Urban, Schofield and Karaganis (2017b)
Contents
Source Details
| Urban, Schofield and Karaganis (2017b) | |
| Title: | Takedown in two worlds: an empirical analysis |
| Author(s): | Urban, J.M., Schofield, B.L., Karaganis, J. |
| Year: | 2017 |
| Citation: | Urban, J.M., Schofield, B.L. and Karaganis, J. (2017) Takedown in two worlds: an empirical analysis. 64 J. Copyright Soc’y 483 |
| Link(s): | Open Access |
| Key Related Studies: | |
| Discipline: | |
| Linked by: | |
| About the Data | |
| Data Description: | Data was obtained from the Lumen database, comprising 288,675 takedown notices gathered over a six-month period. This is largely to substantiate and corroborate, quantitatively the findings in Study 1 (Urban, Karaganis and Schofield (2017a)), which provides qualitative data on the subject of notice and takedown.
• Study 2 comprises a randomised sample of all of these notices, totalling 1,827 takedown requests. • Study 3 comprises a focussed study on a further randomised sample of 1,732 takedown requests sent exclusively to Google’s Image Search service (a service which is disproportionately represented in the larger dataset). |
| Data Type: | Primary and Secondary data |
| Secondary Data Sources: | |
| Data Collection Methods: | |
| Data Analysis Methods: | |
| Industry(ies): | |
| Country(ies): | |
| Cross Country Study?: | No |
| Comparative Study?: | No |
| Literature review?: | No |
| Government or policy study?: | No |
| Time Period(s) of Collection: | |
| Funder(s): |
|
Abstract
“In its nearly two decades of existence, the Digital Millennium Copyright Act’s “notice and takedown” process has become both a source of controversy and a primary means of addressing online copyright disputes. Yet there is little empirical research into the use of notice and takedown or its effectiveness. Because it depends on the private communications and actions of copyright holders, online service providers, and targets, notice and takedown operates largely within a difficult-to-study “black box.” This article contributes to the literature with an empirical look at two sets of takedown notices. It digests findings from two quantitative studies of takedown notices in the Lumen database, reported fully in the three-study report, Notice and Takedown in Everyday Practice (available at https://ssrn.com/abstract=2755628), and peer-reviewed here.
The first study, of notices sent to Google Web Search, shows a heavy reliance by large rightsholders on automated infringement detection and notification, and tends to support their assertions that they focus their enforcement resources on large-scale infringement. At the same time, 31% of these notices raised questions related to their accuracy and statutory compliance. The second study provides a snapshot of takedown as used by one set of smaller notice senders — individuals and small businesses sending to Google Image Search. In this study, notices were more likely to target social media sites, blogs, and personal websites, and 70% of the notices raised questions. While both studies revealed surprisingly high percentages of notices of questionable validity, the questions raised tended to differ. The automated notices in the first study predominantly exhibited issues identifying the works in question, while the notices sent by smaller rightsholders in the second study predominantly exhibited issues with the underlying claim. Accordingly, while notice sending appears to need improvement, there is no “one-size-fits-all” approach. Both “worlds” of notice sending need to be accommodated. We therefore suggest a set of voluntary best practices, educational efforts, and modest legal reforms directed to the different issues we observed.“
Main Results of the Study
The findings are split across two studies (further corroborating evidence detailed in Study 1 - Urban, Karaganis and Schofield (2017a)), each presenting a different “world” of notice and takedown in practice:
Study 2 - Focussing on notices sent to Google Web Search, this study finds a heavy reliance by large rightholders with major entertainment industry copyrights. Notices are usually sent in order to tackle large-scale, and obvious, infringements (with two thirds of requests being directed towards torrent or file-search sites), limiting the potential negative effects of erroneous notices. Nonetheless, almost a third of the notices in this study have issues with validity, including fundamental mismatches between the identified work and targeted work, and problems identifying fair uses.
Study 3 - Focussing on notices sent to Google Image Search, this study finds that these notices are mainly issued by individuals and small businesses, and directed towards personal websites, including social media accounts and blogs. As with Study 2, the notices are dominated by a limited number of senders, with one individual accounting for over half of the notices analysed. Approximately 70% of the notices in this category had substantive issues with validity, including claims over uncopyrightable subject-matters, fair uses, and non-copyright related disputes (including privacy and defamation). The authors attribute the higher instance of notice errors both due to the nature of individuals and smaller businesses to be less-professionalised, and the fact that over 50% of senders were based outside of the US (suggesting a lack of familiarity with US copyright laws).
Policy Implications as Stated By Author
The authors recommend voluntary measures by righthholders and OSPs to improve the notice and takedown system and challenge problematic notices. This is largely to be achieved by promoting greater awareness and identification of non-infringing uses, and includes:
• Increased human review, both for initial filtering and idenfitication of “rogue” sites, and for ongoing spot-checking of algorithmic decision making.
• More advanced technological methods, including the development of improved identification algorithms (that detects more than just e.g. URL or file-name), and a system which automatically rejects flawed notices (e.g. that are directed at a defamation disputes, or have incomplete or missing information).
The authors also recommend copyright education for rightholders and a limited degree of statutory reform (particularly in regards e.g. standards of misrepresentation and damages).
