“Tool Without a Handle”: 21st Century Privacy – A Quantum Puzzle
As I have been analyzing the ways in which the “tool” metaphor has better explanatory power than spatial or landscape metaphors (the Internet as “cyberspace,” e.g.), I’ve been regularly amazed at the extent to which the two metaphors are often used simultaneously. For example, Microsoft’s General Counsel Brad Smith, speaking recently in Brussels: “The Internet is a tool. And the good and bad uses to which this tool can be applied are limited only by people’s imagination… In short, the Internet has become the new terrain on which ideas are shared and deeds are shaped.”
Here, in the same sentence, a leading observer of technology law and policy uses both metaphors to illustrate the nature of the Internet. This has led me to ask whether, and where, networked technologies might exhibit both the properties of tools and the properties of landscapes (terrain, forests, or even gardens) simultaneously. This, in turn, leads me to ask if quantum mechanics might help illuminate a path towards better dialogue about the Internet and data privacy in particular. As Omer Tene reminded me recently, he similarly noted that privacy law and quantum mechanics share certain characteristics.
Are network technologies tools? Or landscapes? Is the Internet a tool you use or a place you go? An article by Bob Frankston argued the Internet is neither a tool nor a place. Perhaps networked technologies exhibit properties of both, depending on the beliefs and perspective of the observer.
Perhaps, as in quantum physics, where the objects of study exhibit properties of both waves and particle during the course of a single event, the answer in part depends on the nature of the observation. It is certainly possible, and perhaps even demonstrable, that the Internet’s properties change depending on the beliefs of the observer.
This statement is more profound than simply saying the Internet looks different from different perspectives. This is not the metaphor of the Internet as an elephant surrounded by blind men, each seeing different aspects. In that metaphor, the elephant’s properties are static and unchanged by the acts of observation – the point of the fable is that the observer’s perspective is limited, both by the failure of one form of sensory input (sight) and reliance on only one other form of sensory input (touch). Eventually, a combination of inputs forms a picture of the elephant as a unified whole. In the quantum world, though, it is not possible to make a visual model of what happens that indicates a unified whole.
Among the reasons, all the observable entities are in motion, and the behavior of atoms cannot be accurately illustrated with a two-dimensional picture. Similarly, in Internet policy and data privacy debates, the objects of discussion are constantly in motion, exhibit multiple properties depending on context and the nature of the data can be influenced by the act of observation.
Thus, it is impossible to make a completely accurate visual model of data flows which includes all the ways in which data is simultaneously used, categorized, and regulated. Networked information technologies have wrought such a meaningful change on the way information flows that the mechanistic models of foundational privacy law are no longer precise tools for data protection.
To illustrate, consider the concept of “personal data.” With few exceptions, the first 200-odd years of Privacy in the Unites States primarily concerned contexts where the identifiable nature of the data in question was not seriously in doubt. “Privacy” issues generally concerned security in private correspondence through the mails, privacy in one’s home, effects, and telephone calls, rights of the press to publish gossip (and photographs) about named individuals, and access to records a government held about a named individual. Whether data was “personal” – linked to a named individual – was rarely at issue in the case of addressed correspondence, private homes, captioned photographs, and easily traceable phone calls.
Networked information technologies, which move both data and meta-data at the speed of light, and where data is rapidly linked and un-linked from names or other obvious personal identifiers, have changed the privacy debate as profoundly as other technical advances have changed the landscape of modern physics. Thus, key questions for modern privacy law include the nature of de-identified data, defining what data is “sensitive” (and to whom), and data ownership in contexts where individuals interact with others and other systems.
Similarly, networked information technologies have changed the extent to which data elements have fixed qualities, independent of their observers. In the “macro” world, any given physical object is unique – this vase is either on my table or in the store; it cannot be in both places simultaneously. Online, however, the identical data object can exist in multiple databases simultaneously. Moreover, the identical data object can be subject to different rights, treatments, and ownership simultaneously.
For example, my name and address is “personal data” in that it can be used to identify me. My name and address, while personal to me, can also be subject to property rights of others – for example, it might appear in a customer list that is proprietary to a service provider (who considers the data competitively sensitive), and it might appear in a service directory in which the compiler claims intellectual property rights. It might well appear in government real estate records which are public data. And that name and address is of course displayed openly on my postal mail, handled by a wide variety of parties. All of these instantiations can co-exist. In some cases the personal data is proprietary to someone other than me. In some cases its protection is subject to regulation, while in others, the data is public and unsecured, by necessity or by rule. The differentiator in all these contexts is the perspective of the observer.
Finally, in the macro world, direct physical engagement is necessary for objects to affect one another. I cannot use the vase on my table (in Seattle) to hold flowers growing in Maryland unless either the vase or the flowers move location. Nor could I crack it over the head of a burglar sneaking into a neighbor’s house without traveling across the street with the vase.
Quantum particles, in contrast, exhibit what physicists call quantum entanglement (a concept Einstein mocked as “spooky action at a distance”). By quantum entanglement, physicists generally mean a state where two particles, either because of how they were generated or how they have engaged, become a system such that action on one particle affects the other, no matter what the other’s location. Personal data has this quality as well.
The fraudulent use of my personal data (name and credit card #) by an overseas party to purchase goods online can affect the qualities of my personal data, and other systems on which that data depends. For example, I not only have to call the bank to dispute that charge, but if the bank disables my card on its server, the data on the servers of my e-commerce provider is impacted (the card # they store is no longer valid). I have to replace the now corrupted credit card number, and in turn, update all my other online accounts where that card is stored. The qualities of data stored on one server are impacted by quality and behavior of data on another server, even if those servers are not themselves directly linked.
To summarize, personal data exhibits quantum properties in the following ways:
1) It can be in more than two places at once, or at least appear to be. The same data element can be both “private” (treated as confidential) and “public” at the same time.
2) The trajectory of data is not always subject to the same mechanical laws of physics that allow for relatively simple predictions of motion;
3) Data are “entangled”: one data element can be influenced by another unrelated and seemingly disconnected data element, even at a distance;
To illustrate further, take a more subjective fact such as “Chuck likes recordings of classical guitar music.” That data element also exists “online” in multiple places. Multiple online services (e.g., Amazon, Pandora, and Facebook), using multiple servers, will have some inkling of my musical preferences. But their data will not be identical; the nature of the data depends on the perspective of the observer.
Where I am the observer, my preferences in Amazon are shopping preferences, while my preferences in Pandora are my listening preferences (as Pandora does not offer music directly for sale). That said, for Pandora, my listening preferences are inferable as shopping preferences and it may well link those preferences with a unique identifier that, in turn, is used to build a profile used by advertisers, ad publishers, and other online platforms on the assumption that my listening preferences are my shopping preferences.
A potential disconnect between the perspective of Pandora and myself in terms of the qualities of the same data lies at the heart of modern privacy discussions regarding fairness in data collection and use. Written privacy policies seek to bridge this disconnection by making it clear to online users what observations a data processor holds – i.e., Pandora may disclose it will use my listening preferences in this way. The ability of users to understand the nature and implications of such uses, and the extent to which we should encumber innovators with ensuring users share the same view on data use, are key factors in contemporary discussions of privacy law. At the heart of this is the fact that the same data has different qualities depending on the beliefs of the observer.
A final point: much has already been written (and written well) about the importance of context and contextual integrity in establishing the privacy-relevant values of data. It is difficult to summarize those insights in a few words, but an attempt would be: privacy values vary in ways relative to context, we should use integrity to context as guidance for the appropriateness of data use, and contexts are more granular than simply a public/private dichotomy. My points are additional: 1) a reason the public/private dichotomy fails is because data can be both simultaneously public and private; 2) the way in which context is created and which it changes the rules applicable to personal data is via the perspective of the observer; 3) the nature of data in one context can be changed by alterations to data in a completely different context.
In sum, policy discussions about privacy should reflect the quantum nature of personal data and networked information technologies. Just as the familiar mechanical principles that lead to predictability in the macro world inaccurately describe the quantum world, so too do metaphors that apply mechanical principles (i.e., data as particles) insufficiently describe the modern information economy. What to do about this will be the subject of a following blog.
 “Safety, Privacy, and the Internet Paradox: 2015 and the Need for New Trans-Atlantic Rules,” Brad Smith,
Executive Vice President and General Counsel, Microsoft Corporation Prepared Remarks at the Centre for European Policy Studies Brussels, Belgium January 20, 2015 http://mseu.blob.core.windows.net/eumedia/2015/01/Brad-Smith-Brussels-January-20-2015.pdf
Pierre de Vries, “Gardening the Internet/Web: A Framework for Communications Policy,” Draft 31 January 2008, online at: https://sites.google.com/site/jpdevries/earlierwork
“The Internet is neither a thing nor place. It doesn't carry packets like a railroad carries freight. If anything, it's more like the way we use paths, streets and sidewalks to get from point A to B.” http://www.circleid.com/posts/20150219_internet_native_policies/
Anyone who has taken their kids to the zoo must wonder why this story never referenced what anyone thought of the elephant’s smell…
For a good review of this history, see “A Brief History of Information Privacy Law, Daniel J. Solove, PROSKAUER ON PRIVACY, PLI, 2006, GWU Law School Public Law Research Paper No. 215.”
These issues are not entirely absent from current discussions. A recent blog post discusses the concept of “autonomy privacy”: “an individual’s ability to conduct activities without concern of or actual observation.” http://blogs.berkeley.edu/2015/02/27/privacy-vs-privacy/ But in the digital world, there is no such thing as “conducting activities without actual observation.” Some system must observe the data. A better way to frame this might be “without concern for unfair private or government actions with regard to data observed.”
See list of examples of PII at http://csrc.nist.gov/publications/nistpubs/800-122/sp800-122.pdf ; see also EU Data Protection Directive, 95/46/EC, (“personal data” shall mean any information relating to an identified or identifiable natural person”) http://eur-lex.europa.eu/legal-content/en/ALL/?uri=CELEX:31995L0046
Compare Feist Publications, Inc., v. Rural Telephone Service Co., 499 U.S. 340 (1991) (no copyright exists for mere compilations of data with no additional creative element, such as plain phone directory listings) with Key Publications, Inc. v. Chinatown Today Publishing Enterprises Inc., 945 F.2d 509 (2nd Circuit, 1991)(Chinese-language yellow pages contained sufficient element of creativity to be copyrightable).
Study of quantum mechanics has also lead directly to inquiry about information processing: quantum computing. See http://www.quantiki.org/wiki/What_is_Quantum_Computation%3F for a basic tutorial on this field of study.