Comments about the article in Nature: An autonomous debating system

Following is a discussion about this article in Nature Vol 591 18 March 2021, by Noam Slonim e.a.
To study the full text select this link: https://www.nature.com/articles/s41586-021-03215-w

• The text in italics is copied from the article
  
• Immediate followed by some comments
  

In the last paragraph I explain my own opinion.

Reflection

• Reflection 1 - The complexity of debating
• Reflection 2 -
• Reflection 3 -

Introduction

"System Architecture - page 380"

"Argument Mining"

Argument rebuttal

Debate construction

Evaluation and results

Comparison to beseline systems

Evaluation of the final system

In-depth analysis

Discussion

On line content

Any methods, additional references etc are available at: https://doi.org/10.1038/s41586-021-03215-w

Reflection 1 - The complexity of debating.

Consider two persons "Pro" and "Contra" who are having a bebat about General Relativity. "Pro" is favour of General Relavity and supports all details. "Contra" is not against General Relativity but is in favour "Newtons Law". The purpose of the debat is to answer the question: Which of the two theories Einstein's Law (GR and SR) and Newton's is the most powerfull to understand the physical reality. To be more specific to predict the future.

The strucure of the debat will very much follow the outlay in the article.

• First "Pro" will make his opening statement and explain how important GR is compared with Newton's Law. As part he will also explain what all the benefits are.
• First "Contra" will make his opening statement and explain how important Newton's Law is compared to GR. As part she will also explain what all the benefits are.
• Secondly "Pro" will responds to openening statement of "Contra" and specific chalenge the argument that Newton's Law is important and accurate.
• Secondly "Contra" will responds to openening statement of "Pro" and specific chalenge the argument that GR is important and accurate.
• In the third Phase "Pro" and "Contra" will start in a more direct discussion between each other, trying to convins that he or she is right and the other party is wrong

The question to ask: what will be the result of this debat?
In any debat original the two debaters have each their own opinion. The debat is a succes if after the debat there is only one opinion or one conclusion.
To reach a common conclusion it is important to identify first all the aspects of the debat that both parties agree upon. Secondly to define as clear as possible the opinion of "Pro" that both understand and agree upon and thirdly to define as clear as possible the opinion of "Contra" that both understand and agree upon
The debat makes progress if the subjects that handle the common aspects increases and the subjects that span the individual aspects decreases. The debat comes to a close when there are no individual aspects any more. This is also the case when the result is a list of open issues or tasks to be done.

The problem in this type of debate is that there is a clear individual who is in favour and a second individual who is against. In many debates this seems to be the case, but in reality this is not. In reality any person can be in favour but can have doubts. At the same time for your oponent it can be completely the other way around.

Reflection 2 - The complexity of debating between computers.

In Reflection 1 the purpose of the debat is to answer the question: Which of the two theories Einstein's Law (GR and SR) and Newton's is the most powerfull to understand the physical reality. In Reflection 1 the debat is performed between two humans.
When you perform the same debat between a human and a computer, the most chalenging part is the opinion of the computer and what his arguments are.
An even more interesting debat is a debat between two computers and to listen to the opening statements of each. In principle they should be identical.

At the same time I have a serious problem with the whole experiment. As explained in Reflection 1 the evolution of the debat is to define what is the common opinion and what are the pecularities of each of the participants i.e. a human and a computer. The whole problems lies in the details of the reasoning involved.
In fact there are two types of reasoning: mathematical and physical. Mathematical reasoning is simple. Physical reasoning is complex. In fact it means to what extend we understand the news and view articles in Nature, and identify possible errors in the logic used.
The second step is to 'feed' a computer with the same text and 'ask' the computer to identify the possible error sources and how to correct these errors.

Reflection 3

Back to my home page Index
Back to Nature comments Nature Index