# Dialectic algorithm - For calculating if an argument is sustained or refuted

post by Sophivorus · 2016-11-20T21:06:46.479Z · LW · GW · Legacy · 11 comments

This is a link post for https://en.wikiversity.org/wiki/Dialectic_algorithm

comment by Jayson_Virissimo · 2016-11-21T02:07:03.130Z · LW(p) · GW(p)

This could be a great technique for adding structure to internet discussions, but this algorithm says more about which debate team has more time on their hands, than which arguments have been refuted or not.

Replies from: ignoranceprior, Sophivorus
comment by ignoranceprior · 2016-11-21T04:30:07.927Z · LW(p) · GW(p)

Yes, for cases of Gish gallop it would be impractical to refute every single point.

comment by Sophivorus · 2016-11-22T15:30:19.267Z · LW(p) · GW(p)

The Wikidebate project has some wisdom on how to avoid that, and the fact that the algorithm is run on a wiki ensures that repeated, misguided or otherwise useless arguments are removed or improved. Also, coming up with arguments where there are none is pretty difficult, and stupid ones are much easier to delete than to write down.

Replies from: ChristianKl
comment by ChristianKl · 2016-11-23T22:21:11.448Z · LW(p) · GW(p)

Also, coming up with arguments where there are none is pretty difficult

That sounds to me like you have never spoken with people who are skilled at debate. It might be difficult for the average person but there are plenty of people who can debate well and argue nearly any point.

comment by MrMind · 2016-11-21T07:57:02.717Z · LW(p) · GW(p)

A Bayesian network is a generalization of a dialectic tree, where instead of 'sustained' or 'refuted' you have probabilities.

comment by sarahconstantin · 2016-11-20T23:13:56.572Z · LW(p) · GW(p)

This sounds like potentially a mechanic for a discussion site that keeps track of arguments.

Replies from: Sophivorus
comment by Sophivorus · 2016-11-22T15:26:30.661Z · LW(p) · GW(p)

I made such a site but I wasn't good at spreading it, so in the end I closed it and moved the project to Wikiversity and called it Wikidebate :-)

comment by Sophivorus · 2017-07-06T21:24:53.552Z · LW(p) · GW(p)

There's a wikidebate now on wether or not to use the dialectic algorithm and the answer seems to be clearly YES.

comment by ChristianKl · 2016-11-23T22:19:44.514Z · LW(p) · GW(p)

Given the way I have seen the term dialectic be used, calling this algorithm dialetic seems misleading.

There's no mechanism for synthesis.

Replies from: Sophivorus
comment by Sophivorus · 2016-11-24T02:31:25.264Z · LW(p) · GW(p)

Good point, yet the meaning of "dialectics" has changed a lot through the centuries, and on Wikipedia it currently is "a discourse between two or more people holding different points of view about a subject but wishing to establish the truth through reasoned arguments" which quite fits the function of the algorithm.

But how would you call it?

Replies from: ChristianKl
comment by ChristianKl · 2016-11-24T10:24:50.424Z · LW(p) · GW(p)

But how would you call it?

The first step would be to research whether there is prior art. "Refutation tree" might be a description that gives a better impression of what it does.

"a discourse between two or more people holding different points of view about a subject but wishing to establish the truth through reasoned arguments" which quite fits the function of the algorithm.

The algorithm doesn't care whether or not the arguments are "reasoned" it cares about whether they have counterarguments.

It's also not clear that the statements have a truth value to begin with. Appeals about what should be done, are value judgments. The example claims are also ill-defined. It starts with the question of whether everybody should be vegan and later people make arguments based on them defending that people should eat little meat. Motte-and-bailey is in full effect.