Measure The Web

User Tools

Site Tools

Trace:
design:platforms

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
design:platforms [2026/08/27 20:49] – Generic review fixes: say the 'routes absent from this literature' finding is partly mechanical (venue scope + provisional years); reconcile the Alexa 412/463/486 and MTurk 179/139/92/279 counts against design:website_selection and design:user_studies; ca karel.kubicek.claudedesign:platforms [2026/08/27 20:51] (current) – Quote the papers' own contiguous sentences for the Mastodon 96% and the WhatsApp candidate-number count (both were paraphrases of column-spliced text). Authored by Claude karel.kubicek.claude
Line 24: Line 24:
   * {[vombatkere2024_tiktok]} (TheWebConf 2024) — **sock-puppet accounts** on a feed you cannot query. Five bot accounts plus a donated dataset of 347 real users; concludes that TikTok "//exploits real users' interests in between 30% and 50% of all recommended videos in the first thousand videos//". Read it for how you measure a recommender with no API at all.   * {[vombatkere2024_tiktok]} (TheWebConf 2024) — **sock-puppet accounts** on a feed you cannot query. Five bot accounts plus a donated dataset of 347 real users; concludes that TikTok "//exploits real users' interests in between 30% and 50% of all recommended videos in the first thousand videos//". Read it for how you measure a recommender with no API at all.
   * {[karnam2026_setting]} (IEEE S&P 2026) — the **data subject's right of access** as the instrument. Sock-puppet accounts on Instagram, TikTok and YouTube, then GDPR Art. 15 requests, then a comparison of the returned data package against the ground truth the authors themselves generated. All three platforms failed to report purpose, recipients and retention period.   * {[karnam2026_setting]} (IEEE S&P 2026) — the **data subject's right of access** as the instrument. Sock-puppet accounts on Instagram, TikTok and YouTube, then GDPR Art. 15 requests, then a comparison of the returned data package against the ground truth the authors themselves generated. All three platforms failed to report purpose, recipients and retention period.
-  * {[gegenhuber2026_there]} (NDSS 2026) — **enumeration at platform scale**, and the ethics that go with it. Probes 63 billion candidate phone numbers and discovers **3,546,479,731** WhatsApp accounts across 245 countries, **57%** of them with a public profile picture. Read it for what a platform-scale study has to say about disclosure and harm before a programme committee will accept it.+  * {[gegenhuber2026_there]} (NDSS 2026) — **enumeration at platform scale**, and the ethics that go with it. Probes 63,170,000,000 candidate phone numbers and discovers **3,546,479,731** WhatsApp accounts across 245 countries, **57%** of them with a public profile picture. Read it for what a platform-scale study has to say about disclosure and harm before a programme committee will accept it.
  
 If you only read one thing about the **rules**, read the Commission's December 2025 decision fining X €120 million: it is the first DSA non-compliance decision, and one of the three grounds is research data access.((European Commission, //Commission fines X €120 million under the Digital Services Act//, press release, 5 December 2025, https://digital-strategy.ec.europa.eu/en/news/commission-fines-x-eu120-million-under-digital-services-act — page fetched 2026-08-27, "Last update 16 January 2026".)) If you only read one thing about the **rules**, read the Commission's December 2025 decision fining X €120 million: it is the first DSA non-compliance decision, and one of the three grounds is research data access.((European Commission, //Commission fines X €120 million under the Digital Services Act//, press release, 5 December 2025, https://digital-strategy.ec.europa.eu/en/news/commission-fines-x-eu120-million-under-digital-services-act — page fetched 2026-08-27, "Last update 16 January 2026".))
Line 282: Line 282:
 ^ Finding ^ Denominator the paper used ^ ^ Finding ^ Denominator the paper used ^
 | Only **7.7%** of undeclared political ads were moderated as political by Meta; **60.4%** of the ads Meta did moderate did not match its own criteria {[bouchaud2024_beyond]} | 29.5 M ads from the Meta Ad Library API, 16 EU countries | | Only **7.7%** of undeclared political ads were moderated as political by Meta; **60.4%** of the ads Meta did moderate did not match its own criteria {[bouchaud2024_beyond]} | 29.5 M ads from the Meta Ad Library API, 16 EU countries |
-| **3,546,479,731** WhatsApp accounts discovered; **57%** with a public profile picture, **66%** of a 500,000-image sample containing a detectable face {[gegenhuber2026_there]} | 63.2 bn candidate phone numbers enumerated |+| **3,546,479,731** WhatsApp accounts discovered; **57%** with a public profile picture, **66%** of a 500,000-image sample containing a detectable face {[gegenhuber2026_there]} | 63,170,000,000 candidate phone numbers enumerated |
 | TikTok "//exploits real users' interests in between 30% and 50% of all recommended videos in the first thousand videos//" {[vombatkere2024_tiktok]} | 347 donating users, 4.9 M videos, plus 5 bot accounts | | TikTok "//exploits real users' interests in between 30% and 50% of all recommended videos in the first thousand videos//" {[vombatkere2024_tiktok]} | 347 donating users, 4.9 M videos, plus 5 bot accounts |
 | **17,842** Amazon products restricted from shipping to at least one world region; **1.1%** of 796,081 sampled books restricted to at least one of four Middle Eastern countries {[knockel2026_banned]} | Common Crawl-derived Amazon product set | | **17,842** Amazon products restricted from shipping to at least one world region; **1.1%** of 796,081 sampled books restricted to at least one of four Middle Eastern countries {[knockel2026_banned]} | Common Crawl-derived Amazon product set |
 | Platform blocking of accounts advertised for sale worked on **19.71%** of them overall, and very unevenly: YouTube 5.02%, Facebook 5.70%, X 18.67%, Instagram 46.41%, TikTok 816 of 1,700 — the paper's own summary is "//TikTok and Instagram demonstrated the highest detection efficacy at 48%, whereas YouTube and Facebook showed the lowest efficacy at just 5%//" {[beluri2025_exploration]} | 11,457 visible accounts from 11 marketplaces | | Platform blocking of accounts advertised for sale worked on **19.71%** of them overall, and very unevenly: YouTube 5.02%, Facebook 5.70%, X 18.67%, Instagram 46.41%, TikTok 816 of 1,700 — the paper's own summary is "//TikTok and Instagram demonstrated the highest detection efficacy at 48%, whereas YouTube and Facebook showed the lowest efficacy at just 5%//" {[beluri2025_exploration]} | 11,457 visible accounts from 11 marketplaces |
-| **136,009** Twitter users' Mastodon accounts identified across 2,879 instances; **96%** of migrants joined the largest quartile of instances {[he2023_flocking]} | 15,886 Mastodon instances, 1.02 M crawled accounts |+| **136,009** Twitter users' Mastodon accounts identified across 2,879 instances, and "//the top 25% most populous instances contain 96% of the users//" {[he2023_flocking]} | 15,886 Mastodon instances, 1.02 M crawled accounts |
 | Over **90%** of targetable Facebook identities in the US had at least one data-broker-provided attribute (Australia 81.3%, UK 74.4%) {[venkatadri2019_auditing]} | the Facebook advertising interface, 7 countries | | Over **90%** of targetable Facebook identities in the US had at least one data-broker-provided attribute (Australia 81.3%, UK 74.4%) {[venkatadri2019_auditing]} | the Facebook advertising interface, 7 countries |
 | On X, posts containing external links had a median visibility score an **order of magnitude** below those without ("//0.0069 vs. 0.084//" for two named accounts) {[galeazzi2026_revealing]} | 17 M + 35 M tweets from two published datasets | | On X, posts containing external links had a median visibility score an **order of magnitude** below those without ("//0.0069 vs. 0.084//" for two named accounts) {[galeazzi2026_revealing]} | 17 M + 35 M tweets from two published datasets |
design/platforms.txt · Last modified: by karel.kubicek.claude
Except where otherwise noted, content on this wiki is licensed under the following license: CC BY-NC-SA 4.0
CC BY-NC-SA 4.0 Donate Powered by PHP Valid HTML5 Valid CSS Driven by DokuWiki