Comment by zozbot234 | Hacker News Reader

Comment by zozbot234 | Hacker News Reader

I'm pretty sure that the semantics of natural language are a lot more complex than can be accounted for by these seemingly very ad-hoc translations into comparatively straightforward FOL formulas, as are given in this paper. A common approach for the understanding of NL semantics from a strictly formal POV is Montague semantics https://en.wikipedia.org/wiki/Montague_grammar https://plato.stanford.edu/entries/montague-semantics/ - even a cursory look at these references is enough to clarify the level of complexity that's involved. Very loosely speaking one generally has to work with multiple "modalities" at the same time each of which, when understood from the POV of ordinary FOL, introduces its own separate notion of abstract "possible worlds" (representing, e.g. an agent's set of beliefs) and ways in which these "worlds" can relate to one another. More complex cases will usually degenerate in some sort of very generic "game semantics" https://en.wikipedia.org/wiki/Game_semantics https://plato.stanford.edu/entries/logic-games/ where any given use of natural language is merely seen as a "game" (in the abstract strategic, game-theoretical sense) with its own set of possibly very ad-hoc 'rules'. The philosopher Ludwig Wittgenstein https://en.wikipedia.org/wiki/Ludwig_Wittgenstein https://plato.stanford.edu/entries/wittgenstein/ gave quite a good description of both of these approaches (from a very naïve approach based on a supposedly straightforward translation to some kind of abstract logic, to a far richer one based on notions of strategies and games) to a "formal" understanding of natural language, throughout his extensive philosophical inquiry.

Which is to say, I'm not sure how this paper's results are generally expected to be all that useful in practice.

Your argumets and links are interesting, I hope to study these materials some day.

But.

To be useful in practice the approach does not need to work in all cases of natural language usage. Even if works in some limited cases there may be useful applications.

The authors evaluate their approach on two datasets. One is LOGIC consisting of learning examples of logical fallacies. The other is LOGICCLIMATE, consisting of logical fallacies collected from real world news articles about climate change.

The datasets are here, if anyone is interested to see the type of natural language they try to adress currently: https://github.com/causalNLP/logical-fallacy

I guess this csv contains the LOGICCLIMATE: https://github.com/causalNLP/logical-fallacy/blob/main/data/...

So a possible practicle utility for the approach - spot individual wrong sentences in a long article and highlight them.

Another real world example. I propose a solution at work, based on some statistics. And a colleague dismisses it by saying that there is a book "6 Ways to Lie with Statistics". If there was a smart assistant in the room who gently explained his logical fallacy to the colleague, it would save a lot of efforts for me and made the discusdion more productive. I doubt the difficulties you mention apply to this simple case.

"And a colleague dismisses it by saying that there is a book "6 Ways to Lie with Statistics"."

Except, that's going in the right direction towards a better argument: empiricism requires your statistics to be peer reviewed for errors or deception before being believed. That takes a skilled individual.

So, you either think they're very good at statistics or you want them to put faith in your work. Otherwise, they need a smart assistant they trust to review the statistics. Then, they have increased confidence in your solution but it still might be wrong.

He was not calling for better statistics, he suggested to ignore statistics.

It was a simple case and actually I was not presenting a statistics I collected, I just suggested to try using some numerical evidence to chose a decision.

On another occasion I mentioned to somebody that it's necessary to chose drugs or medical approaches verified with medical trials and double blind method. And they replied that there is a book about how to lie with statistics and continued to consider unverified methods.

I mean that in real life sometimes very simple fallacies happan.

Some statistics-based deсisions may be wrong => right decision must avoid statistics.

These cases could probably be adressed with automated tools of the near future.

Idk. When I see something like that in real life my conclusion is that people have different levels of intelligence.

Also I don't know what or how to teach someone who falls into this pitfalls.

I think it is easy to highlight the mistake by rephrasing the argument as OP did. By rewriting it that way, I think it is easy to see that the fallacy lies in taking some specific observations and saying that those are valid in general, without providing a solid explanation as of why that should be the case. Another way of rephrasing it could be, "I have evidence of some statistics-based decisions that were wrong, therefore all statistics-based decisions are wrong". If a person still doesn't get it, put it as "I have evidence of you being wrong once, therefore you must be always wrong (including in this specific case :D )".

I believe that is a misunderstanding of what the other party is expressing in this case. It's not "wrong once, so always wrong" but rather "intentional deception in the past, therefore might intentionally deceive again, and I don't know how to verify, therefore I shouldn't trust it".

Uhm, but then, it would work only if applied to the same entity that was deceptive in the past. Applying this to some other party would still make little to no sense, unless you've got some reason to believe this other party wants to deceive you as well.

Unless someone obviously shares common goals with you they are a potential adversary. When faced with a tool that you are confident can be used to deceive you, and a potential adversary who you are confident is aware of this fact, you should then clearly distrust that tool in that context.

In science, we're required to take a default deny approach. The scientific method treats everything as wrong or a lie by default. Then, a combination of independent review and replication add confidence to the claims. Assuming you trust that person.

Later on, media reported pervasive problems in reviewer independence (especially drug studies), statistical claims (especially p hacking), and replication ("replication crisis"). So, we now have more reason to not trust scientific claims without independent replication. That's double true for statistical claims.

If I read one, I say "maybe true, maybe not." At least brings me above a random guess in a belief on that topic. If they're highly biased, I ignore their claims on that topic entirely since cherry-picking evidence is so common. Certain sources have enough independent betting or positive outcomes to trust by default in a probabilistic sense. A person whose chips work is fairly trustworthy on basic, chip design and their own product designs. I default on believing their essential claims but know they might be disproven later.

That's how empiricism works. Anyone doing less is probably using some combo of faith, testimony, logic, or feelings. They can also dress these up with scientific language or mathematical formulas, too. But was it rigorously reviewed by someone who doubted everything about it? Often not for statistical or scientific-sounding claims.

> The scientific method treats everything as wrong or a lie by default

No, the scientific method requires proof both for a positive and for a negative answer. If we can't prove neither, all we can say is that we don't know whether something is true or false. Think i.e. about Riemann's hypothesis or the Collatz conjecture: should we say those are wrong because no one so far proved them to be correct?

But what could I do to help people get better in the __act__ of spotting this or doing what you desribe themselves.

Idk for me it is subconcious, I just "feel" it or know that it is logically faulty.

Also the rephrasing doesnt always work imo, you could have a logic statement, that is totally valid in some context and not valid in others. And you also need to think about the validity of the premise and if it is legit to draw the conclusion in natural language.

The paper proposes to use automated tools to help people avoid logic errors. :)

I doubt anyone is completely free from logic errors. I am certainly not.

But while seeing logic errors other make, we are often blind to our own errors logic - if we were aware of them we would correct them.

Just recently while considering a logical fallacy in somebody's argument I realised that on another occasion I myself used similar argumentation. So only when noticed the problem in somebody else I was able to recognize my own error.

There are many reasons people make logic errors. For example, if they like the conclusion they easier accept even flawed argumets in support.

I'm not sure that "distrust of thing I don't understand" can really be considered a fallacy. Certainly it sounds like the other party's tone wasn't constructive in this case. It also sounds like they are fairly ignorant.

Still, the underlying sense that you shouldn't trust people making claims based on things that you don't understand is probably a fairly solid survival strategy in general. Better to miss out than get scammed.

To put it another way, a call to "trust the science" in the absence of further elaboration is itself an appeal to authority. Despite that, it's not actually wrong - you generally should trust openly published science that has been reproduced by at least one unrelated party. Which serves to illustrate the rather glaring issue with the premise of the linked article, at least for practical everyday use.

Distrust would be right.

The fallacy was that people consider presence of statistical evidence as a negative sign. Not realizing its possible to lie without statistics as well.

Lets imagine a book "100 ways to harm your health with medicine", and a sick person choosing between magic and medicine: "Aha, the book has proven that medicine is harmful, so of course magic".

Indeed that would be the wrong conclusion to jump to.

However it isn't how I read the original example. I saw it more as "A is backed by evidence B" rebutted with "I don't trust evidence B because ...". Despite the described tone being poor and the individual obviously horribly ignorant, when assessed from their (apparent) point of view instead of my own that position seems fairly reasonable to me.

In other words, not so much "magic instead of medicine" as rejecting the claim that medicine is superior to magic while also declining to hold the view that magic is superior to medicine.

What you describe can be called healthy critical thinking.

The case I mentioned was different, I just described it poorly due to my limited English profficiency and typing on mobile, so you and others suspect the opponent meant reasonable doubts.

Anyways, logical fallacies are ubiquitous. And of the level of the simplest Aristotelian logic, not even requiring first order logic.

An automated tool capturing may probably be useful.

Arguments similar to "bananas are yellow, so if I see something yellow that's a banana" are quite often in politics.

The paper cites some previous studies of fallacies in an online forum and in argumentative essays.

Maybe I’m missing something, but how is calling out every time a news article cites a government agency as an appeal to authority a list of logical fallacies?

What were the alternative solutions you discussed?

Did a worse one get picked?

Did you already have a solution in place, and you were actually suggesting a change?

I've worked on classical NLP models for quite some time, and this indeed looks way too simple to be of any practical use. If you mention Montague, I'm going to refer you to "Pedro owns a donkey," the poster kid sentence for Discourse Representation Theory. That's 1980s work, and for simple sentences it's already complicated beyond what the OP article suggests, and fails on anything remotely complex. I think it goes 2nd order the moment a complement is introduced (I think that ...).

And even if you can translate a sentence into a predicate, you haven't begun understanding what lies behind all those predicates. E.g., "Zelensky is ready to work under Trump's 'strong leadership' after 'regrettable' showdown." What good does it do to have that in FOP?

[1] https://plato.stanford.edu/archIves/sum2011/entries/discours...

It looks like classic models of NLP semantics mostly punt on the "logical" point of view precisely due to these difficulties, and focus mostly on the more surface level problem of describing how each word of the source text correlates with a deeper description of the "meaning" of the text as a whole. So it is simply assumed that the meaning of the text as a whole must be derived compositionally from the meaning of each part (usually described by a somewhat ad-hoc "frame" structure), but exactly what that entails in a "logical" sense is left unspecified. UMR (Universal Meaning Representations) seems to be a typical example of such a system https://github.com/umr4nlp/umr-guidelines/blob/master/guidel... The expected use case seems to be something like building a common intermediate language for an automated translation system; individual meaning elements can then be "mapped" in a useful way, even across different languages, but there's not much interest apparently in "inferring" further knowledge from what one already has, or even on verifying that any given inference is valid (as proposed by OP).

Even beyond that you have a ton of pragmatics post-grice to deal with. Computing implicatures is complex and requires a lot of knowledge about context etc. The truth value of a statement and the 'truth value' of a speech act are pretty different things - not sure it's really feasible to convert between them.

Text that is written in Natural Language is open to interpretation. There are many formal statements that can be said to interpret a given Natural Language text. Can we determine which formal representation is correct? What about most useful?

The obvious answer to these questions is, "no". There is no such thing as a conclusive interpretation. If there was, then Natural Language wouldn't be ambiguous in the first place!

So we're all doomed to constantly misinterpret each other forever, right? No? We humans use Natural Language all the time, and usually figure out what the other person actually means!? How do we do it? Are we all just really good at guessing?

No, we have something better: context.

Context exists both in and around Natural Language text. Context determines which formal meaning is used to interpret the text. If we don't know which context is appropriate, there may be clues in the text itself that help us construct one that is useful or correct.

---

I've been trying to work out an approach to language processing that interprets text into logical formalisms (arbitrary meaning). I call them "Stories". A Story is an arbitrary interpretation of text. A Story is never conclusive: instead it is used as arbitrary context to interpret the next text. I call this process "Backstory".

We could even do the process backwards, and "write" an arbitrary formalism (meaning) in the same language/style/voice as a previously interpreted Story.

Given enough example instances of Story, we should be able to read and write to each other through explicitly shared context. I call this process "Empathizing". I call my idea the Story Empathizer.

I'm definitely out of my depth when it comes to the details, though...

I find humans have variation in ability for this as well though. Like some people need waaay more context, and need everything spelled out in granular detail to understand a topic, vs others who can more easily adapt, pick up clues and other relevant context information.

That's definitely true. I also suspect that holding too much potential context can be counterproductive, because then you have too many options to choose from. This happens a lot with jokes: there are a lot of unique backstories offered by different pop culture references, and pop culture is quickly diversifying to an overwhelming size. There is a lot of entropy in human expression.

The good news is that context can sometimes merge stories together. When we do explicitly find shared context, we tend to leverage that knowledge.

My idea is about offloading as much of this process as possible to a computer. We would still need to choose backstories, but the rest could be done in plain view, leveraging the incredible speed and memory size computers have.

I don't think that's the reason it won't be very useful. I think there are two reasons it won't be very useful:

1. Most natural language arguments are not sound because the argument is not deductive logic. Most natural language arguments are persuasive, not formal reasoning.

2. Formal logic is method of preserving truth. It doesn't really create truth. That makes it a lot less useful. Critically, while a deductively valid argument has a true conclusion if all the premises are true, an invalid argument can still have a true conclusion. Formal logic, then, is very narrow.

This is why finding a logical fallacy in an argument is often not convincing by itself. It doesn't say "your logic is flawed therefore I am right". It says "your logic is flawed and therefore should be revised and improved."

> Most natural language arguments are not sound because the argument is not deductive logic. Most natural language arguments are persuasive, not formal reasoning

related notes that there is some evidence that "Language is primarily a tool for communication rather than thought" [1]. ie, that language is neither necessary nor sufficient for the so-called psychic thinking process. It serves as a communication mechanism. Meanwhile, there is a hypothesis that the psychic thinking process lies beyond computation as we know it [2] in the form of turing machines etc.

[1] https://www.nature.com/articles/s41586-024-07522-w [2] https://www.amazon.com/Emperors-New-Mind-Concerning-Computer...

One way to slightly mitigate the difficulties of nuance in language when translating to formal arguments is to attwmpt to always steelman the argument. Afford it all the guarded language and nuance you can, and then formalize in premises and conclusion.

This would also make interaction much more civil as well, given so much proclivity to do the opposite (straw man).

It's not a perfect approach, but it helps. LLMs are quite decent at steelmanning as well, because they can easiky pivot language to caveat and decorate with nuamce.

See also for example V.H. Dudman on the interpretation of "If" sentences: https://www.scribd.com/document/478756656/Dudman-1984-Condit...

It could be useful for domains in which all or at least many problems are solvable (i.e., they can be stated and satisfied) with first-order logic.

It could also be useful as a lower-level component of general-purpose systems that internally rely on chains of thought computed by sub-component LLMs.

It wouldn't be useful if, as the parent comment is saying, it won't do a decent job of translating natural language.

Ah, got it. Thanks!

would be interesting if they had adversarial/null llms attempting the noisy nlp reductions as well. then one could make arguments about the sturdiness of the noisy bit.

Reply on news.ycombinator.com