[ExI] The Argumentative Theory of Reasoning

[Gordon]

Hi Richard. I remember you from discussions over on Ben's AI discussion group.

> Like many evolutionary accounts, it can sometimes appear to
> be saying something substantial when it is not.
> 
> (For example, consider the logical reasoning involved when
> Pythagoras discovered his theorem.  Does it make a
> difference that that kind of reasoning was born from a
> history of argumentation?  Not really).

Pythagoras argued in defense of his idea. His argument won the day. Did he really discover some absolute mathematical truth? Or did he merely defend a defensible idea with cogent arguments? 

Euclid might make a better example. We all know that Euclidean geometry represents at best an approximation of what we might call Truth. It would seem then that Euclid did not really discover and communicate any truisms; he merely argued successfully in defense of some propositions that seemed to him to make sense.

According to the Argumentative Theory of Reasoning (ATR) this is the real role of reason: to find and articulate arguments to justify a position. It is not the role of reason to actually find the truth.

> So, I think you are perfectly
> right to see a connection with the demise of the
> correspondence theory of truth:  if truth is about
> coherence, then argumentation is a way to increase the
> coherence of a system of "truths" by stirring the pot up as
> much as possible and letting it settle into a better (more
> coherent) state.

Glad you agree, but I'm still critiquing my own thoughts on this subject (oh no, according to ATR I should not reason alone! :)

> As for explaining confirmation bias (and all the other so
> called biasses):  we don't exactly need this new theory
> to explain that.

ATR seems to me unique in that it explains cognitive biases as adaptive traits. If ATR is true (whatever that might mean) when we should expect humans to have cognitive biases, in particular the confirmation bias. We should not consider the confirmation bias as a shortcoming. It allows for an efficient division of labor: each member of the group will see the pros of his own position (while remaining mostly blind to the cons) even while weighing the pros and cons of those contradictory positions argued by others. This is easier and less expensive for the group than for each member to weigh both the pros and cons of his own position and the pros and cons of the positions of each of the others. Or says the theory, if I understand it correctly.

-gts