Wednesday, February 10, 2016

Copyright Infringement and Type I/II error



 “It is more important that innocence be protected than it is that guilt be punished, for guilt and crimes are so frequent in this world that they cannot all be punished.”

--John Adams, defending British Soldiers involved in Boston “so-called Massacre”

 Motivation

When one thinks of copyright protection, one thinks of barring certain behaviors that violate copyright, like massively duplicating a movie and offer copies for sale. That view and impression is outdated. Starting with Universal City Studios Inc. v. Sony Corp. of America, copyright holders have increasingly used copyright protection as a vehicle for shaping product design, devices or platform. They exert influences in three ways: litigations (Universal City Studios v. Sony Corp. of America; Cahn v. Sony; Viacom v. YouTube.), regulations (Audio Home Recording Act; Digital Millennium Copyright Act), and agreements formed by private parties (User User Generated Content Principles). This shift in focus has shifted the power of determining and executing copyright disputes from the court to private parties. The result, so far, has been dire. This blog will focus on the agreement that shaped today's User Generated Content services, like YouTube.

User User Generated Content Principles(“UGC services”), agreed to by leading commercial copyright holders and UGC services, set the current framework for UGC services. Within this framework, copyright owners provide reference data (video footage or audio track) to which they “believe” to own copyright. Then UGC services then employ some matching technology to identify content that matches key elements of those reference data. Finally, UGC services will “block such matching content before that content would otherwise be made available”. The rest of the blog will focus on the problems of this framework, exemplified by YouTube’s Content ID system (“the System”).

Contrast this with a court handling a copyright case. The plaintiff has to prove two things: 1) that he has the right to the relevant portion of the work in question and 2) that the defendant has violated one of the “exclusive rights”. After that, 3) only if the defendant fails to come up with an affirmative defense, the court will issue injunction or grant damages. However, on YouTube, the System effectively becomes the “judge, jury, and executioner” of copyright disputes. In the rest of the blog, I will first point out the fallibilities of the System with respect to the three elements and the problems arising from that. Then I will propose a framework for thinking about this issue.1.

Fallibility of the System

Copyright ownership. 

This framework swaps the burden of proof with “good faith”.[1] This low standard of “proof” enabled fraudulent copyright claim. Additionally, even when one has a legitimate copyright claim to some content, he does not necessarily own every single part of the content. For example, when media companies broadcast some news, they will use some material in public domain, that is material with no copyright owner. The media companies have copyright to the overall news broadcast or comments, but not to those public domain footage. Everyone, should be free to use those footage. However, the System is not able to distinguish which part of the reference data is copyrightable, and when it finds other videos matches the portion of public domain material, it automatically makes the infringement judgement. This is exemplified by YouTube taking down NASA’s Mars landing clip and Lon Seidman’s video discussion of the landing.

Existence of infringement.

 The matching technology is, in the first place, imperfect. As any statistical procedure, it inevitably makes “spurious matching”. For example, a YouTube user called eeplox uploaded a video with only sounds of nature like bird calls. The System matches its audio track to a composition licensed by Rumblefish, a music licensing firm (Wired report).

Affirmative defense.

Whereas the System is fallible in the first two dimensions, it is incapable of respecting fair use. Works involving fair use, such as parody and critical review, usually remixes or resemble copyrighted material. However, the System does not understand fair use, and automatically finds infringement. A most ironic example concerns Prof. Larry Lessig, who posted a lecture video on the cultural importance of remixes. The system muted the lecture, because it uses excerpts of music owned by Warner Music (report on EFF).
Given the inadequacies discussed above, the power yielded by the System is huge. It can and often does automatically and block the content. One can dispute the block, but the accuser gets to block the video for 10 business days for free. This is in stark contrast to the way court handles copyright infringement—unless copyright ownership and infringement is proven, and fair use defense fails, the content would not be blocked for a single second.

A Framework

When making judgements, any system make both Type I errors (falsely classify legitimate material as infringement) and Type II errors (fails to catch infringing material). One trades off these two errors by making the system to be either more stringent (more type I errors and fewer type II errors) or less stringent. A rational decision maker should weigh the costs of making these two types error and adjust the stringency of the system accordingly.
What is the cost of type I error? Stifling creativity and censorship. What about type II error? Possibly loss of revenue to copyright holders. However, as is clear from the US constitution, the purpose of copyright protection is to give copyright holders some rewards to incentivize them to do creative work. Copyright protection is a means to an end (stimulate creativity) not an end in itself. Copyright holders are not categorically entitled to all possible rewards. Type II errors are a cost for copyright holders, but not necessarily to the society at large. The real cost of type II error emerges only if the erosion of reward fails to incentivize creative work. It is not obvious creator are primarily motivated by financial incentives (though intermediaries like media corporations certainly do): In fact, some musicians willingly give up their financial rewards: allow their fans to stream or download their works for free (see Zoe Keating's complaint of YouTube). Furthermore, psychology research has pointed out that financial incentive could demotivate creativity. Thus there are few arguments for tipping the balance in favor of over-vigilance, and creating a “copyright scare”. I will come back to this issue in a later blog.

Concluding thoughts

Can one expect private parties to sort out the issue? 
No. 
YouTube always adjusted their practice to serve their business priorities, not public interest. In the past, when it was optimal to attract infringing material to mobilize their service, it chose intentional negligence. Now, its business priorities have shifted. It want musicians to make all of their music available on YouTube first. To do this, YouTube bundled this commitment with the privilege of the System (of near perfect copyright protection) (see Zoe Keating's complaint of YouTube).This privilege, comes at society’s cost (over vigilance), but the profit it generated is reaped by YouTube.

Note:

This change in business priorities might seem strange at first sight, so it deserves a careful explanation. In the beginning, when YouTube started, big media companies were unwilling to to put their contents on YouTube or online in general. For one things, they fear increased piracy risks. Second, they are not willing to let some third-party be cut into the distribution business, interrupting their vertical integration (an anti-trust issue, but no time for it now). Third, YouTube was just a small start up then, and there is no way big media companies will want to sign any contract with YouTube. Given the three reasons, it will be impossible for YouTube to get those contents in any legitimate way. The only solution is to let users upload pirated material so that YouTube can attract users. In fact, one of YouTube founders himself uploaded such material.
Now, media companies take it as a given that many media contents will be distributed online by third parties. The business question becomes how to attract them onto my platform. Due to network effects, a platform with the best and most inclusive content will attract the most viewers, and this in turn make the platform more attractive. Google is a big players, but so is its competitors, like Amazon. As a result, Google has adopted carrot (over vigilance on copyright if you join us and have content ID) and stick (negligence if you do not join us) tactic.


[1] Paragraph 3(g) of Principles for User Generated Content Services.



No comments:

Post a Comment