7
$\begingroup$

In connection with the moderator elections, we are holding a Q&A thread for the candidates. Questions collected from an earlier thread have been compiled into this one, which shall now serve as the space for the candidates to provide their answers.

Due to the submission count, we have selected all provided questions as well as our back up questions for a total of 8 questions.

As a candidate, your job is simple - post an answer to this question, citing each of the questions and then post your answer to each question given in that same answer. For your convenience, I will include all of the questions in quote format with a break in between each, suitable for you to insert your answers. Just copy the whole thing after the first set of three dashes.Please consider putting your name at the top of your post so that readers will know who you are before they finish reading everything you have written, and also including a link to your answer on your nomination post.

Once all the answers have been compiled, this will serve as a transcript for voters to view the thoughts of their candidates, and will be appropriately linked in the Election page.

Good luck to all of the candidates!

Oh, and when you've completed your answer, please provide a link to it after this blurb here, before that set of three dashes. Please leave the list of links in the order of submission.

To save scrolling here are links to the submissions from each candidate (in order of submission):

Dawny33

Kasra Manshaei


  1. Would you prefer to more aggressively filter out low-quality newbie questions and improve the experience for advanced data science users, or prefer to leave open and encourage improvement of newbie questions to make the site more useful to newcomers in the field?

  2. Do you think code debugging-help style questions are on-topic on this site? If yes, then how do we decide on the scope of them? For ex: Why is my code not running? style questions vs Why is this module not taking this hyperparameter? style questions. And so on.

  3. What types of questions are on-topic for Data Science but not Cross Validated, or vice versa? or, what types of questions are on-topic for both?

  4. How would you deal with a user who produced a steady stream of valuable answers, but tends to generate a large number of arguments/flags from comments?

  5. How would you handle a situation where another mod closed/deleted/etc a question that you feel shouldn't have been?

  6. In your opinion, what do moderators do?

  7. A diamond will be attached to everything you say and have said in the past, including questions, answers and comments. Everything you will do will be seen under a different light. How do you feel about that?

  8. In what way do you feel that being a moderator will make you more effective as opposed to simply reaching 10k or 20k rep?

$\endgroup$
3
$\begingroup$

Would you prefer to more aggressively filter out low-quality newbie questions and improve the experience for advanced data science users, or prefer to leave open and encourage improvement of newbie questions to make the site more useful to newcomers in the field?

As Data Science and ML is a domain which is being adopted by a lot of students and engineers these days, the huge influx of newbie questions is justified.

However, questions which demonstrate lack of basic googling and search efforts would need to be discouraged. However, there is no need to be harsh on them, as they might not know how and what to google. So, we can help them by commenting on their questions and closing them off.

These kind of helpful gestures are already a part of this site's culture, where even top users of the site like Neil Slater, Sean Owen, myself, Emre, etc to name a few, help new users by commenting the links to helpful resources.

So, questions like How do I get started in Data Science? or How do I build my career in Data science will be closed as they're very broad. But, pin-pointed questions about references and helpful resources like Are there any helpful resources for learning the concept of trend smoothening in time series analysis welcome as questions.

Do you think code debugging-help style questions are on-topic on this site? If yes, then how do we decide on the scope of them? For ex: Why is my code not running? style questions vs Why is this module not taking this hyperparameter? style questions. And so on.

The Why is my code not workng? belongs in the SO's main site, even if that code is completely ML-related, cause it's debugging help and not specifically related to Data science. However, questions like these [Why is this module not taking this hyperparameter?] should be on-topic, as they speak about the algorithm under the hood, and less about code or the library specifically.

What types of questions are on-topic for Data Science but not Cross Validated, or vice versa? or, what types of questions are on-topic for both?

I have answered that question in a similar discussion in the Meta StackOverflow site.

How would you deal with a user who produced a steady stream of valuable answers, but tends to generate a large number of arguments/flags from comments?

As the user is already a regular contributor, there is a very high probability he/she understands the culture of the site. So, dealing with his/her problems in chat with the mods and other contributors is the ideal way forward. However, I am a big believer of the Be Nice policy of SE. A site would be much better off not having toxic users, even though they're very high-profile. If the culture is friendly and welcoming, we can attract more users everyday.

How would you handle a situation where another mod closed/deleted/etc a question that you feel shouldn't have been?

I have already experienced this in the site which I currently moderate [Devops SE]. We have a chat group for the mods on the site, and the discussion is taken there for reaching a consensus.

In your opinion, what do moderators do?

Moderators just help fellow users on the site by helping resolve flagged posts and handle exceptional cases like conflict resolving, scope alterations of the site, etc. So, nothing more than a facilitator for the community.

In what way do you feel that being a moderator will make you more effective as opposed to simply reaching 10k or 20k rep?

Moderators have more privileges which the 10k+ users don't which including binding votes, taking decisions on users' suspension, unlimited flag votes, etc; which all come into play while conflict resolving and while moderating over toxic posts and users, helping them maintain the culture of the site.

$\endgroup$
  • $\begingroup$ "belongs in the SO's main site, even if that code is completely ML-related, cause it's debugging help and not specifically related to Data science" what if the question is Why is my code not achieving less performance than expected? It does not necessarily imply a bug, hence a solution may still require specific knowledge. $\endgroup$ – E_net4 is sad Aug 4 '17 at 18:33
  • $\begingroup$ Oops, I meant to ask about questions like Why am I getting less performance with this code than expected? (erroneous double negative in my previous comment) $\endgroup$ – E_net4 is sad Aug 5 '17 at 9:55
  • $\begingroup$ I did talk about such questions in the same line. Quoting myself from above: "However, questions like these [Why is this module not taking this hyperparameter?] should be on-topic, as they speak about the algorithm under the hood, and less about code or the library specifically." $\endgroup$ – Dawny33 Aug 5 '17 at 10:23
  • $\begingroup$ To be honest, I didn't find that kind of question similar enough. Are you suggesting that questions requesting for improvements in performance of a model should specify what aspects of the model should be improved, other than just "review my design"? $\endgroup$ – E_net4 is sad Aug 5 '17 at 12:04
  • $\begingroup$ @E_net4 Yeah + review my network design style questions are also welcome. Only ones which aren't are, the Please debug my code style questions, like for example, whenever the code throws type/size mismatch errors for matrices, etc. They're not ML-related question imho :) $\endgroup$ – Dawny33 Aug 5 '17 at 14:11
  • 1
    $\begingroup$ That is fine by me as well! Thank you for the clarification. $\endgroup$ – E_net4 is sad Aug 5 '17 at 14:12
-3
$\begingroup$

Kasra Manshaei

  1. Would you prefer to more aggressively filter out low-quality newbie questions and improve the experience for advanced data science users, or prefer to leave open and encourage improvement of newbie questions to make the site more useful to newcomers in the field?

So there are two different important points in one question:

  • The whole point is to answer questions in our field yes? So any question is welcome as long as they are valid which I write about bellow. We will be happy if we can motivate newbies to learn more and improve their theoretical and practical answers for sure!
  • But there is a BUT!!! No one can deny the recent raise of "let's be a Data Scientist" and I believe massive popularity usually reduces the intellectual properties of concepts. This has happened to Data Science recently and it's a good approach to protect the scientific soul of the field by editing and improving answers and questions instead of ignoring them. We can help keeping the scientific aspect of the field this way. For example:
  • The last but not the least is the archive aspect of SE. Let's keep it as a massive collection of QA in the field so filtering questions and answers out is the opposite way! We need to correct them.
  1. Do you think code debugging-help style questions are on-topic on this site? If yes, then how do we decide on the scope of them? For ex: Why is my code not running? style questions vs Why is this module not taking this hyperparameter? style questions. And so on.

Absolutly yes! They are on-topic as long as they are prog. questions related to DS. It's great that we can act like between theory and practice. And coding sometimes pains! We can help each other to save much time by answering coding questions. The other point is that it enriches the archive characteristic of SE by keeping questions in different aspects.

Deciding about the scope is not easy because there is no border. I would say any coding question showing up in a DS project is welcome.

  1. What types of questions are on-topic for Data Science but not Cross Validated, or vice versa? or, what types of questions are on-topic for both?

If there is a must to have Cross-Validated, then I have a concrete answer for this question which was my question as well. DS-related fields are heavily based on Statistics plus the fact that a huge part of Data Analysis is actually Statistics itself. So I would say Statistics-related questions in DS field better to go on Cross-Validated and the other parts here. A concrete example is PCA algorithm. I would say if someone has question about calculating Covariance matrix we can say $X^TX$ here. But if someone asks "Why $X^TX$ can be used as Cov matrix?" the math explanation can be in Cross-Validated. I would like to mention that at the end of the day I wouldn't mind combining both sites actually! (I even recommend it and I have a reason for that)

  1. How would you deal with a user who produced a steady stream of valuable answers, but tends to generate a large number of arguments/flags from comments?

I don't see any problem in this user! Thanks for his activities and contributions! Flags, etc. are also contributions. It's what a collaborative discussion means.

  1. How would you handle a situation where another mod closed/deleted/etc a question that you feel shouldn't have been?

As I said, for me SE is all about discussions. You just make your comment about it. And yes, sometimes mistakes happen in an irreversible way.

  1. In your opinion, what do moderators do?

To be honest I am new to this! I think it's a bunch of experts who maintain the quality of DSSE not a team of dictators who think anyone should ask/answer a question in the way they like. Diversity of minds is the only treasure we have here.

  1. A diamond will be attached to everything you say and have said in the past, including questions, answers and comments. Everything you will do will be seen under a different light. How do you feel about that?

I don't see a problem. Not only moderators in SE but in general anyone who takes a responsibility toward the benefit of many others should be as transparent as possible. I support this POV.

  1. In what way do you feel that being a moderator will make you more effective as opposed to simply reaching 10k or 20k rep?

They are different things I suppose! Being a moderator is about maintaining the quality of everything but having high-rep may even mean a huge passion in the field without feeling responsibility! There is a different between the star of a football team and the manager! None of them does (and is supposed to do) the job of other one.

Hope my answers helps making decision :)

$\endgroup$
  • $\begingroup$ 1. Do you believe that every question is salvageable? $\endgroup$ – E_net4 is sad Aug 10 '17 at 11:50
  • $\begingroup$ 2. On the other hand, why should we be attracting questions such as How to add regularizations in TensorFlow? here when they are on-topic on Stack Overflow? Drawing a line is hard, but a take-all strategy only sounds like an "easy way out" that could cripple this site with lots of poor programming questions. $\endgroup$ – E_net4 is sad Aug 10 '17 at 11:56
  • $\begingroup$ 4. You have misunderstood the question. Suppose that a user has significant contributions in the form of questions and answers, but also tends to have a behaviour that does not comply with the Be Nice policy, or other rules of the site. $\endgroup$ – E_net4 is sad Aug 10 '17 at 11:59
  • $\begingroup$ 1. The actual answer is of course not, but I would like to say "yes I do"! Because this is a professional community and the probability of having "that off-topic" questions is not that high. For me a question is not valid if it's that unclear that you can not even edit it. How often do they happen? Let's trust users and "tell them how to benefit from this community" if they "don't know". Look at this datascience.stackexchange.com/questions/22166/… . This guy just has a question! that's all! Why negative vote? $\endgroup$ – Kasra Manshaei Aug 11 '17 at 12:01
  • $\begingroup$ 2. Moving to Stack Overflow is fine but sometimes the guy does not know where to ask and removing the question may ends up in not asking this question at all. If one cares about this then it's fine and SO is a better place. But we already see reactions to the questions which are too "unnecessary". There is a spectrum between take-all strategy and the example I provided above. I just say let's see gray-levels more precisely which is not an "easy way out" but even needs more time. Check the link bellow to see how tried to correct a very "bad" question so that others don't delete it. $\endgroup$ – Kasra Manshaei Aug 11 '17 at 12:05
  • $\begingroup$ 2. Cont. Now, an initially "vary bad, wrong and fuzzy" question turned into a question with 3 upvote and my answer to it got an upvote as well. So in this case, the "easy way out" was actually to filter it out instantly ;) datascience.stackexchange.com/questions/22051/… $\endgroup$ – Kasra Manshaei Aug 11 '17 at 12:11
  • $\begingroup$ 4. Ahaa .. thanks for correcting me ... Then Reinforcement Learning I would say. They get reward for behaving well and punishment otherwise. After a while they learn how to behave. Even if the guy continues the bad behavior as well as providing good answers, we keep good answers and delete his wrong behavior! Users see these activities and report them usually. $\endgroup$ – Kasra Manshaei Aug 11 '17 at 12:14
  • $\begingroup$ I would argue that the first question linked is not a good fit because it's primarily opinion based. It's not something that attracts good quality answers. And not knowing how to ask is not an excuse to open the gates wide and let everything in. There is a limit to how much we should be shaping the site to newcomers. Having moderators and other high rep users improving bad quality questions is very hard work. $\endgroup$ – E_net4 is sad Aug 11 '17 at 12:46
  • $\begingroup$ As for the other subject, it's true that choosing the place for machine learning model "debugging/improvement" is not a solved problem. The questions that you linked are not in this category however. For instance, questions like this one keep showing up on Stack Overflow, and not many people there are suggesting to migrate them to Data Science. Nevertheless, let's see how things go in terms of machine learning questions with programming significantly involved. :) $\endgroup$ – E_net4 is sad Aug 11 '17 at 13:15
  • $\begingroup$ Oh, don't forget to update your answer! $\endgroup$ – E_net4 is sad Aug 12 '17 at 14:17
  • $\begingroup$ Well ... "And not knowing how to ask is not an excuse to open the gates wide and let everything in" is something I don't understand indeed! ... "excuse"? This is a place for people to get their answers. Opinion-based question is a question in which "being right/wrong" aspect of the answer is opinion-based. But "suggest a project which lets me show my DS skills" does not have right/wrong answer. Yes it attracts different opinions according to how to start getting hands dirty in DS and being opinion-based is actually something good there! I do not agree but fortunately we can express it here. $\endgroup$ – Kasra Manshaei Aug 14 '17 at 11:28
-3
$\begingroup$

VividD

Would you prefer to more aggressively filter out low-quality newbie questions and improve the experience for advanced data science users, or prefer to leave open and encourage improvement of newbie questions to make the site more useful to newcomers in the field?

I tend to like the later option more, but I would definitely implement some elements of the former.

Do you think code debugging-help style questions are on-topic on this site? If yes, then how do we decide on the scope of them? For ex: Why is my code not running? style questions vs Why is this module not taking this hyperparameter? style questions. And so on.

No.

What types of questions are on-topic for Data Science but not Cross Validated, or vice versa? or, what types of questions are on-topic for both?

DS and CV are overlapping, that is the reality. In my opinion, CV should contain questions with emphasis to pure statistics. DS covers fairly more diverse areas. In the end, the decision would be made by community.

How would you deal with a user who produced a steady stream of valuable answers, but tends to generate a large number of arguments/flags from comments?

I judge people equally. If one consistently behaves in a bad manner in comment areas, I will treat him/her the same, according to guidelines, and no matter how many answers he/she gave.

How would you handle a situation where another mod closed/deleted/etc a question that you feel shouldn't have been?

I would express my opinion to the other moderator.

In your opinion, what do moderators do?

The point of being the moderator is making community more interesting to participants. The means of achieving that are various.

A diamond will be attached to everything you say and have said in the past, including questions, answers and comments. Everything you will do will be seen under a different light. How do you feel about that?

I do not see any problem about that.

In what way do you feel that being a moderator will make you more effective as opposed to simply reaching 10k or 20k rep?

To me, making me more or less effective is not the point, I don't think about that at all.

$\endgroup$

You must log in to answer this question.

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