In 1973, the U.S. Department of Health, Education and Welfare published a report in which they outlined a Code of Fair Information Practices. Over the intervening years, these practices have been adopted, modified, and updated and form the basis of many modern privacy laws, both in the United States and in other countries. One incarnation, the U.S. Federal Trade Commissions Fair Information Practice Principles consist of five principles:
- Notice/Awareness
- Choice/Consent
- Access/Participation
- Integrity/Security
- Enforcement/Redress
Notice is considered the foundation upon which the other principles rest. Without notice as to what information is collected and how it will be used, it is hard for consumers (data subjects) to make meaningful use of the other principles. For many digital services, the purpose of a privacy statement, often labeled a privacy policy, is to supply this notification.
Unfortunately, for consumers the costs in their time to read all the privacy statements they encounter make them mostly a ostentatious exercise more for regulators benefit than consumers. ( See The Cost of Reading Privacy Policies) A more modern trend, in line with the principles of Privacy By Design, is to supply this notice inline with the collection of information. While this reduces the transaction costs associated with a consumers being put on notice as to what information is collected and for what purpose, it helps only slightly in the context of Big Data.
Big Data Benefits
As readers of this blog will no doubt be aware, one of the biggest promises of big data is finding hidden correlations amongst reams of seemingly unrelated data. Google Flu Trends is an excellent example. Google has found that the search of specific terms is fairly reliably correlated with influenza surveillance data. Such information could be used by health officials in various countries to mobilize their resources more effectively, sending medical professionals into communities with a rise in flu related searches or spending more on media campaigns emphasizing hygiene and washing hands.
This is clearly a benefit that was not anticipated by Google when they created their search engine. Had they narrowly tailored their privacy policy to exclude such revolutionary data mining, only allowing the search terms entered by their users for returning of search results and then, following good data security practices, destroying the information, such correlations could never have been discovered and the potential social benefits of reducing the spread of influence would not exist.
Proper Notice
What sort of notice should Google then provide? A generic statement that they may use your search data for other purposes seems over broad. A narrow one that they can use your search terms to help prevent influenza seems equally too narrow and proscribes other, potentially, beneficial uses. In fact, in the realm of Big Data, the concept of notification of future use seems rather archaic, since data collectors cant possibly know all the possible uses of the data they collect and therefore the data subjects are even more so in the dark. The miracle of Big Data is finding heretofore unknown correlations; knowing what youre going to look for is antithetic to the true benefit of Big Data analytics.
While notification of ultimate use is of little help in the context of Big Data, consumers still should be put on notice that their data will be used for purposes other than the obvious one for which it was collected, or such should be readily apparent from the context of the collection process. This clearly precludes the use of surveillance, whether covert or overt, as a means of collection.
The Limits of Notice
Even with truthful and full disclosure, it is difficult for consumers to understand the risks their disclosure of information presents. Consumers have cognitive biases that discount future risks for current benefits; we behave irrationally with regards to information about us. The history of privacy rests on informed consent, but current notions about the reasoning ability of humans seems to suggest that informed consent in privacy decision making may not be possible.
We protect children because they cant fully appreciate the ramifications of their actions, but adults also seem prone to the mistakes, even if they appear intelligent and introspective. In other words, no matter how much information is provided in a notice about the collection and use of data, very few people will be able to fully appreciate and rationalize the tradeoffs.
Limitations on Use
While giving notice is important, is does not give a Big Data firm carte blanche to use the information in whatever fashion they desire. Notice has limited capabilities in the Big Data context and humans have limited capabilities to making wise choices even when presented with all available information. It therefore falls on the Big Data firm to make those choices.
The use should ultimately be beneficial to the data subject, if not directly then indirectly through social benefits like Googles Flu Trends. Customer expectations do not match the use to which most companies put data and if they want to develop trust relationships with their customers, that much change. (See What Misconceptions Do Consumers Have About Data Privacy? )
Firms should absolutely not perform analytics that would harm the data subjects interests and they should not aggregate data in a way that would be unexpected to the data subject. The latter is something that Big Data is prone to do by compiling and cross referencing multiple data points.
The Software & Information Industry Association recently came up with a guide for policy makers which suggests that policy makers should allow firms to weigh benefits and risks of privacy harms rather than be formulaic in proscribing collection and use principles. Data stewardship is more than just proper security controls but is about making informed and ethical decisions for consumers.
