A lot was said about the accuracy of labels, chosen metrics and strange data split in this thread: https://www.kaggle.com/c/tensorflow-great-barrier-reef/discussion/308329
\nLet me add a nature conservation enthusiast's point of view.
The competition title was very attractive: \"TensorFlow - Help Protect the Great Barrier Reef\" and I hoped that GBR would be the main winner here. But as it shows now, the protection effort got nearly lost somewhere in the data arrangement. In fact nearly the only what remained from the title was promotion of Tensorflow.
\nBut I may be wrong and there was some well hidden reason or purpose. So with all the respect to great winners' solutions and your work I have some genuine questions to you, @addisonhoward and @wcukierski :
\nHow did the strange data split (and all the above listed factors) serve to help protecting the GBR?
\nCouldn't the main beneficiary (CSIRO) have obtained even better and more practically useful solutions if the data arrangement had been different?
\nWill Kaggle staff have the same hardly understandable approach to other conservationists' competitions? As I understand this is always Kaggle that makes the decisions about data arrangement.
Thanks in advance for your response.
", "messages": [ { "id": "1700061", "postDate": "02/21/2022 16:22:39", "content": "A lot was said about the accuracy of labels, chosen metrics and strange data split in this thread: https://www.kaggle.com/c/tensorflow-great-barrier-reef/discussion/308329
\nLet me add a nature conservation enthusiast's point of view.
The competition title was very attractive: \"TensorFlow - Help Protect the Great Barrier Reef\" and I hoped that GBR would be the main winner here. But as it shows now, the protection effort got nearly lost somewhere in the data arrangement. In fact nearly the only what remained from the title was promotion of Tensorflow.
\nBut I may be wrong and there was some well hidden reason or purpose. So with all the respect to great winners' solutions and your work I have some genuine questions to you, @addisonhoward and @wcukierski :
\nHow did the strange data split (and all the above listed factors) serve to help protecting the GBR?
\nCouldn't the main beneficiary (CSIRO) have obtained even better and more practically useful solutions if the data arrangement had been different?
\nWill Kaggle staff have the same hardly understandable approach to other conservationists' competitions? As I understand this is always Kaggle that makes the decisions about data arrangement.
Thanks in advance for your response.
", "rawMarkdown": "A lot was said about the accuracy of labels, chosen metrics and strange data split in this thread: [https://www.kaggle.com/c/tensorflow-great-barrier-reef/discussion/308329](url)\nLet me add a nature conservation enthusiast's point of view.\n\nThe competition title was very attractive: \"TensorFlow - Help Protect the Great Barrier Reef\" and I hoped that GBR would be the main winner here. But as it shows now, the protection effort got nearly lost somewhere in the data arrangement. In fact nearly the only what remained from the title was promotion of Tensorflow.\n\nBut I may be wrong and there was some well hidden reason or purpose. So with all the respect to great winners' solutions and your work I have some genuine questions to you, @addisonhoward and @wcukierski :\nHow did the strange data split (and all the above listed factors) serve to help protecting the GBR?\nCouldn't the main beneficiary (CSIRO) have obtained even better and more practically useful solutions if the data arrangement had been different?\nWill Kaggle staff have the same hardly understandable approach to other conservationists' competitions? As I understand this is always Kaggle that makes the decisions about data arrangement.\n\nThanks in advance for your response.", "votes": null }, { "id": "1709788", "postDate": "03/02/2022 13:43:45", "content": "\"How did the strange data split (and all the above listed factors) serve to help protecting the GBR?\"
\nIn fact nearly the only what remained from the title was promotion of Tensorflow. \"
\nI'm so naive that I made some work with Pytorch :)
\nPractical solutions that I read during the comp: fork the COTS (Not the Notebooks) manually, burn and bury them.
\n\"Marine tourism operators and community groups are leading the initiative using injector kits to cull COTS, and remote villages without kits are removing COTS manually and burning or burying them. \" (in \"Controlling crown-of-thorns starfish in a small island developing nation: using collaboration and innovation to overcome resource limitations.\"
\nAuthors: Johanna Johnson, David Welch, Peter Whitelaw, Christina Shaw, Pascal Dumas,
\n, Rocky Kaku, Owen Drew, Kehana Andrew
The sponsoring TF team intends to integrate the learnings from this project into CSIRO's glider for real-time detection of COTS. This takes time--time to understand the models, time to implement, time to test, debug, refine--and so we (Kaggle) don't have an immediate answer on how this competition impacted the GBR. I certainly believe there is already a value in having thousands of people made aware of this problem and commonly focus on a solution. As researchers, you don't know where the state of the art is unless you have a lot of smart people press on it.
\nIt's healthy to ask the general question of \"couldn't this competition have been more impactful if X,Y, or Z was different\" at the end of any competition we run, and the answer is probably always going to be yes, albeit with 20/20 hindsight, unlimited time and resourcing before deadlines, all project stakeholders in perfect agreement, and absent the constraints, leakages, and imperfections real-world data brings with it :) What we've observed from running hundreds of competitions is that--in spite of the imperfections--solid, generalized models still rise to the top of challenges. We do continually listen to feedback on competition design and work to make them as fair and high quality as we can.
", "rawMarkdown": "The sponsoring TF team intends to integrate the learnings from this project into CSIRO's glider for real-time detection of COTS. This takes time--time to understand the models, time to implement, time to test, debug, refine--and so we (Kaggle) don't have an immediate answer on how this competition impacted the GBR. I certainly believe there is already a value in having thousands of people made aware of this problem and commonly focus on a solution. As researchers, you don't know where the state of the art is unless you have a lot of smart people press on it.\n\nIt's healthy to ask the general question of \"couldn't this competition have been more impactful if X,Y, or Z was different\" at the end of any competition we run, and the answer is probably always going to be yes, albeit with 20/20 hindsight, unlimited time and resourcing before deadlines, all project stakeholders in perfect agreement, and absent the constraints, leakages, and imperfections real-world data brings with it :) What we've observed from running hundreds of competitions is that--in spite of the imperfections--solid, generalized models still rise to the top of challenges. We do continually listen to feedback on competition design and work to make them as fair and high quality as we can.", "votes": null }, { "id": "1710952", "postDate": "03/03/2022 13:57:47", "content": "@wcukierski
\nDear Mr. Cukierski and all your loyal supporters,
\nThank you for your highly generalizing response, though the main question \"why such a hardly understandable data arrangement especially in this competition?\" remains unanswered.
\nIn fact no answer can sometimes be the best one as it proves the rightness of doubts.
P.S.
\nIt is really funny that yesterday, at the very same moment of receiving your response, I also got 2 simultaneous downvotes which (mysteriously?) downgraded my comment from silver to bronze. And downvotes will still stay undercover …😎
\nI know very well that Kaggle Competitions' environment is nearly a strictly Big Boys Club and I, as a quite successful Czech female, can always be only an unwelcome intruder. Where are the proclaimed Kaggle Values?
\nFrankly said, at such occasions I feel like staying among little boys on a sandbox playground.😀
Edited:
\nAfter 1 hour my main comment is back on silver.🎶
Hi Allie,
\nDon't bother about the downvotes. It's Not a gender thing since male, female and prefer not to say are downvoted equally. Even the \" Big Boys\" :) are downvoted.
\nWorst that being downvoted, is being ignored.
\nSurviving on the Playground it's part of the whole process.
", "rawMarkdown": "Hi Allie,\n\nDon't bother about the downvotes. It's Not a gender thing since male, female and prefer not to say are downvoted equally. Even the \" Big Boys\" :) are downvoted. \n\nWorst that being downvoted, is being ignored.\n\nSurviving on the Playground it's part of the whole process.", "votes": null }, { "id": "1711299", "postDate": "03/03/2022 19:57:18", "content": "Hi Allie,
\nI am saddened to hear that you feel unwelcome here. You are welcomed, as is your constructive feedback. Our community guidelines specifically call out, \"when you have critical feedback, focus on the ideas that others are sharing, not the person\" as well as:
\n\n\nWe strive to be a community that welcomes and supports people of all backgrounds and identities. This includes, but is not limited to, members of any race, ethnicity, culture, national origin, color, immigration status, social and economic class, educational level, sex, sexual orientation, gender identity and expression, age, size, family status, political belief, religion, and mental and physical ability.
\n
Under these guidelines, upvotes and downvotes should be signals of the quality and constructiveness of the content, not personal affronts. You should continue to voice concerns if you feel like this is not what is happening.
\nRegarding my response being general: I do not feel as though the problem split was unorthodox here (public and private use the same videos split by time), but as I alluded to above, we often pick between an imperfect set of choices and do so prior to all the information that is surfaced while the competition is running. There are tens of critiqueable design decisions that feed into each launch and people willing to argue both sides of most of them. Trying to defend them publicly, with 20/20 hindsight exposing subtle differences buried within the data, can sometimes feel unproductive for us and unsatisfying for you.
", "rawMarkdown": "Hi Allie,\n\nI am saddened to hear that you feel unwelcome here. You are welcomed, as is your constructive feedback. Our [community guidelines](https://www.kaggle.com/community-guidelines) specifically call out, \"when you have critical feedback, focus on the ideas that others are sharing, not the person\" as well as:\n\n> We strive to be a community that welcomes and supports people of all backgrounds and identities. This includes, but is not limited to, members of any race, ethnicity, culture, national origin, color, immigration status, social and economic class, educational level, sex, sexual orientation, gender identity and expression, age, size, family status, political belief, religion, and mental and physical ability.\n\nUnder these guidelines, upvotes and downvotes should be signals of the quality and constructiveness of the content, not personal affronts. You should continue to voice concerns if you feel like this is not what is happening.\n\nRegarding my response being general: I do not feel as though the problem split was unorthodox here (public and private use the same videos split by time), but as I alluded to above, we often pick between an imperfect set of choices *and do so prior to all the information that is surfaced while the competition is running*. There are tens of critiqueable design decisions that feed into each launch and people willing to argue both sides of most of them. Trying to defend them publicly, with 20/20 hindsight exposing subtle differences buried within the data, can sometimes feel unproductive for us and unsatisfying for you.", "votes": null }, { "id": "1722524", "postDate": "03/14/2022 16:04:29", "content": "@wcukierski
\nDear Mr. Cukierski,
First only a note.
\nI didn't complain about the fact of getting downvotes, I was only pointing at the mysterious time coincidence of your response and the downvotes. At least I know, why downvotes will never be made public.
\nBack to the main topic which you opened citing the Community Guidelines.
\n\"We strive …\" - I tend to believe that you really do. But let the data speak about the bitter reality.
\nFemale Kagglers form only 1.2% in top 600 competition rankings and, as far as I know, there is only 1 female competition Grandmaster (of 248) as well as only 1 female Triple Grandmaster - my deepest respect to both of them.
\nI hope that nobody would say that women and girls are not clever enough to successfully compete with boys in DS. But do you have an acceptable explanation for such a huge imbalance? What or who plays the main role in the situation that we can participate but hardly ever be long-term successful?
\nIsn't it accidentally a consequence of the atmosphere of escalated ruthlessness, aggressiveness and disrespect for anybody and anything including Kaggle Competitions' rules which you tend to overlook?
\nSometimes some Kaggle staff members can even contribute to the feeling of unfairness and injustice - e.g. no reaction to lost parts of GPU quota due to system maintenance error, strange and highly subjective system of disqualification.
\nYou also cite: \" We’re all here to learn and share ideas.\" But for many competitors the main goal is only getting points and tiers for a prospect of carrier boost no matter how they reach it. And some \"minor\" rules violations are broadly considered more acceptable than publicly pointing at them.
For some competitors this atmosphere one day simply becomes unbearable and they quit with active kaggling. No wonder that it holds true mostly for women. And, please, don't try to only tell me that it saddens you. Proclamations not followed by an action can hardly ever help.
\nI hope that my feedback is constructive enough and it isn't necessary to add more facts or evidence.
To wrap it up:
\nI don't ask for any special treatment for anybody or anything like it. I only ask for Kaggle staff more actively supporting the environment of dignity, honesty, fairness and mutual respect as described in Community Guidelines. This could help to every decently behaving Kaggler. Kaggle Competitions should really be a platform for learning and practising DS skills not for training surviving (as Marilia noted above).
Hi Allie,
\nEven GM are downvoted: take a look at this:
\nhttps://www.kaggle.com/product-feedback/76394
\n\"I've been debating making this kernel public, but given how this thread went down with all the downvoting- I figure it's worth releasing. Last week I made a comment on a kernel mentioning that much of the code was copied without giving credit to the creators. This comment had 5 upvotes and then suddenly dropped to -4 (sound familiar?). I was curious so I did some digging into the meta kaggle dataset. The kernel below reveals what I believe to be ~9 accounts associated with one user, these accounts have peculiar names (Jobs S, Cook T, and Bill G stood out to me). While the 9 accounts upvoted many kerenels, they all overlap on one user. I can't prove anything definitively, but it just seems odd.\"
\n\"I'm not trying to ruin anyone's reputation, but if these are fake accounts being used to downvote comments in mass… they should probably be removed….\"
\nhttps://www.kaggle.com/robikscube/fake-accounts/
\nI still think that downvotes should at least be Justified. As I listed in another topic.
", "rawMarkdown": "Hi Allie,\n\nEven GM are downvoted: take a look at this: \n\nhttps://www.kaggle.com/product-feedback/76394\n\n\"I've been debating making this kernel public, but given how this thread went down with all the downvoting- I figure it's worth releasing. Last week I made a comment on a kernel mentioning that much of the code was copied without giving credit to the creators. This comment had 5 upvotes and then suddenly dropped to -4 (sound familiar?). I was curious so I did some digging into the meta kaggle dataset. The kernel below reveals what I believe to be ~9 accounts associated with one user, these accounts have peculiar names (Jobs S, Cook T, and Bill G stood out to me). While the 9 accounts upvoted many kerenels, they all overlap on one user. I can't prove anything definitively, but it just seems odd.\"\n\n\"I'm not trying to ruin anyone's reputation, but if these are fake accounts being used to downvote comments in mass… they should probably be removed….\"\n\nhttps://www.kaggle.com/robikscube/fake-accounts/\n\nI still think that downvotes should at least be Justified. As I listed in another topic.", "votes": null }, { "id": "1722677", "postDate": "03/14/2022 19:13:19", "content": "Hi Marilia,
\nthank you for your response and the links.
\nBut in the end downvotes are the things which bother me least. As I explained it in my comment there are more serious issues which hurt me nearly every time when I go to Kaggle.
That imbalance is not a \"privilege\" of this Platform. It just reproduces what we see on our professional world (thanks Not mine on both professions).
\nHowever, women give up opportunities since some achieved GM positions and didn't try a single EDA on WiDS (Women in Data Science) Competition or any Competiton. It's their option, though their participation could incentivate more women to join competitions too. And since they stop to Publish things, they lose relevance too.
\nBesides, Kaggle is a Platform with many paths to our projects but it' s still seen as a Competition site. Therefore GM MALE Competitors are worshiped by young male students. Since they long to be like them. Which is exactly what occurs in many professions.
\nIf any woman intend to get recognition (the Real recognition) she should participate actively on Competitions, reach top positions, make discussion topics about Metrics/methods. Not just saying congrats to anyone or spreading links. And even there, she will be downvoted by some angry bird that thinks she doesn't belong there :)
\nThough there are plenty of friendly, supportive, skilled, talented kagglers that behave with maturity.
\nThose users are who we choose to share, learn and improve skills. They make our participation on Kaggle a pleasure. What makes worthy to log in daily.
", "rawMarkdown": "That imbalance is not a \"privilege\" of this Platform. It just reproduces what we see on our professional world (thanks Not mine on both professions).\n\nHowever, women give up opportunities since some achieved GM positions and didn't try a single EDA on WiDS (Women in Data Science) Competition or any Competiton. It's their option, though their participation could incentivate more women to join competitions too. And since they stop to Publish things, they lose relevance too. \n\nBesides, Kaggle is a Platform with many paths to our projects but it' s still seen as a Competition site. Therefore GM MALE Competitors are worshiped by young male students. Since they long to be like them. Which is exactly what occurs in many professions.\n\nIf any woman intend to get recognition (the Real recognition) she should participate actively on Competitions, reach top positions, make discussion topics about Metrics/methods. Not just saying congrats to anyone or spreading links. And even there, she will be downvoted by some angry bird that thinks she doesn't belong there :) \n\nThough there are plenty of friendly, supportive, skilled, talented kagglers that behave with maturity. \n\nThose users are who we choose to share, learn and improve skills. They make our participation on Kaggle a pleasure. What makes worthy to log in daily.", "votes": null }, { "id": "1728074", "postDate": "03/18/2022 15:07:12", "content": "Marilia,
\nwith all the respect let me strongly disagree with you.
\nEither of us views the situation from different position and with different experience from different Kaggle parts. I would never speak about notebooks or datasets as I am not deeply involved in this part of Kaggle and my view would be only superficial as your view of competitions' environment probably is. And I would never tell anybody what she should do to become \"recognized\" in this different part. BTW I don't seek any recognition, I seek only fairness, justice and mutual respect.
\nSuch a superficial view can downplay the serious issues which I am trying to raise and address to Kaggle staff.
\nAnd I don't speak about general gender imbalance on Kaggle, this isn't so dramatic, I speak about imbalance closer to the top of competitions' rankings.
\nIf something isn't unique, it doesn't mean that it is correct. We live in the year 2022.
@wcukierski
\nTo repeat the main question to Mr. Cukierski.
Data show that apparently there are some insurmountable obstacles preventing vast majority of female competitors from reaching a long-term success in competitions' ranking.
\nIn my previous comments I showed possible reasons and solutions. So I ask again: how will Kaggle address this real inequality and the issues connected with fairness, justice and mutual respect?
Do you seriously think that WiDS competitions which you directly organize as second class competitions can make any positive difference in this problem? Of course I fully respect those who decide to participate, but I don't want to be segregated on some special playground \"for girls only\" and with no topics choice. I am an equal human being and I view all these actions for women as in fact rather humiliating than encouraging.
\nAs if the organizing companies (including Kaggle) were saying to us - well, girls, you are not good enough to compete with boys, so we had to prepare something (of second class) especially for you. Consequently it also supports some male competitors' opinion that big competitions are only for them and girls should go only to their designated playground.
@blankaf I have worked with the WiDS organizers for 5 years and they, like all of our hosts, come to Kaggle with their own goals for success. The WiDS organizers each year seek to inspire new entrants by crafting short, accessible, competitions to inspire teamwork and learning. WiDS competitions have a higher than average number of participants who are new to Kaggle and their satisfaction surveys (conducted by WiDS) are consistently high. The WiDS Datathon has grown every year and we fully support their mission and methods.
", "rawMarkdown": "blankaf I have worked with the WiDS organizers for 5 years and they, like all of our hosts, come to Kaggle with their own goals for success. The WiDS organizers each year seek to inspire new entrants by crafting short, accessible, competitions to inspire teamwork and learning. WiDS competitions have a higher than average number of participants who are new to Kaggle and their satisfaction surveys (conducted by WiDS) are consistently high. The WiDS Datathon has grown every year and we fully support their mission and methods.", "votes": null }, { "id": "1746000", "postDate": "04/05/2022 13:05:19", "content": "@maggiemd , @wcukierski
\nMy reply is a bit delayed because, though you don't seem to pay any attention to it, there is a war nearly at our doorstep and helping war refugees takes some time.
Anyway Maggie, your comment is another off-topic one. The problem here is not in attracting new female entrants but in the environment which prevents them from continuing for a longer time.
\nA satisfaction survey made after the first experience presenting only the sunny side of Kaggle can do more harm than good for the life of female newcomers if not accompanied with additional warnings for their future activity.
\nThe survey should also contain such questions as:
\n\"Do you plan to continue in competing on Kaggle?\" - If the answer is \"yes\", then:
\n\"Are you prepared to turn a blind eye to injustice and unfairness of any kind and to ignore abusive behaviour?\" If the answer is \"no\", then:
\n\"Especially if you are from a country not heavily present on Kaggle, are you prepared to face various kinds of unscrupulous personal attacks, jeering and insults without any chance for help?\"
Yes, proclamations sound nice and don't cost anything but the reality here is different.
\nThe girls should know, what to expect if they eventually decide to join the real world of Kaggle competitions, which choices they will stand before and what kind of decisions touching even their personal lives they will have to make.
\nAnd as I can see from Kaggle staff reactions there is hardly any chance for change. Simply nobody seems to care.
Or am I wrong Mr. Cukierski and I will finally after 3 weeks get an answer to my main questions formulated above and supported by the statistics?
\nBTW the data about female competitors presence in TOP 100 got even worse in the meantime. The number is one big zero now.
\nYour silence might look like you suppose me and before all the presented problem to be unworthy your reaction.
\nDoes Kaggle need only high absolute numbers of participants and newly recruited paid Colab services users or do some other numbers matter as well? Who does Kaggle in fact support and who is disregarded? Does Kaggle deserve its reputation at all?
Hi Allie,
\nWe stand by the community guidelines referenced above and reiterate, in strongest terms possible, our aim for people of any background to feel welcome to participate on Kaggle. Large, online communities are imperfect places on their best days. Both staff and the community have a responsibility to focus on the ideas shared and not the person behind them. You are encouraged to continue to call out issues you see with respect to fairness, justice, and equality in the Kaggle community.
\nIf you experience behavior that violates community guidelines, please either directly flag the comment or summarize the specific issue via support@kaggle.com. Reports will be kept confidential.
", "rawMarkdown": "Hi Allie,\n\nWe stand by the community guidelines referenced above and reiterate, in strongest terms possible, our aim for people of any background to feel welcome to participate on Kaggle. Large, online communities are imperfect places on their best days. Both staff and the community have a responsibility to focus on the ideas shared and not the person behind them. You are encouraged to continue to call out issues you see with respect to fairness, justice, and equality in the Kaggle community.\n\nIf you experience behavior that violates community guidelines, please either directly flag the comment or summarize the specific issue via support@kaggle.com. Reports will be kept confidential.", "votes": null }, { "id": "1775144", "postDate": "05/02/2022 18:28:48", "content": "@wcukierski
\nDear Mr.Cukierski,
Your words:
\n\"You are encouraged to continue to call out issues you see with respect to fairness, justice, and equality … \"
\nReality:
\nWhen I write and ask about rules, fairness and justice, even with support of statistical data, I am totally ignored by all members of Kaggle team. The final leaderboard of Happywhale stayed as it had been i.e. with the seeds of future problems which are now happily flourishing again in NBME and H&M competitions.
Kaggle is apparently a place of unlimited opportunities:
\nI have 1 more question about team names. Can anybody have any team name (or write in a discussion) in other than English language even with non-Latin alphabet? I don't know whether it is fun or provocation but it definitely shows disrespect to other Kagglers and it seems to be widely tolerated by Kaggle team. Where is the equality here?
\nThe competition rules are somewhat vague and I know that Kaggle team is the only and final authority to decide about disqualifications. Kaggle Terms of Use are carefully formulated and in fact give you unlimited power and hardly any legal responsibility.
\nBut don't you think that there is another type of responsibility in place here? I mean responsibility not only to educate but also to form characters of the very young users. When they once succeed in getting the best reward using the easiest way without any work, what will they do next time not only on Kaggle? If data science is the future of the world what perspective do such DS practitioners offer to the world?
Big evil always starts with ignorance and/or tolerance of its humble roots.
\nOne more example. Imagine that a company hires a \"triple gold competitions master\" based on his medals and tier. How happy will the company be after learning that the individual is hardly able to write a single line of code? How will it influence Kaggle reputation and its business value?
In fact I should feel satisfaction because with the latest developments every single sentence, which I have written in this thread, proves to be correct. But I am only sad and disgusted while actively observing the situation heading to similarly scandalous endings as it was 2 weeks ago in Happywhale now in the 2 currently finishing competitions.
\nSimply it's the time of \"instant geniuses\" taking Kaggle competitions by storm and occasionally leaving unpunished. Based on my current research 20% of teams in top 200 in NBME show clear signs of presence of \"instant geniuses\" and/or \"silent companions\" in their squads.
\nWill the justice be finally fully served?
Manual for safe downvoting of my topics
\n\"How did the strange data split (and all the above listed factors) serve to help protecting the GBR?\"
\nIn fact nearly the only what remained from the title was promotion of Tensorflow. \"
\nI'm so naive that I made some work with Pytorch :)
\nPractical solutions that I read during the comp: fork the COTS (Not the Notebooks) manually, burn and bury them.
\n\"Marine tourism operators and community groups are leading the initiative using injector kits to cull COTS, and remote villages without kits are removing COTS manually and burning or burying them. \" (in \"Controlling crown-of-thorns starfish in a small island developing nation: using collaboration and innovation to overcome resource limitations.\"
\nAuthors: Johanna Johnson, David Welch, Peter Whitelaw, Christina Shaw, Pascal Dumas,
\n, Rocky Kaku, Owen Drew, Kehana Andrew
The sponsoring TF team intends to integrate the learnings from this project into CSIRO's glider for real-time detection of COTS. This takes time--time to understand the models, time to implement, time to test, debug, refine--and so we (Kaggle) don't have an immediate answer on how this competition impacted the GBR. I certainly believe there is already a value in having thousands of people made aware of this problem and commonly focus on a solution. As researchers, you don't know where the state of the art is unless you have a lot of smart people press on it.
\nIt's healthy to ask the general question of \"couldn't this competition have been more impactful if X,Y, or Z was different\" at the end of any competition we run, and the answer is probably always going to be yes, albeit with 20/20 hindsight, unlimited time and resourcing before deadlines, all project stakeholders in perfect agreement, and absent the constraints, leakages, and imperfections real-world data brings with it :) What we've observed from running hundreds of competitions is that--in spite of the imperfections--solid, generalized models still rise to the top of challenges. We do continually listen to feedback on competition design and work to make them as fair and high quality as we can.
", "votes": null, "replies": [ { "id": 1710952, "author_name": "blankaf", "author_url": "", "post_date": "03/03/2022 13:57:47", "content": "@wcukierski
\nDear Mr. Cukierski and all your loyal supporters,
\nThank you for your highly generalizing response, though the main question \"why such a hardly understandable data arrangement especially in this competition?\" remains unanswered.
\nIn fact no answer can sometimes be the best one as it proves the rightness of doubts.
P.S.
\nIt is really funny that yesterday, at the very same moment of receiving your response, I also got 2 simultaneous downvotes which (mysteriously?) downgraded my comment from silver to bronze. And downvotes will still stay undercover …😎
\nI know very well that Kaggle Competitions' environment is nearly a strictly Big Boys Club and I, as a quite successful Czech female, can always be only an unwelcome intruder. Where are the proclaimed Kaggle Values?
\nFrankly said, at such occasions I feel like staying among little boys on a sandbox playground.😀
Edited:
\nAfter 1 hour my main comment is back on silver.🎶
Hi Allie,
\nDon't bother about the downvotes. It's Not a gender thing since male, female and prefer not to say are downvoted equally. Even the \" Big Boys\" :) are downvoted.
\nWorst that being downvoted, is being ignored.
\nSurviving on the Playground it's part of the whole process.
", "votes": null, "replies": [] }, { "id": 1711299, "author_name": "wcukierski", "author_url": "", "post_date": "03/03/2022 19:57:18", "content": "Hi Allie,
\nI am saddened to hear that you feel unwelcome here. You are welcomed, as is your constructive feedback. Our community guidelines specifically call out, \"when you have critical feedback, focus on the ideas that others are sharing, not the person\" as well as:
\n\n\nWe strive to be a community that welcomes and supports people of all backgrounds and identities. This includes, but is not limited to, members of any race, ethnicity, culture, national origin, color, immigration status, social and economic class, educational level, sex, sexual orientation, gender identity and expression, age, size, family status, political belief, religion, and mental and physical ability.
\n
Under these guidelines, upvotes and downvotes should be signals of the quality and constructiveness of the content, not personal affronts. You should continue to voice concerns if you feel like this is not what is happening.
\nRegarding my response being general: I do not feel as though the problem split was unorthodox here (public and private use the same videos split by time), but as I alluded to above, we often pick between an imperfect set of choices and do so prior to all the information that is surfaced while the competition is running. There are tens of critiqueable design decisions that feed into each launch and people willing to argue both sides of most of them. Trying to defend them publicly, with 20/20 hindsight exposing subtle differences buried within the data, can sometimes feel unproductive for us and unsatisfying for you.
", "votes": null, "replies": [] }, { "id": 1722524, "author_name": "blankaf", "author_url": "", "post_date": "03/14/2022 16:04:29", "content": "@wcukierski
\nDear Mr. Cukierski,
First only a note.
\nI didn't complain about the fact of getting downvotes, I was only pointing at the mysterious time coincidence of your response and the downvotes. At least I know, why downvotes will never be made public.
\nBack to the main topic which you opened citing the Community Guidelines.
\n\"We strive …\" - I tend to believe that you really do. But let the data speak about the bitter reality.
\nFemale Kagglers form only 1.2% in top 600 competition rankings and, as far as I know, there is only 1 female competition Grandmaster (of 248) as well as only 1 female Triple Grandmaster - my deepest respect to both of them.
\nI hope that nobody would say that women and girls are not clever enough to successfully compete with boys in DS. But do you have an acceptable explanation for such a huge imbalance? What or who plays the main role in the situation that we can participate but hardly ever be long-term successful?
\nIsn't it accidentally a consequence of the atmosphere of escalated ruthlessness, aggressiveness and disrespect for anybody and anything including Kaggle Competitions' rules which you tend to overlook?
\nSometimes some Kaggle staff members can even contribute to the feeling of unfairness and injustice - e.g. no reaction to lost parts of GPU quota due to system maintenance error, strange and highly subjective system of disqualification.
\nYou also cite: \" We’re all here to learn and share ideas.\" But for many competitors the main goal is only getting points and tiers for a prospect of carrier boost no matter how they reach it. And some \"minor\" rules violations are broadly considered more acceptable than publicly pointing at them.
For some competitors this atmosphere one day simply becomes unbearable and they quit with active kaggling. No wonder that it holds true mostly for women. And, please, don't try to only tell me that it saddens you. Proclamations not followed by an action can hardly ever help.
\nI hope that my feedback is constructive enough and it isn't necessary to add more facts or evidence.
To wrap it up:
\nI don't ask for any special treatment for anybody or anything like it. I only ask for Kaggle staff more actively supporting the environment of dignity, honesty, fairness and mutual respect as described in Community Guidelines. This could help to every decently behaving Kaggler. Kaggle Competitions should really be a platform for learning and practising DS skills not for training surviving (as Marilia noted above).
Hi Allie,
\nEven GM are downvoted: take a look at this:
\nhttps://www.kaggle.com/product-feedback/76394
\n\"I've been debating making this kernel public, but given how this thread went down with all the downvoting- I figure it's worth releasing. Last week I made a comment on a kernel mentioning that much of the code was copied without giving credit to the creators. This comment had 5 upvotes and then suddenly dropped to -4 (sound familiar?). I was curious so I did some digging into the meta kaggle dataset. The kernel below reveals what I believe to be ~9 accounts associated with one user, these accounts have peculiar names (Jobs S, Cook T, and Bill G stood out to me). While the 9 accounts upvoted many kerenels, they all overlap on one user. I can't prove anything definitively, but it just seems odd.\"
\n\"I'm not trying to ruin anyone's reputation, but if these are fake accounts being used to downvote comments in mass… they should probably be removed….\"
\nhttps://www.kaggle.com/robikscube/fake-accounts/
\nI still think that downvotes should at least be Justified. As I listed in another topic.
", "votes": null, "replies": [] }, { "id": 1722677, "author_name": "blankaf", "author_url": "", "post_date": "03/14/2022 19:13:19", "content": "Hi Marilia,
\nthank you for your response and the links.
\nBut in the end downvotes are the things which bother me least. As I explained it in my comment there are more serious issues which hurt me nearly every time when I go to Kaggle.
That imbalance is not a \"privilege\" of this Platform. It just reproduces what we see on our professional world (thanks Not mine on both professions).
\nHowever, women give up opportunities since some achieved GM positions and didn't try a single EDA on WiDS (Women in Data Science) Competition or any Competiton. It's their option, though their participation could incentivate more women to join competitions too. And since they stop to Publish things, they lose relevance too.
\nBesides, Kaggle is a Platform with many paths to our projects but it' s still seen as a Competition site. Therefore GM MALE Competitors are worshiped by young male students. Since they long to be like them. Which is exactly what occurs in many professions.
\nIf any woman intend to get recognition (the Real recognition) she should participate actively on Competitions, reach top positions, make discussion topics about Metrics/methods. Not just saying congrats to anyone or spreading links. And even there, she will be downvoted by some angry bird that thinks she doesn't belong there :)
\nThough there are plenty of friendly, supportive, skilled, talented kagglers that behave with maturity.
\nThose users are who we choose to share, learn and improve skills. They make our participation on Kaggle a pleasure. What makes worthy to log in daily.
", "votes": null, "replies": [] }, { "id": 1728074, "author_name": "blankaf", "author_url": "", "post_date": "03/18/2022 15:07:12", "content": "Marilia,
\nwith all the respect let me strongly disagree with you.
\nEither of us views the situation from different position and with different experience from different Kaggle parts. I would never speak about notebooks or datasets as I am not deeply involved in this part of Kaggle and my view would be only superficial as your view of competitions' environment probably is. And I would never tell anybody what she should do to become \"recognized\" in this different part. BTW I don't seek any recognition, I seek only fairness, justice and mutual respect.
\nSuch a superficial view can downplay the serious issues which I am trying to raise and address to Kaggle staff.
\nAnd I don't speak about general gender imbalance on Kaggle, this isn't so dramatic, I speak about imbalance closer to the top of competitions' rankings.
\nIf something isn't unique, it doesn't mean that it is correct. We live in the year 2022.
@wcukierski
\nTo repeat the main question to Mr. Cukierski.
Data show that apparently there are some insurmountable obstacles preventing vast majority of female competitors from reaching a long-term success in competitions' ranking.
\nIn my previous comments I showed possible reasons and solutions. So I ask again: how will Kaggle address this real inequality and the issues connected with fairness, justice and mutual respect?
Do you seriously think that WiDS competitions which you directly organize as second class competitions can make any positive difference in this problem? Of course I fully respect those who decide to participate, but I don't want to be segregated on some special playground \"for girls only\" and with no topics choice. I am an equal human being and I view all these actions for women as in fact rather humiliating than encouraging.
\nAs if the organizing companies (including Kaggle) were saying to us - well, girls, you are not good enough to compete with boys, so we had to prepare something (of second class) especially for you. Consequently it also supports some male competitors' opinion that big competitions are only for them and girls should go only to their designated playground.
@blankaf I have worked with the WiDS organizers for 5 years and they, like all of our hosts, come to Kaggle with their own goals for success. The WiDS organizers each year seek to inspire new entrants by crafting short, accessible, competitions to inspire teamwork and learning. WiDS competitions have a higher than average number of participants who are new to Kaggle and their satisfaction surveys (conducted by WiDS) are consistently high. The WiDS Datathon has grown every year and we fully support their mission and methods.
", "votes": null, "replies": [] }, { "id": 1746000, "author_name": "blankaf", "author_url": "", "post_date": "04/05/2022 13:05:19", "content": "@maggiemd , @wcukierski
\nMy reply is a bit delayed because, though you don't seem to pay any attention to it, there is a war nearly at our doorstep and helping war refugees takes some time.
Anyway Maggie, your comment is another off-topic one. The problem here is not in attracting new female entrants but in the environment which prevents them from continuing for a longer time.
\nA satisfaction survey made after the first experience presenting only the sunny side of Kaggle can do more harm than good for the life of female newcomers if not accompanied with additional warnings for their future activity.
\nThe survey should also contain such questions as:
\n\"Do you plan to continue in competing on Kaggle?\" - If the answer is \"yes\", then:
\n\"Are you prepared to turn a blind eye to injustice and unfairness of any kind and to ignore abusive behaviour?\" If the answer is \"no\", then:
\n\"Especially if you are from a country not heavily present on Kaggle, are you prepared to face various kinds of unscrupulous personal attacks, jeering and insults without any chance for help?\"
Yes, proclamations sound nice and don't cost anything but the reality here is different.
\nThe girls should know, what to expect if they eventually decide to join the real world of Kaggle competitions, which choices they will stand before and what kind of decisions touching even their personal lives they will have to make.
\nAnd as I can see from Kaggle staff reactions there is hardly any chance for change. Simply nobody seems to care.
Or am I wrong Mr. Cukierski and I will finally after 3 weeks get an answer to my main questions formulated above and supported by the statistics?
\nBTW the data about female competitors presence in TOP 100 got even worse in the meantime. The number is one big zero now.
\nYour silence might look like you suppose me and before all the presented problem to be unworthy your reaction.
\nDoes Kaggle need only high absolute numbers of participants and newly recruited paid Colab services users or do some other numbers matter as well? Who does Kaggle in fact support and who is disregarded? Does Kaggle deserve its reputation at all?
Hi Allie,
\nWe stand by the community guidelines referenced above and reiterate, in strongest terms possible, our aim for people of any background to feel welcome to participate on Kaggle. Large, online communities are imperfect places on their best days. Both staff and the community have a responsibility to focus on the ideas shared and not the person behind them. You are encouraged to continue to call out issues you see with respect to fairness, justice, and equality in the Kaggle community.
\nIf you experience behavior that violates community guidelines, please either directly flag the comment or summarize the specific issue via support@kaggle.com. Reports will be kept confidential.
", "votes": null, "replies": [] }, { "id": 1775144, "author_name": "blankaf", "author_url": "", "post_date": "05/02/2022 18:28:48", "content": "@wcukierski
\nDear Mr.Cukierski,
Your words:
\n\"You are encouraged to continue to call out issues you see with respect to fairness, justice, and equality … \"
\nReality:
\nWhen I write and ask about rules, fairness and justice, even with support of statistical data, I am totally ignored by all members of Kaggle team. The final leaderboard of Happywhale stayed as it had been i.e. with the seeds of future problems which are now happily flourishing again in NBME and H&M competitions.
Kaggle is apparently a place of unlimited opportunities:
\nI have 1 more question about team names. Can anybody have any team name (or write in a discussion) in other than English language even with non-Latin alphabet? I don't know whether it is fun or provocation but it definitely shows disrespect to other Kagglers and it seems to be widely tolerated by Kaggle team. Where is the equality here?
\nThe competition rules are somewhat vague and I know that Kaggle team is the only and final authority to decide about disqualifications. Kaggle Terms of Use are carefully formulated and in fact give you unlimited power and hardly any legal responsibility.
\nBut don't you think that there is another type of responsibility in place here? I mean responsibility not only to educate but also to form characters of the very young users. When they once succeed in getting the best reward using the easiest way without any work, what will they do next time not only on Kaggle? If data science is the future of the world what perspective do such DS practitioners offer to the world?
Big evil always starts with ignorance and/or tolerance of its humble roots.
\nOne more example. Imagine that a company hires a \"triple gold competitions master\" based on his medals and tier. How happy will the company be after learning that the individual is hardly able to write a single line of code? How will it influence Kaggle reputation and its business value?
In fact I should feel satisfaction because with the latest developments every single sentence, which I have written in this thread, proves to be correct. But I am only sad and disgusted while actively observing the situation heading to similarly scandalous endings as it was 2 weeks ago in Happywhale now in the 2 currently finishing competitions.
\nSimply it's the time of \"instant geniuses\" taking Kaggle competitions by storm and occasionally leaving unpunished. Based on my current research 20% of teams in top 200 in NBME show clear signs of presence of \"instant geniuses\" and/or \"silent companions\" in their squads.
\nWill the justice be finally fully served?
Manual for safe downvoting of my topics
\n