I'm a little confused, I'm a newcomer to the StackExchange community in general and I've been wondering why there's an abundance of downvotes for certain questions that seem to be non-homework and fairly genuine, interesting questions.

Could anyone elaborate on these reasons, (for reference, if you search for some of the newest questions, you'd immediately see a stream of questions, with downvotes for a reason that is not immediately obvious), if anyone could shed some light on this, I'd appreciate it.

EDIT: Some questions:

  1. Probability Distribution and Likelihood in the context of Deep Learning
  2. Q-learning with a state-action-state reward structure and a Q-matrix with states as rows and actions as columns
  3. Time Series Clustering, then Classification, need consistent data points? 4.Sklearn function to show probability of each word in cluster generated through LDA
  4. Is there a way to measure the performance of a neural network without using all the data?

Mind you, these are all of the newest questions.

| |
  • $\begingroup$ Pl link some such qns in your post $\endgroup$ – Dawny33 Jul 29 '17 at 17:47
  • $\begingroup$ I've lined 5 posts which demonstrate this issue, I don't really see a reason for the downvote (please let me know if I am missing something), it seems like a genuine question... $\endgroup$ – PSub Jul 29 '17 at 23:27

I did not downvote them, but I can understand why, to some extent. My quick evaluation (and speculation?) of the example questions follows. Please take them with a bit of salt. Also note that 4 out of 5 were deleted (as of 2017-08-12), which also proves their lack of quality.

  • In the first question, I had left a comment already: it could use some rewriting in order to make it look less broad, as right now it looks like 4 questions in one.
  • The 2nd one could probably benefit from a minimal example in code of what they have done so far, while cutting the initial explanation a bit shorter. It would be unfortunate if askers and future visitors had to read all 7 paragraphs before getting to the actual question.
  • The third one could use some clarifications, so I left a comment inviting the OP to provide an example and to show what was attempted so far.
  • Number 4 could also use some more code to better understand the kind of function they're looking for.
  • And finally, 5 is asking two questions and does not seem to show much research effort. In fact, at least the first part of the question has been covered before on this site: HOW TO: Deep Neural Network weight initialization and What is the purpose of setting an initial weight on deep learning model?

So no, I don't believe that questions are being downvoted because of being about deep learning, nor do I believe that deep learning questions are a problem here. In fact, I hope to see the Data Science SE community embrace good quality questions about deep learning, as they often reach the two ends of the science-technology spectrum to the point of being hard to know where to ask them across the SE network.

As an ending note: do not forget to upvote good questions. With sufficient triage from the community, those downvotes will not be as significant.

| |
  • $\begingroup$ I see, but what I don't understand is, these are generally, as I see it, clarifications that are to be asked of the original poster, right? It doesn't seem intelligent to willy-nilly downvote these questions, but rather to suggest improvements? $\endgroup$ – PSub Jul 30 '17 at 4:12
  • 1
    $\begingroup$ On that end, please note that downvotes already are a form of requesting for improvements, and there is no way to control someone's voting criteria (unless in cases of vote fraud) or to demand explanations to a vote in the comments. Besides, as far as those questions go, they only have a score of -1. Although it seems that they could use some explicit guidance to improve the question, I can't find this alarming, not even to deep learning questions. $\endgroup$ – E_net4 explains Jul 30 '17 at 8:38
  • 2
    $\begingroup$ I agree with this assessment, although I did not downvote them myself or know who did or why. It is reasonable to downvote questions that aren't well-formed. A comment would be nice explaining why though to help. This is all normal behavior. $\endgroup$ – Sean Owen Jul 30 '17 at 11:23
  • 1
    $\begingroup$ Also, yes, deep learning is entirely on topic here. $\endgroup$ – Sean Owen Jul 30 '17 at 11:23
  • $\begingroup$ Sounds, good, I'll try and do my part to request for clarifications and updates. $\endgroup$ – PSub Jul 30 '17 at 14:53

It can be correlated to some human characteristics that are studied by psychology: in humans being the cognitive capabilities are limited, so we use some heuristics that, although inaccurate, provide a good approximation.

Trying to be more precise, psychologists speak of heuristics, cognitive biases and cognitive distortions, but beyond the classification one of these universal mechanisms is the tendency to label what surrounds us, reducing it to just two opposite labels: black and white, right and wrong, healthy and sick, full or empty. The complexity of reality goes far beyond the shades of grey. A television is already made up of millions of pixels each colored (not shades of gray), the reality is at 3 dimensions and changes in time. Dichotomic thinking is the name of this way of thinking.

So, as human being, we are not logic neither rational, we rationalize our emotions. Once a distorted thought has been formulated, we all tend to defend it (confirmation bias) in order to avoid something called cognitive dissonance (studied by Festinger). We experience almost physical pain when two opposing beliefs coexist in our minds.

| |
  • 2
    $\begingroup$ Unless you are actually going to explain in a constructive manner why that relates to this meta question, we'll simply not take this argument seriously. What psychological issues could be present in the SE communities at large, other than the non-problem of striving for a repository of high-quality questions and answers? $\endgroup$ – E_net4 explains Sep 5 '17 at 15:22
  • $\begingroup$ @E_net4: sorry, you are right, I will modify the answer. $\endgroup$ – Revious Sep 6 '17 at 6:53
  • 1
    $\begingroup$ I wasn't expecting a lecture on psychology, but a concrete, non-passive explanation of how this relates to this question at hand (granted, it could involve the former, but not on its own). Also note that whatever is being implied here is not specific to Data Science SE. Other sites in the SE network also rate content with upvotes and downvotes, an approach known to be successful on Stack Overflow. $\endgroup$ – E_net4 explains Sep 6 '17 at 9:55
  • $\begingroup$ @E_net4: your sentence shows is how my explanation relates to the question: " an approach known to be successful on Stack Overflow". Ok, Upvoting and Downvoting in general proved to be a successful approach, but can you provide me ANY hint or evidence that excessive, compulsive and emotional, downvoting has showed ANY advantage? This is an example of Dichotomic thinking: since downvoting in general works well on average it doesn't matter if some users make an excessive use of it and it is only counter-productive for the community. Empty or full, why taking measures into considerations? $\endgroup$ – Revious Sep 6 '17 at 9:59
  • $\begingroup$ @E_net4: I know it may look annoying or boring.. if you don't have time to think about don't worry. It's more fitting as a topic on the community management section. $\endgroup$ – Revious Sep 6 '17 at 10:00
  • $\begingroup$ "evidence of excessive, compulsive and emotional, downvoting has showed any advantage" Likewise, you cannot prove that there is any of that, or that such forms of downvoting has occurred in the examples above. Users are given the right to vote anonymously as they please, because applying constraints to content ranking in a site where questions need to be ranked hardly make any sense. The opposite issue happens too: excessive upvoting on low-quality questions. You might understand all this better with a deeper look into the Meta Stack Overflow and Meta Stack Exchange sites. $\endgroup$ – E_net4 explains Sep 6 '17 at 10:10
  • $\begingroup$ @E_net4: Yes, you are right, sorry, my question doesn't apply to the specific examples brought from the guy. I was just trying to provide information which, in my opinion, shoule be understood thoroughly to improve this community. Thanks for your time. Just a mistake: "applying constraints to content ranking in a site where questions need to be ranked hardly make any sense" (could you provide any evidence? always be critical against your thoughts to avoid confirmation bias..) $\endgroup$ – Revious Sep 6 '17 at 10:28
  • $\begingroup$ See this example of emotional voting: stackoverflow.com/questions/26655534/… which pro can you see in a question which has been also revised by stackoverflow.com/users/19068/quentin . The excessive number of downvotes are easily explained by social psychology, people tend to act as the one before and find a pseudorational motivation $\endgroup$ – Revious Sep 6 '17 at 10:28

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .