical scrutiny? How often does anyone stop and ask - wait, does that reasoning even make sense?
Here's what I've noticed. Smart people - genuinely intelligent, well-meaning people - make predictable reasoning errors in meetings all the time. Not because they're dumb, but because that's just what humans do when they're under pressure, navigating social dynamics, and trying to think on their feet. The slippery slope that kills a good idea before anyone really considers it. The appeal to tradition that keeps a broken process limping along way past its expiration date. The ad hominem that derails a productive debate into something personal and ugly.
These patterns repeat everywhere. Every organization, every meeting, every decision. And because nobody's tracking them, the same bad arguments keep winning.
That's why I built Elenchus.
So What Does It Actually Do?
Let me just be straight about what this is.
Elenchus looks at meeting transcripts and finds logical fallacies. Not perfectly - nothing does that - but accurately enough to catch patterns that humans miss when they're in the middle of actual conversation.
Give it a transcript and you get back:
- Flagged fallacies with confidence scores and explanations
- Conversation dynamics - who's dominating, who's getting cut off, who went quiet after their first idea got shot down
- Severity ratings that separate minor rhetorical stumbles from reasoning that actually tanks an argument
- Recommendations for doing better next time
The confidence scores matter here. When Elenchus flags something at 0.85, it's saying "this really looks like fallacious reasoning." At 0.55, it's more like "maybe - worth a second look." The tool raises questions. Humans decide the answers.
Why Call It "Elenchus"?
Fair question. It's not just a cool-sounding name.
The elenchus was Socrates's go-to method back in ancient Athens - basically systematic cross-examination designed to expose the holes in someone's reasoning. If you've read Plato's dialogues, you've seen the pattern: someone makes a confident claim, Socrates asks a bunch of careful questions that draw out their assumptions, those assumptions turn out to contradict what they originally said, and suddenly the person realizes their thinking wasn't as solid as they thought.
Here's the thing though - the goal was never to embarrass anyone. It was aporia. Productive confusion. That moment when you realize you don't actually know what you thought you knew.
Why does that matter? Because you can't fix reasoning you don't know is broken. You can't find truth if you're already convinced you have it.
Elenchus the app does the same thing, just differently. It doesn't tell you what to think. It shows you where thinking went sideways and invites you to reason better. Different method, same purpose.
Or as Socrates might put it: turns out the examined meeting is worth having.
The Fallacy Lineup
I broke the fallacies into seven categories, pulling from Aristotle's work on sophistical refutations all the way through modern informal logic. Here's the tour.
Relevance Fallacies happen when the premises don't actually connect to the conclusion. Ad hominem goes after the person instead of their argument - like somehow discrediting the messenger makes the message go away. Straw man twists what someone said so you can attack an easier target. Red herrings throw in unrelated stuff to distract from the actual question.
Presumption Fallacies assume what they're supposed to prove. False dilemmas give you two choices when there are actually ten - "either we do this or we fail" conveniently ignores everything in between. Slippery slopes claim one thing will inevitably lead to disaster without showing any of the connections. Begging the question hides the conclusion in the premises, like a magic trick where the card was up the sleeve the whole time.
Weak Induction draws big conclusions from small evidence. Hasty generalization takes one example and runs with it - one bad experience becomes "they always do this." False cause assumes correlation means causation, the classic mistake of thinking that just because B happened after A, A must have caused B.
Emotional Fallacies swap feelings for actual arguments. Appeals to fear, pity, popularity, tradition - they all skip past reasoning and go straight for the gut. Sometimes effective. Not actually sound.
Formal Fallacies are structural problems in the logic itself. Affirming the consequent, denying the antecedent - these are invalid no matter what content you plug in. Like a building with a cracked foundation.
Informal Miscellany is the catch-all for stuff that doesn't fit neatly elsewhere. No True Scotsman dismisses counterexamples by just redefining terms on the fly. Moving the goalposts shifts what "success" means the second you achieve it, so nothing ever counts.
Ambiguity Fallacies let slippery language do the work. Equivocation switches what a word means mid-argument. Sunk cost confuses what you've already spent with what's actually worth doing - "I've come this far" becomes a reason to keep going even when stopping makes more sense.
Each fallacy in the system comes with detection patterns, context factors, and edge cases. It knows "we've always done it this way" might be appeal to tradition or might be legitimate institutional knowledge. Context tells you which.
It's Not Just About Logic
Here's something I figured out along the way: catching fallacies isn't enough.
Think about it - a meeting could have zero logical errors and still be completely broken. One person talks for forty out of fifty minutes. Someone else makes one contribution, gets pushback, and doesn't say another word. Certain voices get interrupted every single time; others never do.
These patterns matter. They decide whose reasoning actually makes it into the discussion and whose gets filtered out before anyone hears it. A perfect argument that gets cut off mid-sentence might as well not exist.
So Elenchus also tracks conversation dynamics:
- Who's talking how much
- How often people get turns and how long those turns are
- Interruption patterns - telling the difference between supportive "yeah, exactly" moments and actual cutoffs
- Dominance scores
- Who got silenced
There's a "conversation health score" that pulls all this together. It's not a moral judgment - presentations are supposed to have one main speaker - but it's a diagnostic tool for seeing how your discussions actually work.
Being Honest About Limits
Let me be real about what confidence scores actually mean.
A confidence score isn't a probability that someone committed a fallacy. It's how closely the detected text matches patterns associated with fallacious reasoning, adjusted for context.
High confidence means strong pattern match. It doesn't mean definite fallacy. The Principle of Charity still applies - that old idea that you should interpret ambiguous statements in the most reasonable way. Someone saying "everyone knows this is the right approach" might be pulling a bandwagon fallacy. Or they might just be emphasizing something that's actually well-supported. The words alone can't tell you which.
That's exactly why Elenchus is a tool for humans, not a replacement for them. Final calls require what only humans bring: full context, understanding of intent, knowledge of how things work in your specific domain, and charitable interpretation where it's deserved.
The system will have false positives and false negatives. It'll struggle with sarcasm and irony and creative new ways of being wrong. It needs human oversight.
These aren't bugs I'm planning to fix. They're built into the problem itself. Hell, Socrates got accused of committing the same fallacies he called out in other people. The method can be misused. That doesn't mean the attempt to reason better isn't worth making.
Who's This For?
Elenchus makes sense for organizations where:
- Big decisions come out of group conversations
- Getting better at discussion over time actually matters
- You want to hear everyone, not just the loudest people
- There's genuine willingness to look at your own reasoning honestly
It's not for scoring political debates or automating decisions that need human wisdom. It's for teams that want to think better together and have the guts to examine how they're thinking now.
Socrates got executed for his method - Athens got tired of having their assumptions questioned. Hopefully your organization handles it better.
But that's really the question Elenchus helps you answer: Is your team's discourse healthy enough to stand up to scrutiny? And if it isn't - would you even want to know?
Ironwright builds tools that help organizations think more clearly. Elenchus is my focus on bringing logic and critical thinking into actual business practice. If your team's decisions could use more rigorous conversation, let's talk.