Thank you to XYZ10 and MSR for hosting this competition! This is a very interesting problem and finding models to optimizing navigation will certainly help millions (or even billions) of people in their daily life.
\nFrom the description, it's clear that user permission and cooperation is important to XYZ10 and I believe that such data will be used very responsibly by the organizers to produce impactful research. However, as other people start to use this research to build their own localization tools, some might end up using it in malicious ways; for example, hackers could use it to track unsuspicious targets, retailers would collect data on users to see if they go to competing stores, etc.
\nAll of this might be unavoidable if we want to make progress in research and build better products. However, I would love to hear about the considerations put into privacy and ethics for this type of work. As we develop such models and tools, are we also simultaneously finding ways to protect a user's identity? Is there UI/UX design recommendations for releasing such products to the public, such that non-tech savvy users are aware that they are giving permissions for \"advanced\" localization? Is there ways to maintain accurate localization through the resulting approaches while keeping the identity secret?
\nI think an excellent example of transparent and useful data \"policy\" is Google's My Activity, which lets you know what is being collected and gives you the option to disable or delete past data. I'd love to hear if such approach was tackled in previous works by the organizers, or if it's something that's already being discussed in the upcoming manuscript(s).
\n\n", "messages": [ { "id": 1175355, "postDate": "2021-01-29T05:24:30.030Z", "content": "As a follow up to this post, I shared some links about privacy-conscious localization research in this post. I invite you to take a look if you are interested about how we can address the aforementioned challenges.
\n
Thank you to XYZ10 and MSR for hosting this competition! This is a very interesting problem and finding models to optimizing navigation will certainly help millions (or even billions) of people in their daily life.
\nFrom the description, it's clear that user permission and cooperation is important to XYZ10 and I believe that such data will be used very responsibly by the organizers to produce impactful research. However, as other people start to use this research to build their own localization tools, some might end up using it in malicious ways; for example, hackers could use it to track unsuspicious targets, retailers would collect data on users to see if they go to competing stores, etc.
\nAll of this might be unavoidable if we want to make progress in research and build better products. However, I would love to hear about the considerations put into privacy and ethics for this type of work. As we develop such models and tools, are we also simultaneously finding ways to protect a user's identity? Is there UI/UX design recommendations for releasing such products to the public, such that non-tech savvy users are aware that they are giving permissions for \"advanced\" localization? Is there ways to maintain accurate localization through the resulting approaches while keeping the identity secret?
\nI think an excellent example of transparent and useful data \"policy\" is Google's My Activity, which lets you know what is being collected and gives you the option to disable or delete past data. I'd love to hear if such approach was tackled in previous works by the organizers, or if it's something that's already being discussed in the upcoming manuscript(s).
\n\n", "rawMarkdown": "Thank you to XYZ10 and MSR for hosting this competition! This is a very interesting problem and finding models to optimizing navigation will certainly help millions (or even billions) of people in their daily life.\n\nFrom the description, it's clear that user permission and cooperation is important to XYZ10 and I believe that such data will be used very responsibly by the organizers to produce impactful research. However, as other people start to use this research to build their own localization tools, some might end up using it in malicious ways; for example, hackers could use it to track unsuspicious targets, retailers would collect data on users to see if they go to competing stores, etc.\n\nAll of this might be unavoidable if we want to make progress in research and build better products. However, I would love to hear about the considerations put into privacy and ethics for this type of work. As we develop such models and tools, are we also simultaneously finding ways to protect a user's identity? Is there UI/UX design recommendations for releasing such products to the public, such that non-tech savvy users are aware that they are giving permissions for \"advanced\" localization? Is there ways to maintain accurate localization through the resulting approaches while keeping the identity secret?\n\nI think an excellent example of transparent and useful data \"policy\" is Google's [My Activity](https://myactivity.google.com/), which lets you know what is being collected and gives you the option to disable or delete past data. I'd love to hear if such approach was tackled in previous works by the organizers, or if it's something that's already being discussed in the upcoming manuscript(s). \n\n# Edits\n\n> As a follow up to this post, I shared some links about privacy-conscious localization research in [this post](https://www.kaggle.com/c/indoor-location-navigation/discussion/215979). I invite you to take a look if you are interested about how we can address the aforementioned challenges.", "votes": 79 }, { "id": 1177603, "postDate": "2021-01-30T11:53:50.797Z", "content": "As a follow up to this post, I shared some links about privacy-conscious localization research in this post. I invite you to take a look if you are interested about how we can address the aforementioned challenges.
\n
This is an important topic and this post raises the issues in a very well-considered way. Interested to hear responses.
", "rawMarkdown": "This is an important topic and this post raises the issues in a very well-considered way. Interested to hear responses.", "votes": 7 }, { "id": 1186345, "postDate": "2021-02-04T19:07:28.887Z", "content": "Thank you for posting this. I agree that the objectives of this competition raises serious potential ethical concerns. Governments in repressive countries could easily use technology developed in this competition to track and arrest political dissidents or whistleblowers.
\nFor example, suppose State Employee X meets with Newspaper Reporter Y at a food court in a shopping mall to \"blow the whistle\" about government human rights abuses, but their smartphone locations are tracked by the government. Individuals X and Y are known to be in close proximity, and can be swiftly arrested (perhaps after an automated alert from an AI system).
\nRepressive governments are already known to collect smartphone data without user permission, or install secret data collection \"backdoors\" into apps, so developing this technology assuming it will only be used on consenting users does not seem reasonable.
\nWith so many other opportunities on Kaggle to use AI to do things that we all agree are for the good of humanity, like curing diseases, reducing medical errors, and improving environmental conservation, it's hard to see how this competition fits into the broader mission of using AI for good.
\nI'm disappointed there was not a more careful consideration of the ethical issues raised by this competition before it was launched.
", "rawMarkdown": "Thank you for posting this. I agree that the objectives of this competition raises serious potential ethical concerns. Governments in repressive countries could easily use technology developed in this competition to track and arrest political dissidents or whistleblowers. \n\nFor example, suppose State Employee X meets with Newspaper Reporter Y at a food court in a shopping mall to \"blow the whistle\" about government human rights abuses, but their smartphone locations are tracked by the government. Individuals X and Y are known to be in close proximity, and can be swiftly arrested (perhaps after an automated alert from an AI system). \n\nRepressive governments are already known to collect smartphone data without user permission, or install secret data collection \"backdoors\" into apps, so developing this technology assuming it will only be used on consenting users does not seem reasonable. \n\nWith so many other opportunities on Kaggle to use AI to do things that we all agree are for the good of humanity, like curing diseases, reducing medical errors, and improving environmental conservation, it's hard to see how this competition fits into the broader mission of using AI for good. \n\nI'm disappointed there was not a more careful consideration of the ethical issues raised by this competition before it was launched. ", "votes": 4, "replies": [ { "id": 1186550, "postDate": "2021-02-04T22:16:02.760Z", "content": "@johnjdavisiv We did discuss this extensively as a team internally, along with various other teams (legal, AI principles), and I personally pressed on both the pros and cons of launching this competition. There are a considerable number of positive use cases for this type of technology (safety, consumer navigation, etc.) To your point, though, it would be naive to believe it could not be exploited by bad actors. Unfortunately, this is true of much of ML and AI.
\nWe absolutely respect that individuals in the community will have different levels of comfort working on various technologies, as well as for various competitions we host on Kaggle, and encourage each Kaggler to carefully consider the actions of their work, not only in the competitions here, but also professionally.
\nThanks for the feedback and adding to the discussion.
", "rawMarkdown": "@johnjdavisiv We did discuss this extensively as a team internally, along with various other teams (legal, AI principles), and I personally pressed on both the pros and cons of launching this competition. There are a considerable number of positive use cases for this type of technology (safety, consumer navigation, etc.) To your point, though, it would be naive to believe it could not be exploited by bad actors. Unfortunately, this is true of much of ML and AI.\n\nWe absolutely respect that individuals in the community will have different levels of comfort working on various technologies, as well as for various competitions we host on Kaggle, and encourage each Kaggler to carefully consider the actions of their work, not only in the competitions here, but also professionally.\n\nThanks for the feedback and adding to the discussion.\n", "votes": 7 }, { "id": 1188592, "postDate": "2021-02-06T11:15:11.413Z", "content": "I think you make some great points. As @inversion implied, much of ML and AI is a double edged sword that can be used for good and bad purposes.
\nExample:
\nThe ability to track everyone, anywhere, anytime, would have massive dystopian consequences: making autocratic governments virtually invincible and allowing any nation to become a strict police state overnight.
\nHowever, this ability would also have profound advantages: essentially no unsolved crimes such as theft, murders and rapes and no more missing children.
\nI'm not arrogant enough to try and say whether it's worth it or not, but it is something people will likely have to consider in the semi-near future.
As for this competition, there is already the ability to track people, as long as they have a smartphone for example. This competition is about predicting where people will be. This will the tracking process easier for people but will not change the underlying ethics or the motives of the people using these technologies.
\nI think the ideal outcome is a balance between people's privacy and security, and the commercial interests of businesses, and legal interests of governments, without any of these factors becoming dominant.
It's great to see that increasing amount of concern into people's privacy. But I do want to put a thumb down on the example you putted here. Without any ML work, the GPS in your phone can already do the exact thing, every byte of your info is there, transmitted via a government owned satellite and potentially government controlled telecom. There are some question marks on the backdoor, so plz, ML & AL & Kaggle does not have a play in that area.
I agree. Recently I had a job interview in one company, which is tracking big shopping Malls visitors in China using their special tools for positioning. The task is sending targeted ads to people based on their position in the shop. Let's say you visited jewelry shop and didn't buy anything, but in two days based on where you were standing the company sends a reminder about the product which you were looking at. At the beginning I thought it's cool, but later I imagined myself being the person receiving such advertising. A bit creepy. Important note is that they can't do business in Europe, only in China where privacy is hard to obtain.
\nI suppose, that just a small share of people would agree on giving info about their position even in this case, so for business needs companies will definitely try to incline clients to this decision or just will put a small note about it in their long privacy agreements.
From a tech perspective it is quite cool that this can be done but I agree from a privacy point of view it could be a bit creepy.
", "rawMarkdown": "From a tech perspective it is quite cool that this can be done but I agree from a privacy point of view it could be a bit creepy.", "votes": 1 } ] }, { "id": 1177644, "postDate": "2021-01-30T12:24:32.490Z", "content": "Just imagine if thiefs could track when your house is empty and they can \"freely\" enter… This is indeed a delicate topic.
", "rawMarkdown": "Just imagine if thiefs could track when your house is empty and they can \"freely\" enter... This is indeed a delicate topic.", "votes": 2, "replies": [ { "id": 1178288, "postDate": "2021-01-30T19:22:54.940Z", "content": "Thieves could probably already do it without fine-grained localization; i.e. they only need to know you are at a mall rather than were exactly.
", "rawMarkdown": "Thieves could probably already do it without fine-grained localization; i.e. they only need to know you are at a mall rather than were exactly. ", "votes": 3 } ] }, { "id": 1212446, "postDate": "2021-02-21T08:27:44.440Z", "content": "Really happy to see that these concerns are being raised.
\nIn my opinion, it's good to have competitions that fall in the \"grey area\" of ML & AI use-cases, because I much prefer the solution be built by a global community with various perspectives, in a collaborative space, instead of it being developed in a closed environment or a single company. This gives us a chance to gain exposure to the gray area use case and be informed. Also, there is a requirement that the winning team should share their solution, so we get to make sure that the best solution doesn't have any blatant privacy concerns.
\nAnd these concerns exist for most of the use cases we work in this field, every SOTA model developed for medical imaging or Imagenet has the potential to advance unethical use of face recognition, etc.
", "rawMarkdown": "Really happy to see that these concerns are being raised.\n\nIn my opinion, it's good to have competitions that fall in the \"grey area\" of ML & AI use-cases, because I much prefer the solution be built by a global community with various perspectives, in a collaborative space, instead of it being developed in a closed environment or a single company. This gives us a chance to gain exposure to the gray area use case and be informed. Also, there is a requirement that the winning team should share their solution, so we get to make sure that the best solution doesn't have any blatant privacy concerns.\n\nAnd these concerns exist for most of the use cases we work in this field, every SOTA model developed for medical imaging or Imagenet has the potential to advance unethical use of face recognition, etc.", "votes": 1, "replies": [ { "id": 1212521, "postDate": "2021-02-21T09:48:29.713Z", "rawMarkdown": "", "votes": 1, "isDeleted": true } ] }, { "id": 1207750, "postDate": "2021-02-18T02:48:40.510Z", "content": "Everything has twos side, a knife can be used to cook or can be used to kill human being, so we don't make knife anymore. Or like car speed, the higher speed the higher risk.
\nMaybe it's not because of the tool, It's how we use it.
I thinks it's very hard to have a correct answer.
", "rawMarkdown": "Everything has twos side, a knife can be used to cook or can be used to kill human being, so we don't make knife anymore. Or like car speed, the higher speed the higher risk.\nMaybe it's not because of the tool, It's how we use it.\n\nI thinks it's very hard to have a correct answer.", "votes": 1 }, { "id": 1181561, "postDate": "2021-02-02T01:19:45.580Z", "content": "Agreed. This objective is incredibly creepy and disturbing. US does have consent laws, but private companies get around that by requiring opt-in (e.g. \"pay\" to use as in Facebook Pixel background tracking and Google advertising tracker). We can't stop the evolution of creepy marketing and tracking. The best thing is to be vocal publicly about what companies are doing and why you object. That's the only way to keep them in check.
", "rawMarkdown": "Agreed. This objective is incredibly creepy and disturbing. US does have consent laws, but private companies get around that by requiring opt-in (e.g. \"pay\" to use as in Facebook Pixel background tracking and Google advertising tracker). We can't stop the evolution of creepy marketing and tracking. The best thing is to be vocal publicly about what companies are doing and why you object. That's the only way to keep them in check.", "votes": 1 }, { "id": 1218619, "postDate": "2021-02-26T03:57:15.997Z", "content": "Just curious, what exactly is the evidence that \"user permission and cooperation is important to XYZ10\"? The contest description says the data is collected \"with the cooperation of the user.\" Could the hosts elaborate on what this means? I'm assuming that the \"user\" is the participant/labeler, not everyone who could be tracked by this technology. Later, the description says XYZ10 has been \"accumulating privacy-sensitive\" data. But again, does this mean that only the ground truth labels (and corresponding features) were taken with consent, or that the features of any person could only be collected with consent going forward into deployment? I'm assuming it's just the former.
\nAfter poking around a little on (a crappy google translation of) xyz10.com, I couldn't find a single mention of user consent or privacy, and in fact the descriptions of the use cases certainly seem like folks could be tracked without consenting. So has XYZ10 done anything to actually demonstrate that they care about this issue?
\nI hope I'm wrong, and I may still participate. This seems like an interesting problem, and as many have pointed out, there could be socially beneficial applications as well.
", "rawMarkdown": "Just curious, what exactly is the evidence that \"user permission and cooperation is important to XYZ10\"? The contest description says the data is collected \"with the cooperation of the user.\" Could the hosts elaborate on what this means? I'm assuming that the \"user\" is the participant/labeler, not everyone who could be tracked by this technology. Later, the description says XYZ10 has been \"accumulating privacy-sensitive\" data. But again, does this mean that only the ground truth labels (and corresponding features) were taken with consent, or that the features of any person could only be collected with consent going forward into deployment? I'm assuming it's just the former.\n\nAfter poking around a little on (a crappy google translation of) xyz10.com, I couldn't find a single mention of user consent or privacy, and in fact the descriptions of the use cases certainly seem like folks could be tracked without consenting. So has XYZ10 done anything to actually demonstrate that they care about this issue?\n\nI hope I'm wrong, and I may still participate. This seems like an interesting problem, and as many have pointed out, there could be socially beneficial applications as well.", "replies": [ { "id": 1218914, "postDate": "2021-02-26T09:30:08.240Z", "content": "Here is a video from the organizers to explain how data is acquired. It seems that they paid surveyors to collect the data.
", "rawMarkdown": "Here is a [video](https://www.youtube.com/watch?v=xt3OzMC-XMU) from the organizers to explain how data is acquired. It seems that they paid surveyors to collect the data. " }, { "id": 1219167, "postDate": "2021-02-26T13:54:23.913Z", "content": "Right, but I think most of the concerns people are raising go beyond the surveyors. Once the model is deployed, they will be tracking everyone in the building (no waypoints needed thanks to the ML) and it is less than clear whether folks will have to opt in to being tracked. And it seems very plausible that a person could opt in to providing WiFi/magnetic/accelerometer info without fully understanding that this allows them to be tracked to within a few feet.
\nA more comprehensive approach to privacy would, for example, commit to a transparent opt-in interface with plain language, so that ‘users’ (not just surveyors) would only be tracked if they actually understood that they are being tracked. I’m thinking of how iPhones are now fairly transparent about apps using gps to track your location, and requiring opt in.
", "rawMarkdown": "Right, but I think most of the concerns people are raising go beyond the surveyors. Once the model is deployed, they will be tracking everyone in the building (no waypoints needed thanks to the ML) and it is less than clear whether folks will have to opt in to being tracked. And it seems very plausible that a person could opt in to providing WiFi/magnetic/accelerometer info without fully understanding that this allows them to be tracked to within a few feet.\n\nA more comprehensive approach to privacy would, for example, commit to a transparent opt-in interface with plain language, so that ‘users’ (not just surveyors) would only be tracked if they actually understood that they are being tracked. I’m thinking of how iPhones are now fairly transparent about apps using gps to track your location, and requiring opt in. ", "votes": 1 } ] }, { "id": 1195684, "postDate": "2021-02-11T02:19:38.010Z", "content": "Hey, Just an Idea !! Cant, we make a notebook on how to avoid being tracked public places!! . Since we are data scientists we can get to know which data these companies use to track us in public places!! So what if we make a manul/Notebook to make people undertand and avoid(if they want to ) these \"LEGAL Stalkers\"?? We can discuss more on this!!
", "rawMarkdown": "Hey, Just an Idea !! Cant, we make a notebook on how to avoid being tracked public places!! . Since we are data scientists we can get to know which data these companies use to track us in public places!! So what if we make a manul/Notebook to make people undertand and avoid(if they want to ) these \"LEGAL Stalkers\"?? We can discuss more on this!!" }, { "id": 1244206, "postDate": "2021-03-18T19:09:11.853Z", "rawMarkdown": "", "votes": 1, "isDeleted": true }, { "id": 1196981, "postDate": "2021-02-11T20:25:52.387Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 1177603, "author_name": "MichaelP", "author_url": "", "post_date": "2021-01-30T11:53:50.797000", "content": "This is an important topic and this post raises the issues in a very well-considered way. Interested to hear responses.
", "votes": 7, "replies": [] }, { "id": 1186345, "author_name": "John Davis", "author_url": "", "post_date": "2021-02-04T19:07:28.887000", "content": "Thank you for posting this. I agree that the objectives of this competition raises serious potential ethical concerns. Governments in repressive countries could easily use technology developed in this competition to track and arrest political dissidents or whistleblowers.
\nFor example, suppose State Employee X meets with Newspaper Reporter Y at a food court in a shopping mall to \"blow the whistle\" about government human rights abuses, but their smartphone locations are tracked by the government. Individuals X and Y are known to be in close proximity, and can be swiftly arrested (perhaps after an automated alert from an AI system).
\nRepressive governments are already known to collect smartphone data without user permission, or install secret data collection \"backdoors\" into apps, so developing this technology assuming it will only be used on consenting users does not seem reasonable.
\nWith so many other opportunities on Kaggle to use AI to do things that we all agree are for the good of humanity, like curing diseases, reducing medical errors, and improving environmental conservation, it's hard to see how this competition fits into the broader mission of using AI for good.
\nI'm disappointed there was not a more careful consideration of the ethical issues raised by this competition before it was launched.
", "votes": 4, "replies": [ { "id": 1186550, "author_name": "inversion", "author_url": "", "post_date": "2021-02-04T22:16:02.760000", "content": "@johnjdavisiv We did discuss this extensively as a team internally, along with various other teams (legal, AI principles), and I personally pressed on both the pros and cons of launching this competition. There are a considerable number of positive use cases for this type of technology (safety, consumer navigation, etc.) To your point, though, it would be naive to believe it could not be exploited by bad actors. Unfortunately, this is true of much of ML and AI.
\nWe absolutely respect that individuals in the community will have different levels of comfort working on various technologies, as well as for various competitions we host on Kaggle, and encourage each Kaggler to carefully consider the actions of their work, not only in the competitions here, but also professionally.
\nThanks for the feedback and adding to the discussion.
", "votes": 7, "replies": [] }, { "id": 1188592, "author_name": "Jared Savage", "author_url": "", "post_date": "2021-02-06T11:15:11.413000", "content": "I think you make some great points. As @inversion implied, much of ML and AI is a double edged sword that can be used for good and bad purposes.
\nExample:
\nThe ability to track everyone, anywhere, anytime, would have massive dystopian consequences: making autocratic governments virtually invincible and allowing any nation to become a strict police state overnight.
\nHowever, this ability would also have profound advantages: essentially no unsolved crimes such as theft, murders and rapes and no more missing children.
\nI'm not arrogant enough to try and say whether it's worth it or not, but it is something people will likely have to consider in the semi-near future.
As for this competition, there is already the ability to track people, as long as they have a smartphone for example. This competition is about predicting where people will be. This will the tracking process easier for people but will not change the underlying ethics or the motives of the people using these technologies.
\nI think the ideal outcome is a balance between people's privacy and security, and the commercial interests of businesses, and legal interests of governments, without any of these factors becoming dominant.
It's great to see that increasing amount of concern into people's privacy. But I do want to put a thumb down on the example you putted here. Without any ML work, the GPS in your phone can already do the exact thing, every byte of your info is there, transmitted via a government owned satellite and potentially government controlled telecom. There are some question marks on the backdoor, so plz, ML & AL & Kaggle does not have a play in that area.
I agree. Recently I had a job interview in one company, which is tracking big shopping Malls visitors in China using their special tools for positioning. The task is sending targeted ads to people based on their position in the shop. Let's say you visited jewelry shop and didn't buy anything, but in two days based on where you were standing the company sends a reminder about the product which you were looking at. At the beginning I thought it's cool, but later I imagined myself being the person receiving such advertising. A bit creepy. Important note is that they can't do business in Europe, only in China where privacy is hard to obtain.
\nI suppose, that just a small share of people would agree on giving info about their position even in this case, so for business needs companies will definitely try to incline clients to this decision or just will put a small note about it in their long privacy agreements.
From a tech perspective it is quite cool that this can be done but I agree from a privacy point of view it could be a bit creepy.
", "votes": 1, "replies": [] } ] }, { "id": 1177644, "author_name": "Imanol Recio Erquicia", "author_url": "", "post_date": "2021-01-30T12:24:32.490000", "content": "Just imagine if thiefs could track when your house is empty and they can \"freely\" enter… This is indeed a delicate topic.
", "votes": 2, "replies": [ { "id": 1178288, "author_name": "xhlulu", "author_url": "", "post_date": "2021-01-30T19:22:54.940000", "content": "Thieves could probably already do it without fine-grained localization; i.e. they only need to know you are at a mall rather than were exactly.
", "votes": 3, "replies": [] } ] }, { "id": 1212446, "author_name": "GauthamKumaran", "author_url": "", "post_date": "2021-02-21T08:27:44.440000", "content": "Really happy to see that these concerns are being raised.
\nIn my opinion, it's good to have competitions that fall in the \"grey area\" of ML & AI use-cases, because I much prefer the solution be built by a global community with various perspectives, in a collaborative space, instead of it being developed in a closed environment or a single company. This gives us a chance to gain exposure to the gray area use case and be informed. Also, there is a requirement that the winning team should share their solution, so we get to make sure that the best solution doesn't have any blatant privacy concerns.
\nAnd these concerns exist for most of the use cases we work in this field, every SOTA model developed for medical imaging or Imagenet has the potential to advance unethical use of face recognition, etc.
", "votes": 1, "replies": [ { "id": 1212521, "author_name": "", "author_url": "", "post_date": "2021-02-21T09:48:29.713000", "content": "", "votes": 1, "replies": [] } ] }, { "id": 1207750, "author_name": "Do Phuc", "author_url": "", "post_date": "2021-02-18T02:48:40.510000", "content": "Everything has twos side, a knife can be used to cook or can be used to kill human being, so we don't make knife anymore. Or like car speed, the higher speed the higher risk.
\nMaybe it's not because of the tool, It's how we use it.
I thinks it's very hard to have a correct answer.
", "votes": 1, "replies": [] }, { "id": 1181561, "author_name": "LDMTWO", "author_url": "", "post_date": "2021-02-02T01:19:45.580000", "content": "Agreed. This objective is incredibly creepy and disturbing. US does have consent laws, but private companies get around that by requiring opt-in (e.g. \"pay\" to use as in Facebook Pixel background tracking and Google advertising tracker). We can't stop the evolution of creepy marketing and tracking. The best thing is to be vocal publicly about what companies are doing and why you object. That's the only way to keep them in check.
", "votes": 1, "replies": [] }, { "id": 1218619, "author_name": "Paul Fornia", "author_url": "", "post_date": "2021-02-26T03:57:15.997000", "content": "Just curious, what exactly is the evidence that \"user permission and cooperation is important to XYZ10\"? The contest description says the data is collected \"with the cooperation of the user.\" Could the hosts elaborate on what this means? I'm assuming that the \"user\" is the participant/labeler, not everyone who could be tracked by this technology. Later, the description says XYZ10 has been \"accumulating privacy-sensitive\" data. But again, does this mean that only the ground truth labels (and corresponding features) were taken with consent, or that the features of any person could only be collected with consent going forward into deployment? I'm assuming it's just the former.
\nAfter poking around a little on (a crappy google translation of) xyz10.com, I couldn't find a single mention of user consent or privacy, and in fact the descriptions of the use cases certainly seem like folks could be tracked without consenting. So has XYZ10 done anything to actually demonstrate that they care about this issue?
\nI hope I'm wrong, and I may still participate. This seems like an interesting problem, and as many have pointed out, there could be socially beneficial applications as well.
", "votes": 0, "replies": [ { "id": 1218914, "author_name": "Jiwei Liu", "author_url": "", "post_date": "2021-02-26T09:30:08.240000", "content": "Here is a video from the organizers to explain how data is acquired. It seems that they paid surveyors to collect the data.
", "votes": 0, "replies": [] }, { "id": 1219167, "author_name": "Paul Fornia", "author_url": "", "post_date": "2021-02-26T13:54:23.913000", "content": "Right, but I think most of the concerns people are raising go beyond the surveyors. Once the model is deployed, they will be tracking everyone in the building (no waypoints needed thanks to the ML) and it is less than clear whether folks will have to opt in to being tracked. And it seems very plausible that a person could opt in to providing WiFi/magnetic/accelerometer info without fully understanding that this allows them to be tracked to within a few feet.
\nA more comprehensive approach to privacy would, for example, commit to a transparent opt-in interface with plain language, so that ‘users’ (not just surveyors) would only be tracked if they actually understood that they are being tracked. I’m thinking of how iPhones are now fairly transparent about apps using gps to track your location, and requiring opt in.
", "votes": 1, "replies": [] } ] }, { "id": 1195684, "author_name": "A/C", "author_url": "", "post_date": "2021-02-11T02:19:38.010000", "content": "Hey, Just an Idea !! Cant, we make a notebook on how to avoid being tracked public places!! . Since we are data scientists we can get to know which data these companies use to track us in public places!! So what if we make a manul/Notebook to make people undertand and avoid(if they want to ) these \"LEGAL Stalkers\"?? We can discuss more on this!!
", "votes": 0, "replies": [] }, { "id": 1244206, "author_name": "", "author_url": "", "post_date": "2021-03-18T19:09:11.853000", "content": "", "votes": 1, "replies": [] }, { "id": 1196981, "author_name": "", "author_url": "", "post_date": "2021-02-11T20:25:52.387000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1175355": "Thank you to XYZ10 and MSR for hosting this competition! This is a very interesting problem and finding models to optimizing navigation will certainly help millions (or even billions) of people in their daily life.\n\nFrom the description, it's clear that user permission and cooperation is important to XYZ10 and I believe that such data will be used very responsibly by the organizers to produce impactful research. However, as other people start to use this research to build their own localization tools, some might end up using it in malicious ways; for example, hackers could use it to track unsuspicious targets, retailers would collect data on users to see if they go to competing stores, etc.\n\nAll of this might be unavoidable if we want to make progress in research and build better products. However, I would love to hear about the considerations put into privacy and ethics for this type of work. As we develop such models and tools, are we also simultaneously finding ways to protect a user's identity? Is there UI/UX design recommendations for releasing such products to the public, such that non-tech savvy users are aware that they are giving permissions for \"advanced\" localization? Is there ways to maintain accurate localization through the resulting approaches while keeping the identity secret?\n\nI think an excellent example of transparent and useful data \"policy\" is Google's [My Activity](https://myactivity.google.com/), which lets you know what is being collected and gives you the option to disable or delete past data. I'd love to hear if such approach was tackled in previous works by the organizers, or if it's something that's already being discussed in the upcoming manuscript(s). \n\n# Edits\n\n> As a follow up to this post, I shared some links about privacy-conscious localization research in [this post](https://www.kaggle.com/c/indoor-location-navigation/discussion/215979). I invite you to take a look if you are interested about how we can address the aforementioned challenges.", "1177603": "This is an important topic and this post raises the issues in a very well-considered way. Interested to hear responses.", "1186345": "Thank you for posting this. I agree that the objectives of this competition raises serious potential ethical concerns. Governments in repressive countries could easily use technology developed in this competition to track and arrest political dissidents or whistleblowers. \n\nFor example, suppose State Employee X meets with Newspaper Reporter Y at a food court in a shopping mall to \"blow the whistle\" about government human rights abuses, but their smartphone locations are tracked by the government. Individuals X and Y are known to be in close proximity, and can be swiftly arrested (perhaps after an automated alert from an AI system). \n\nRepressive governments are already known to collect smartphone data without user permission, or install secret data collection \"backdoors\" into apps, so developing this technology assuming it will only be used on consenting users does not seem reasonable. \n\nWith so many other opportunities on Kaggle to use AI to do things that we all agree are for the good of humanity, like curing diseases, reducing medical errors, and improving environmental conservation, it's hard to see how this competition fits into the broader mission of using AI for good. \n\nI'm disappointed there was not a more careful consideration of the ethical issues raised by this competition before it was launched. ", "1178093": "I agree. Recently I had a job interview in one company, which is tracking big shopping Malls visitors in China using their special tools for positioning. The task is sending targeted ads to people based on their position in the shop. Let's say you visited jewelry shop and didn't buy anything, but in two days based on where you were standing the company sends a reminder about the product which you were looking at. At the beginning I thought it's cool, but later I imagined myself being the person receiving such advertising. A bit creepy. Important note is that they can't do business in Europe, only in China where privacy is hard to obtain. \nI suppose, that just a small share of people would agree on giving info about their position even in this case, so for business needs companies will definitely try to incline clients to this decision or just will put a small note about it in their long privacy agreements.", "1177644": "Just imagine if thiefs could track when your house is empty and they can \"freely\" enter... This is indeed a delicate topic.", "1212446": "Really happy to see that these concerns are being raised.\n\nIn my opinion, it's good to have competitions that fall in the \"grey area\" of ML & AI use-cases, because I much prefer the solution be built by a global community with various perspectives, in a collaborative space, instead of it being developed in a closed environment or a single company. This gives us a chance to gain exposure to the gray area use case and be informed. Also, there is a requirement that the winning team should share their solution, so we get to make sure that the best solution doesn't have any blatant privacy concerns.\n\nAnd these concerns exist for most of the use cases we work in this field, every SOTA model developed for medical imaging or Imagenet has the potential to advance unethical use of face recognition, etc.", "1207750": "Everything has twos side, a knife can be used to cook or can be used to kill human being, so we don't make knife anymore. Or like car speed, the higher speed the higher risk.\nMaybe it's not because of the tool, It's how we use it.\n\nI thinks it's very hard to have a correct answer.", "1181561": "Agreed. This objective is incredibly creepy and disturbing. US does have consent laws, but private companies get around that by requiring opt-in (e.g. \"pay\" to use as in Facebook Pixel background tracking and Google advertising tracker). We can't stop the evolution of creepy marketing and tracking. The best thing is to be vocal publicly about what companies are doing and why you object. That's the only way to keep them in check.", "1218619": "Just curious, what exactly is the evidence that \"user permission and cooperation is important to XYZ10\"? The contest description says the data is collected \"with the cooperation of the user.\" Could the hosts elaborate on what this means? I'm assuming that the \"user\" is the participant/labeler, not everyone who could be tracked by this technology. Later, the description says XYZ10 has been \"accumulating privacy-sensitive\" data. But again, does this mean that only the ground truth labels (and corresponding features) were taken with consent, or that the features of any person could only be collected with consent going forward into deployment? I'm assuming it's just the former.\n\nAfter poking around a little on (a crappy google translation of) xyz10.com, I couldn't find a single mention of user consent or privacy, and in fact the descriptions of the use cases certainly seem like folks could be tracked without consenting. So has XYZ10 done anything to actually demonstrate that they care about this issue?\n\nI hope I'm wrong, and I may still participate. This seems like an interesting problem, and as many have pointed out, there could be socially beneficial applications as well.", "1195684": "Hey, Just an Idea !! Cant, we make a notebook on how to avoid being tracked public places!! . Since we are data scientists we can get to know which data these companies use to track us in public places!! So what if we make a manul/Notebook to make people undertand and avoid(if they want to ) these \"LEGAL Stalkers\"?? We can discuss more on this!!", "1244206": "", "1196981": "" } }