To glimpse the future of information privacy law, you should look at the work coming out of two Stanford Law School centers, the Center for Internet and Society and the CodeX center. In the past few years, these centers have housed a steady stream of fellows and clinical professors who have written some of the most interesting, vibrant, and future-looking scholarship in this field. For example, Lauren Gelman’s article on “blurry-edged” boundaries—already lauded in these pages—is a significant contribution, one that has advanced our understanding of the complicated relationship between social networks and privacy. Another excellent example is Structural Rights in Privacy, written by Harry Surden—now my colleague at the University of Colorado—during his stint as a fellow at CodeX, about how technology sometimes protects privacy in ways we fail to appreciate until the technology changes. I write now to focus on another scholar in the Stanford centers, Ryan Calo, who has embarked on a fascinating project with an excellent article, People Can Be So Fake: A New Dimension to Privacy and Technology Scholarship, forthcoming in the Penn State Law Review.
Calo focuses on “technologies designed to emulate people,” such as robots with expressive eyes or software assistants designed to look like people. We’ve come a long way since Microsoft’s Clippy the paperclip first annoyingly noticed that it looked like I was writing a letter. Computer scientists, roboticists, and companies have poured time and creative energy into designing interfaces and devices that look and act human, and they’ve made great strides in the process. To document these advances, Calo cites with care a rich, emerging, technical literature, importing dozens of studies and papers into law, saving the rest of us a lot of heavy research lifting.
Calo proves to be much more than merely an import-export specialist, however, because he skillfully examines what the rise of anthromorphic robots means for privacy. By building upon a second body of literature—studies from psychology and communications surveying how we behave when we feel as if we’re being watched by synthetic people—he sees both peril and promise in the rise of the robots and, interestingly, finds the source of both in the same psychological observation: we become inhibited in the presence of a human face, even one we know to be artificial. As we invite robots into our homes and program faces into our software, we should be mindful of this psychological response, because this kind of inhibition deprives us, as Calo cites, of Alan Westin’s “moments ‘off stage’,” leading us down the path to Ruth Gavison’s “terrible flatness” and Julie Cohen’s “bland and the mainstream.”
But Calo turns the lemons of inhibition into lemonade, by suggesting that we should sometimes intentionally trigger inhibition, for example when asking for consent to invade someone’s privacy. Perhaps rather than displaying only a traditional, text-laden privacy policy, Calo argues, websites should also include a picture of a pair of eyes above the text, or perhaps, I would add, lawmakers should force them to do so. This is a fresh spin on old, tired debates about notice and consent: tell me how you’re using my data, and I won’t read and I won’t care; put a human face on the page, and I’ll sense “a visceral reminder that the data being collected will be used and shared.”
This is rich, valuable scholarship, but what elevates his article even further is how Calo uses these observations about anthromorphic privacy triggers to critique the broader Information Privacy Law project. Calo argues, echoing Julie Cohen, Arthur Miller, and Dan Solove, that too many other privacy scholars focus only on data collection and use. He wants to broaden our viewscreen because sometimes our privacy can be invaded even when none of our data is released at all. Many of us have had this vague thought—in the practical trenches of public policy, privacy is almost always seen only through the lens of collection and use—but Calo replaces vagueness with substance, giving us a concrete factual context with which to play out the pros and cons.
This is an excellent article by an exciting junior scholar and, I gather, only the first in a planned series of articles. I eagerly await the next installment.
“Perhaps rather than displaying only a traditional, text-laden privacy policy, Calo argues, websites should also include a picture of a pair of eyes above the text, or perhaps, I would add, lawmakers should force them to do so.”
The appearance of eyes (why not charred lungs?) over text is visual noise we’d desensitize to almost immediately, and if anything a step backward from more humane interfaces and plainly stated opt-in conditions.
Aza Raskin’s idea of bolt-on privacy icons makes loads more sense.
Jason,
Thanks for your note, which Paul just brought to my attention.
What evidence there is suggests that the effects will not wear off. In one study, the effect of eyes on participants was about the same at 1 week as at 9. Reeves and Nass have shown that computers as social actors (CASA) effects are just as pronounced on people very familiar with the technology. It’s ultimately an empirical question and more study is certainly welcome.
I’m not sure I know what a “humane interface” might be. Greg Conti talks about “malicious interfaces”; perhaps this is the flip-side? My own sense is that many people reflexively “click through” opt-in conditions. And recent work out of Carnegie Mellon suggests that some privacy options risk giving users a false and potentially harmful sense of control. But anyway: there’s no reason we couldn’t combine visceral notice with a user option (e.g., the user clicks the option to encrypt and the eyes disappear).
I agree Aza’s notion of privacy icons holds promise. (You may have noticed that Aza lists me as an adviser on the project.)
Thanks again. Best,
Ryan
Ryan,
I regret not at least skimming “People Can Be So Fake” before going after one of its conclusions. Reminder to self: RTFA. The literature in there suggesting “our brains ‘rarely make distinctions between speaking to a machine and speaking to a person’ at a viseral level” indeed gives me pause, and reveals to me I’ve wrongly extended a mental model from limited past observation to the future promises and dangers of anthropomorphic design.
Here’s the thing, though: most of today’s anthropomorphic designs are as well-intended as the pair of eyes over a privacy policy, and each turns quickly into a nag that hinders our awareness of choices and their meaning. Think of touchscreen mall directories. A 3D hostess talks you through your search, offers hints along the way, compliments your choices, and gestures at the mall map like an air hostess doing a seatbelt demonstration.
When I encounter one of these directories, it gives me visceral pause. I feel intruded upon. But this presence does not have the effect of bracing me for the intrusive, noisy, public, social nature of retail, nor does it rekindle my awareness of other presences in the mall besides those I can see. Maybe my sense of recoil could be taken as a subliminal placeholder for broader misgivings, but that’s a stretch: what I am recoiling at is the machine. A machine that interrupts to make sure I understand what a map is instead of serving me a (better) map.
That lost distinction is what prompted snark in my first reply. It’s not that ordinary people don’t know what a contract is. It’s that contracts are full of boring, evasive, meaningless words. Visceral human presences in the mediation of unclear contracts do not touch the problem of meaningless consent.
I’m curious, Ryan, what you think of Google’s unofficial position on opt-in dystopias. This passage speaks my worry: “Once consumers are desensitised to opt-in requests and the sequence of interactions required to constitute opting-in, the actual scope can start growing without much awareness on the part of the user.” Whether or not a pair of eyes brings out a lasting inhibitory response is trivial if all they do is invade our thinking space and raise the volume on a “someone’s watching” sensation that may go dull in the presence of machines and animals for adaptive purposes we aren’t even cognizant of.
And I guess by “humane” design I really mean transparent and unobtrusive. Much in the “Magic Ink” vein. Perhaps I’ll extend humane to allow cognitive interruptions that presage hidden risks, if the doses are measured right, like the ones the body might supply when we step through a door. But my design biases, which you anticipate and deal with in the paper, almost prevented me from learning something new. So thanks for the reply.