Difference between revisions of "Seng (2015b)"

From Copyright EVIDENCE
m (Saved using "Save and continue" button in form)
 
Line 23: Line 23:
 
|Data Type=Secondary data
 
|Data Type=Secondary data
 
|Data Source=Google Transparent Report;Chilling Effects (Lumen)
 
|Data Source=Google Transparent Report;Chilling Effects (Lumen)
|Method of Collection=Quantitative Collection Methods, Quantitative data/text mining, Quantitative Collection Methods
+
|Method of Collection=Quantitative Collection Methods, Quantitative data/text mining
|Method of Analysis=Quantitative Analysis Methods, Regression Analysis, Quantitative Analysis Methods
+
|Method of Analysis=Quantitative Analysis Methods, Regression Analysis
 
|Industry=Publishing of books, periodicals and other publishing; Film and motion pictures; Software publishing (including video games); Film and motion pictures; Sound recording and music publishing
 
|Industry=Publishing of books, periodicals and other publishing; Film and motion pictures; Software publishing (including video games); Film and motion pictures; Sound recording and music publishing
 
|Country=United States
 
|Country=United States
Line 33: Line 33:
 
}}
 
}}
 
|Dataset={{Dataset
 
|Dataset={{Dataset
|Sample Size=21,546,901
 
|Level of Aggregation=takedown requests
 
|Data Material Year=2011-2012
 
}}{{Dataset
 
 
|Sample Size=21,546,901
 
|Sample Size=21,546,901
 
|Level of Aggregation=takedown requests
 
|Level of Aggregation=takedown requests

Latest revision as of 15:14, 1 October 2021

Advertising Architectural Publishing of books, periodicals and other publishing Programming and broadcasting Computer programming Computer consultancy Creative, arts and entertainment Cultural education

Film and motion pictures Sound recording and music publishing Photographic activities PR and communication Software publishing (including video games) Specialised design Television programmes Translation and interpretation

1. Relationship between protection (subject matter/term/scope) and supply/economic development/growth/welfare 2. Relationship between creative process and protection - what motivates creators (e.g. attribution; control; remuneration; time allocation)? 3. Harmony of interest assumption between authors and publishers (creators and producers/investors) 4. Effects of protection on industry structure (e.g. oligopolies; competition; economics of superstars; business models; technology adoption) 5. Understanding consumption/use (e.g. determinants of unlawful behaviour; user-generated content; social media)

A. Nature and Scope of exclusive rights (hyperlinking/browsing; reproduction right) B. Exceptions (distinguish innovation and public policy purposes; open-ended/closed list; commercial/non-commercial distinction) C. Mass digitisation/orphan works (non-use; extended collective licensing) D. Licensing and Business models (collecting societies; meta data; exchanges/hubs; windowing; crossborder availability) E. Fair remuneration (levies; copyright contracts) F. Enforcement (quantifying infringement; criminal sanctions; intermediary liability; graduated response; litigation and court data; commercial/non-commercial distinction; education and awareness)

Source Details

Seng (2015b)
Title: "Who Watches the Watchmen": An Empirical Analysis of the Reasons for Rejecting Copyright Takedown Notices
Author(s): Seng, D.
Year: 2015
Citation: Seng, D.K.B (2015). Who Watches the Watchmen": An Empirical Analysis of the Reasons for Rejecting Copyright Takedown Notices. Available at SSRN: https://ssrn.com/abstract=3687861 or http://dx.doi.org/10.2139/ssrn.3687861
Link(s): Definitive , Open Access
Key Related Studies:
Discipline:
Linked by:
About the Data
Data Description: Original dataset comprises about 60 million (N0 = 59,033,579) takedown requests submitted between 2011 and 2012, collected from the Chilling Effects repository (Lumen) and Google Transparent Report.

A classification-based model is developed to classify each request according to their attributes, characteristics or features. Based on these attributes - explanatory or independent variables - the model has the objective of evaluating the influence of each of these attributes to the acceptance or rejection of a takedown request.
For the purposes of model building, in the part involving the determination of the URIaction field, the dataset is reduced to around 21,547,250 (N1) takedown requests. More 359 requests are removed from the data set for missing at least one of the three fields/variables (hasGoodFaithBeliefNotAuthorized/ hasAcknowledgedAccuracy/ hasAuthorityToAct) in the originating notice.
The final dataset comprises 21,546,901 (N) requests submitted in 2012, including 20,301,257 successfully processed takedown requests and 1,245,644 unsuccessfully processed takedown requests, with 1 dependent variable (URIaction) for 19 independent variables.
Logistic regression is used for the classification of the takedown requests. For the purposes of model building, the open source statistical package R and its programming langued are adopted in this research.

Data Type: 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:
  • 2011-2012
Funder(s):

Abstract

“In “Copyrighting Copywrongs”, the scope and extent of mistakes made in compliance with DMCA formalities when reporters issued take-down notices was examined. In this paper, there is a further examination of the actions taken by the Internet intermediaries in response to these take-downs. The Internet intermediary selected for this analysis is Google, Inc., because it is the largest intermediary and accounts for the largest number of take-down notices received. But because the actual mechanisms for Google’s processing of take-down notices is not publicly accessible, an accurate statistical model based on 21.8 million take-down requests (a notice can have multiple requests) received between 2011 and 2014 is instead built to approximate Google’s automated processing mechanisms. One would have expected Google’s mechanisms to adhere to the DMCA formalities rules strictly. The analysis confirms some correct implementations, such as the fact that submitting a complaint which lacks a description of the title of the allegedly infringed copyrighted work significantly increases the odds that the request will be rejected. However, it also yields some counterfactual, such as the fact that submitting a take-down complaint which lacks any description (including the title) of the copyrighted work improves by up to 40 times the rate of a successful take-down request. Also, Google’s processes do not appear to check if the URL in the take-down request is valid. The analysis also shows that notwithstanding Google’s implementation of the Trusted Copyright Removal Program (TCRP) to improve the quality of take-down complaints sent by reporters, some TCRP senders have a very poor track record of sending successful take-down requests. This is the first empirical confirmation of the lack of oversight under the DMCA over Internet intermediaries and their response to take-down notices. To address this very serious issue that would undermine the role of intermediaries as the watchmen of the Internet, this paper advances proposals that would disrupt the strategic behavior exhibited by the intermediaries to re-balance the positions taken by intermediaries and right-holders.”

Main Results of the Study

The study suggests that:
1. Even though Google system for processing takedown requests is heading in the right direction, it was not designed to completely comply with the DMCA formalities. For example, the research findings show that the lack of any description about the copyright work in the takedown request decreases significantly the probability of this request being rejected. In practice, except for the title of the work, it can be better for the complaining parties to not include a description of the infringed work in their requests, even though the DMCA do require that takedown notices provide this information.
2. A significant number of Trusted Copyright Removal Program (TCRP) members has a better chance to have success in their takedown request than non-TCRP members. However, it is important to note that some TCRP members have a poor rate of successful takedown requests.
3. It seems that Google system does not check whether the URL in the takedown request is valid for its approval or rejection.
4. Copyright infringement claims involving computer programs and video games are more likely to be rejected when compared to other works, such as literary works, visual arts, performing arts and motion pictures.
5. Overall, there is a lack of oversight under the DMCA over Internet service providers and their response to takedown notices.

Policy Implications as Stated By Author

From a policy perspective, the author suggests that legally requiring that Internet service providers provide a justification for rejecting a takedown request to the reporter may be an option to ensure more accountability and transparency among these service providers. One possibility is to extend the requirement already provided for in Section 512(c)(3)(B)(ii) of the DMCA including this obligation. The author also proposes that the service provider should not be liable for any claim based on its notification of the reason for rejecting a takedown request in good faith.



Coverage of Study

Coverage of Fundamental Issues
Issue Included within Study
Relationship between protection (subject matter/term/scope) and supply/economic development/growth/welfare
Relationship between creative process and protection - what motivates creators (e.g. attribution; control; remuneration; time allocation)?
Harmony of interest assumption between authors and publishers (creators and producers/investors)
Effects of protection on industry structure (e.g. oligopolies; competition; economics of superstars; business models; technology adoption)
Understanding consumption/use (e.g. determinants of unlawful behaviour; user-generated content; social media)
Green-tick.png
Coverage of Evidence Based Policies
Issue Included within Study
Nature and Scope of exclusive rights (hyperlinking/browsing; reproduction right)
Exceptions (distinguish innovation and public policy purposes; open-ended/closed list; commercial/non-commercial distinction)
Mass digitisation/orphan works (non-use; extended collective licensing)
Licensing and Business models (collecting societies; meta data; exchanges/hubs; windowing; crossborder availability)
Fair remuneration (levies; copyright contracts)
Enforcement (quantifying infringement; criminal sanctions; intermediary liability; graduated response; litigation and court data; commercial/non-commercial distinction; education and awareness)
Green-tick.png

Datasets

Sample size: 21,546,901
Level of aggregation: takedown requests
Period of material under study: 2011-2012