August 4, 2009
Accounting for consumers' isolation and legislators' paid for apathy, I advocated last week to set up a "philosopher's corner" to tackle the issues privacy raises in our Information Age. Who can claim to be an expert in all the fields involved? Yet how can one ignore any one influence which either links or opposes privacy, sharing and money. No lasting solution will emerge unless all reasonable voices come together in a permanent forum.
The list of authors mentioned in these fillips could be a starting point as my formula is to react to the unfolding pageant of events and opinions with a very catholic taste, limited only by resource availability. Today let us turn to Randal C. Picker for his recent paper on online advertising (*).
Proof Randal Picker's voice is reasonable can easily rest on three notes, harmonizing economic imperatives, the need for regulations and the interplay of identity and privacy. Calling reasonable someone whose positions I share may not be the most objective criterion, I admit, but it allows me to quickly focus on what I find the salient ideas of this research.
Randal Picker's main idea is to model advertising as a search for matches between sellers and buyers. "This is as friendly a conception of advertising as possible", he recognizes, but a very legitimate assumption which accounts for a significant part of this activity. His second point is that identity is a complex notion, which goes from the "Google Opera version of me" to the "actual me".
Isn't it fair to mirror my earlier argument and assert that, by seconding these opinions of mine, Randal Picker enhances my own claims to being eminently reasonable? It also sets the stage for looking at where we part from each another.
First Randal Picker is too indulgent regarding the FTC's refusal to limit usage of the behavioral data a company collects directly (1). "Back in the dawn of sales, my local merchant knew me and knew what I had purchased. [...] Amazon does that today, except with remarkable efficiency." I would not count on gentlemanly understatements to convey to the FTC commissioners the message that, when technology allows a "remarkable" jump in efficiency, it is urgent to review the rules. Back in the dawn of the music industry, I could also buy a record and share it with my friends... The HADOPI saga (2) shows how ruthlessly the French government wants to outlaw modern ways of sharing for being too remarkably efficient.
This venial slip is compounded by a serious sin of omission. Having explained how each of our online relation creates a different identity, Randal Picker does note that by favoring "first-party data" over "third-party data", the FTC creates an artificial incentive towards vertical integration. But he fails to account for the reach of data aggregation.
Verifying my ID, the "actual me", for example by linking together several of my ids, each a "partial me", is a pattern recognition task. To minimize false positives, which let a thief escape, policemen tend to increase false negatives and thus challenge the innocent. Tempers can rise in the process. Personal profile aggregators however run no such risk when linking together their "first-party data" with multiple sources of "third-party data". Since the most they and their advertising clients deign to reveal is a mere intent to collect and target, all identification errors remain hidden from consumers.
Thus free from scrutiny, aggregators can maximize their gains by any process which feeds personal profiles with a good enough success rate. This is precisely the case with so called non identifiable information, whose trading remains so far free from any regulation. The AOL scandal proved merging enough anonymous data can make identity emerge. Now, as reported by the Technology Review (**), Arvind Narayanan and Vitaly Shmatikov have given a sound scientific basis to the enrichment of identified profiles with anonymized data (***). Focused on social network data rather than search histories, the authors' process is subject to statistical noise but yields the type of good enough results aggregators look for.
Randal Picker is quite right to assign an advantage to companies whose "first-party data" already contain personally identified information. It is one reason Google acquired DoubleClick. It also explains the growth into online aggregation of traditional consumer data companies such as Acxiom (3), documented by Stephanie Clifford (****). What Randall Picker missed is that Facebook itself is at a disadvantage against an Acxiom which can boast "[its] cookie data is all anonymous" as, contrary to Facebook, its more sensitive information is kept away from the public eye.
Randal Picker is right again to question data disclosure relevance. Acxiom does not entrust its consumer data to online advertisers, contrary to what happens when it rents a list of postal addresses. But then isn't opt-in versus opt-out a bit byzantine when the ad network already owns your data legally? Law professor Paul M. Schwartz speaks of consumers' "unwitting participation", which increases rather than decreases privacy issues.
For instance read David Gelles' report on Cheryl Smith (*****). The reuse of Mrs Smith's photograph for advertising purposes is another clumsy attempt by Facebook to justify its valuation. The scary part though is that Mr Smith will never know why he was presented with "an ad for "hot singles" paired with a picture". Was he targeted? If so, why? Without "real transparency" (4), it does cast Mr Smith in a "false light" (5), doesn't it?
Randal Picker's thesis however is economic in nature. It offers quantitative support on how overall welfare would be raised were both consumers and suppliers to advertise in order to be matched. Consumers of course do not spend money, they instead "reveal information to advertise". Needed then is a market to let consumers control their "spending" by picking among different levels of personal disclosure from several ad space providers.
His thesis though suffers from three flaws. First it explicitly ignores that suppliers compete to advertise. When ad auctions are the very basis for the success of Google, the dominant ad space provider, it is as if one studied global warming on an oxygen free atmosphere. Nice but who cares?
Second it abets bundling service transactions with data transactions. True competition may in theory provide enough choice to enable consumers to virtually unbundle a service from its coerced data exchange. But again, this wish flies in the face of reality. Ad networks are driven by economies of scale. Contextual search requires no data exchange in principle and "[Google] does not currently collect or serve ads based on [...] personal information without user permission". Were Google to degrade this "basic" service, how could consumers be certain to retain their former option?
Last but not least, trading off privacy against better matches is unnecessary as shown by the ePrio technology. Its adoption of course amounts to a reform as problematic as would be showing the door to your mother-in-law after having invited her into your home. Expect some reluctance on her part to leave. Isn't she acting for your own sake? Isn't Randal Picker's solution akin to relying on the marriage market for a wider choice of in-laws?
Hard to believe your mother-in-law's advertising increases global economic welfare. We rather need a new law to oust all those nosy in-laws.
- (*) ............. Online Advertising, Identity and Privacy, by Randal C. Picker (The Law School - The University of Chicago) - June, 2009
- (**) ........... Social Security. Researchers raise privacy concerns about social-network data (Technology Review) - July-August, 2009
- (***) ......... De-anonymizing Social Networks, by Arvind Narayanan and Vitaly Shmatikov (IEEE Symposium on Security and Privacy) - May, 2009
- (****) ....... Ads Follow Web Users, and Get Deeply Personal, by Stephanie Clifford (New York Times) - July 31, 2009
- (*****) ..... Privacy rumours rage on Facebook, by David Gelles (Financial Times) - July 29, 2009
- (1) the documents published by the FTC are listed as recommended readings on the overview page of my FTC campaign
- (2) check the relevant texts on the French HADOPI initiative in my lecture on copying
- (3) see Acxiom in the wikipedia for more information about its traditional business
- (4) a wish from Marc Rotenberg, executive director of the Electronic Privacy Information Center, quoted by Stephanie Clifford
- (5) per Simson Garfinkel's expression for one type of privacy violation, in Privacy Requires Security, Not Abstinence (Technology Review) - July-August, 2009