Artificial Intelligence
(fact checking)
Tag: Visual edit
(downvote answers)
Tag: Visual edit
 
Line 16: Line 16:
 
The question was about novelty search which is an algorithm to improve the learning of neural networks. In the posting it was focused on the problem of controlling a robot. The author has claimed, that novelty search allows in doing so. The problem with the answer is, that the referenced NEAT algorithm comes from a background of biological driven Artificial intelligence research. It was first discovered in the context of understanding real neurons. Using the NEAT algorithm for controlling a robot is possible, but it's not the main objective.
 
The question was about novelty search which is an algorithm to improve the learning of neural networks. In the posting it was focused on the problem of controlling a robot. The author has claimed, that novelty search allows in doing so. The problem with the answer is, that the referenced NEAT algorithm comes from a background of biological driven Artificial intelligence research. It was first discovered in the context of understanding real neurons. Using the NEAT algorithm for controlling a robot is possible, but it's not the main objective.
   
  +
  +
  +
== Is it allowed to downvote answers? ==
  +
From a formal perspective, each user has the ability to downvote existing answers and explain the reason why in a short comment. This would allow to improve existing answers and the conflict may help to identify missing knowledge. A short look into the reality shows, that downvoting an answer is a rare action.
  +
  +
The reality can be determined by browsing through some answers on a random basis. What we will observe is, that the typical answer gets upvotes and sometimes in the comment it's written “very good answer, thanks”. Other comments are asking an additional question, but that is not the same like a downvote.
  +
  +
The missing downvotes to answers can be backup-ed by a look into the downvote statistics. Most users didn't even know, that the downvote button exists. The average user in SE.AI has pressed the downvote button only 3 times, but the upvote button around 300 times and more. The reason why is a certain understanding how the community is working internally. The assumption of the average user is, that the answers in SE.AI are making sense and there is no need to argue against it.
  +
  +
There is a problem with avoiding downvoting of answers. It prevents, that the community can learn from mistakes. If no conflict is there, there is no reason to change anything. And if nobody has to ask himself about missing knowledge and wrong assumption, he won't be able to improve the quality.
  +
  +
This pattern can be described as a missing error culture. It's a collective behavior which doesn't tolerate internal conflicts because the assumption is, that these conflicts will hurt the overall community.
   
 
[[Category:2019]]
 
[[Category:2019]]

Latest revision as of 19:53, 2 December 2019

The user Manuel Rodriguez has received a lot of downvotes for it's posting:

https://ai.stackexchange.com/questions/8261/is-known-math-really-enough-for-ai/8262#8262

https://ai.stackexchange.com/questions/5410/what-is-the-current-state-of-agi-development/5411#5411

https://ai.stackexchange.com/questions/12835/what-approach-should-i-take-to-model-forecasting-problem-in-machine-learning/12836#12836

The referenced posting were deleted by the normal moderation pipeline which means that at first the post gets downvoted and then at least two moderators have to press the delete button.

Fact checking for SE.AI

Before an article can gets deleted because of misformation it make sense to discuss the content first in a reproducible way. The first posting which is criticized is

The question was about novelty search which is an algorithm to improve the learning of neural networks. In the posting it was focused on the problem of controlling a robot. The author has claimed, that novelty search allows in doing so. The problem with the answer is, that the referenced NEAT algorithm comes from a background of biological driven Artificial intelligence research. It was first discovered in the context of understanding real neurons. Using the NEAT algorithm for controlling a robot is possible, but it's not the main objective.


Is it allowed to downvote answers?

From a formal perspective, each user has the ability to downvote existing answers and explain the reason why in a short comment. This would allow to improve existing answers and the conflict may help to identify missing knowledge. A short look into the reality shows, that downvoting an answer is a rare action.

The reality can be determined by browsing through some answers on a random basis. What we will observe is, that the typical answer gets upvotes and sometimes in the comment it's written “very good answer, thanks”. Other comments are asking an additional question, but that is not the same like a downvote.

The missing downvotes to answers can be backup-ed by a look into the downvote statistics. Most users didn't even know, that the downvote button exists. The average user in SE.AI has pressed the downvote button only 3 times, but the upvote button around 300 times and more. The reason why is a certain understanding how the community is working internally. The assumption of the average user is, that the answers in SE.AI are making sense and there is no need to argue against it.

There is a problem with avoiding downvoting of answers. It prevents, that the community can learn from mistakes. If no conflict is there, there is no reason to change anything. And if nobody has to ask himself about missing knowledge and wrong assumption, he won't be able to improve the quality.

This pattern can be described as a missing error culture. It's a collective behavior which doesn't tolerate internal conflicts because the assumption is, that these conflicts will hurt the overall community.