Author Topic: Computer detection of logical fallacies?  (Read 590 times)

0 Members and 1 Guest are viewing this topic.

Offline Ambious

  • Frequent Poster
  • ******
  • Posts: 2020
  • Up the Irons!
Computer detection of logical fallacies?
« on: June 22, 2016, 05:20:58 AM »
So a few weeks ago I gave a lecture about logical fallacies to my Logics class, and yesterday my Logics professor told me my lecture inspired her to research and look into being able to create an algorithm to detect logical fallacies in texts, and asked me if I was interested in helping.
Of course I warned her that logical fallacies are highly context-dependent and that it can't possibly be easy for a computer where people can barely detect them with ease, but it sounds like a topic worth investigating at the very least.
If it goes further the project will not begin until next year, probably, but I thought you guys might have some insights. A quick survey of the internet showed us that this topic is barely researched, but it seems like we're going to have to employ data mining and computer learning, both fields which are very primordial and contested, and by no means easy to implement well.
Personally I'm skeptical, given that we're in a low-tier university with limited resources and that bigger and better people than us can barely manage to teach a computer to understand language WITHOUT searching for logical fallacies and understanding context, something which has only recently been achieved by the likes of Google.
What do you guys think? Is it doable? How would you approach such a task?
Quote
"Teach thy tongue to say “I do not know”, and thous shalt progress."
–Maimonides

Offline brilligtove

  • Stopped Going Outside
  • *******
  • Posts: 4598
  • Ignorance can be cured. Stupidity, you deal with.
    • Valuum
Re: Computer detection of logical fallacies?
« Reply #1 on: June 22, 2016, 06:49:21 AM »
On the face of it as an interested and somewhat informed bystander, I would think it is quite doable. My confidence in my opinion is quite low, however.

Some fallacies may be easier for a computer to see than for a human. I suspect that assuming the conclusion and many other simple fallacies would pop out of the text based on any conversion to formal logic. Others, like ad homenym attacks, might have a characteristic signature of a subject change, for example from (topics) to (people).

I wonder if you could train an AI by having it compete with humans in a gamified scenario. Take the CAPTCHA model as a base for how good the human is with a particular fallacy perhaps?

Sounds pretty interesting. Kudos for what was clearly an inspiring lecture!
evidence trumps experience | performance over perfection | responsibility – authority = scapegoat | emotions motivate; data doesn't

Offline The Latinist

  • Cyber Greasemonkey
  • Technical Administrator
  • Reef Tank Owner
  • *****
  • Posts: 9062
Re: Computer detection of logical fallacies?
« Reply #2 on: June 22, 2016, 07:00:34 AM »
I think formal fallacies might be doable, with the caveat that often there are unstated premises in arguments which are perfectly clear to a human but which might flummox an algorithm.

Informal fallacies strike me as harder, since they are not fallacious simply by the form of the argument.  The algorithm would have to be able to understand context and have a lot of generalized knowledge in order to differentiate between, say, a false continuum and a valid one.
I would like to propose...that...it is undesirable to believe in a proposition when there is no ground whatever for supposing it true. — Bertrand Russell

Online Andrew Clunn

  • Poster of Extraordinary Magnitude
  • **********
  • Posts: 14964
  • Aspiring Super Villain
Re: Computer detection of logical fallacies?
« Reply #3 on: June 22, 2016, 10:06:06 AM »
Okay computer, let's try ethics.  So assuming that human life has innate value...

Code: [Select]
LOGICAL FALLACY!
 :(
I agree with Clunn, which makes me feel all weird inside.

Offline Billzbub

  • Frequent Poster
  • ******
  • Posts: 2591
  • I know you know I know
Re: Computer detection of logical fallacies?
« Reply #4 on: June 22, 2016, 11:27:12 AM »
I would recommend that you start out implementing a limited set of grammar so you can get the rest of your algorithms working, and then expand gradually into letting people type whatever they want.  It will be nearly impossible for you to start out with all of your ultimate goals in mind and try designing to that.  The more you break it down into pieces that can be accomplished on their own, the more you can implement easier-to-accomplish pieces and gradually ramp them up to be more capable.  This sounds like so much fun.

Offline InsanePat

  • Keeps Priorities Straight
  • ***
  • Posts: 292
  • coward among the cowards
Re: Computer detection of logical fallacies?
« Reply #5 on: June 22, 2016, 11:29:45 AM »
So, I assume it would need to :
- identify the conclusion, the premises.
- guess the hidden premises, if any - not an easy one
- eliminate the useless  and redundant premises.
- identify the type of argument (deductive/inductive)
- if it is deductive, identify the form, check it,
- if it is inductive, ,  maybe do a fuzzy logic estimation
- if you want to be comprehensive, check the premises for validity... that one could be hell
all that through semantic analysis...
Sound like a job mainly for dnn(s) and supervised learning... but even with google power, it wouldn't be easy imo... Not that i'm an expert, I hope i'm wrong ;-).
 

Online PANTS!

  • One leg at a time.
  • Reef Tank Owner
  • *********
  • Posts: 8752
  • What seals? I auditioned for this job.
Re: Computer detection of logical fallacies?
« Reply #6 on: June 22, 2016, 11:47:11 AM »
Do you have the ability to detect a formal argument?  I would think that would be a first step. 

Then you you need to separate it into premises (possibly unstated) and syllogisms.  Once you get that, the rest is really easy.   :-*

I wonder if you might be better served by doing the Watson trick, and just search for context sensitive criticism of similar arguments.

EDIT: I see Pat beat me to the punch.
Now where I come from
We don't let society tell us how it's supposed to be
-Uptown, Prince 👉

We cross our bridges when we come to them and burn them behind us, with nothing to show for our progress except a memory of the smell of smoke, and a presumption that once our eyes watered.

Offline ThomasKennedy

  • Brand New
  • Posts: 1
Re: Computer detection of logical fallacies?
« Reply #7 on: March 16, 2017, 05:35:11 AM »
Hey guys--first post on this forum no idea if I'll receive any replies and if I commit any faux pas, please forgive me.  I am currently in the process of writing a grant proposal to fund exploration into a project that would involve centrally (though not exclusively) an alogrithm that could (with some significant margin of error, I have no doubt) identify logical fallacies in standard English, with a particular interest in analyzing journalistic prose, articles, authors, organizations etc.--with however many qualifications wind up ultimately being appended to that description corresponding to whatever aspects ultimately elude debugging within the scope of the project. I expect to get funding for the exploratory stage of my project at least but at stage one--just writing the first proposal--I'm just trying to familiarize myself with the state of the research.

Frankly--though perhaps somewhat naively--I was expecting that there would already be some major beta-stage projects going on in this area and I'm coming up with very little. I think this is an area that's really ripe for development. I would be interested in emailing, skyping with, or by any means communicating with any individual who is currently researching possibilities in this area. I would leave my email or phone number in my comment if this were a closed forum, but if anybody emailed me to vet me or put me in touch with students--or with anyone--currently doing thinking in this area I would be thrilled and grateful.

 There are prestigious cognitive science, linguistics, informatics and other relevant departments at my own university but it would be helpful to have my exploratory grant in hand to start relationships & open up doors in those communities to have the kind of conversations I'm hoping to have. What I'd really love is to have an email exchange or some other communications with anybody who's currently working on this to spitball some nitty gritty stuff. For example, I have no idea what an appropriate hourly or per article wage for a cog sci grad to redline & diagram news articles would be, and that would be something I need to build into my projected budget etc. There's probably about 20 or 30 weird specific issues that an hourlong phonecall with somebody who's already in this research would allow me to either get some clarity on or flag as too problematic to deal with at this stage.

I'm an art & design instructor at Indiana University with a background in a sort of exotic business consulting for Fortune 500 clients during which time I essentially liased between developers (with all kinds of cog sci & linguistic backgrounds) & c-suite executives (a population group notorious for certain types of irrationality) during which time I became better grounded in--among other things--formal logic, as well as textually and linguistically analytical algorithms. I also got experience planning, organizing and administering a few big data-harvesting and toolbuilding projects. This territory is new to me but tactically it seems at least homologous to problems I've worked on before. I expect the solution to problems in this territory will be multidisciplinary with cog sci & informatics people forming being the heart and soul--or maybe the brain & spine?--of the project.




Offline Boßel

  • Go away or I shall taunt you a second time.
  • Frequent Poster
  • ******
  • Posts: 2342
Re: Computer detection of logical fallacies?
« Reply #8 on: March 16, 2017, 11:00:01 AM »
I've been wanting to create an app like this for years, but because of my complete ignorance of algorithms and knowing nothing about code, I didn't think it was doable.

I thought, maybe have examples already available for comparison. Then maybe the computer could detect keys words and phrases from your example, then send you possible matches.

People choose which are correct and that will help refine the matches in the future.
« Last Edit: March 16, 2017, 11:03:39 AM by Boßel »

Online PANTS!

  • One leg at a time.
  • Reef Tank Owner
  • *********
  • Posts: 8752
  • What seals? I auditioned for this job.
Re: Computer detection of logical fallacies?
« Reply #9 on: March 16, 2017, 11:52:01 AM »
Sounds cool - I wrote it for you!

fallacyRegEx String = "^\w+ (O|o)(H|h)\s+(Y|y)(E|e)(A|a)(H|h)\w+"

IF IsMatch(fallacyRegEx, webPostScrape)
      Print "Fallacy Found"

 >:D
Now where I come from
We don't let society tell us how it's supposed to be
-Uptown, Prince 👉

We cross our bridges when we come to them and burn them behind us, with nothing to show for our progress except a memory of the smell of smoke, and a presumption that once our eyes watered.

Online Andrew Clunn

  • Poster of Extraordinary Magnitude
  • **********
  • Posts: 14964
  • Aspiring Super Villain
Re: Computer detection of logical fallacies?
« Reply #10 on: March 16, 2017, 02:24:56 PM »
Sounds cool - I wrote it for you!

fallacyRegEx String = "^\w+ (O|o)(H|h)\s+(Y|y)(E|e)(A|a)(H|h)\w+"

IF IsMatch(fallacyRegEx, webPostScrape)
      Print "Fallacy Found"

 >:D

Awesome, now embed it in a Perl script.
I agree with Clunn, which makes me feel all weird inside.

Offline Boßel

  • Go away or I shall taunt you a second time.
  • Frequent Poster
  • ******
  • Posts: 2342
Re: Computer detection of logical fallacies?
« Reply #11 on: March 16, 2017, 02:31:35 PM »
Sounds cool - I wrote it for you!

fallacyRegEx String = "^\w+ (O|o)(H|h)\s+(Y|y)(E|e)(A|a)(H|h)\w+"

IF IsMatch(fallacyRegEx, webPostScrape)
      Print "Fallacy Found"

 >:D

Bow bow, chik, chika, chika?

 

personate-rain
personate-rain