Science - USA (2022-01-21)

SCIENCE science.org

CREDIT: (PHOTO ILLUSTRATION WILLIAM DUKE, (PHOTO RONALD HUDSON/BIGSTOCK.COM

may be muted or entirely unavailable on-

line. This may explain the specific hurdles

that we face in protecting privacy online,

and the surprising observation of seemingly

careless online behaviors by individuals

who claim to care about their privacy ( 5 ). In

our nondigital life, Altman ( 1 ) noted, behav-

iors that help us manage privacy boundar-

ies are common and ubiquitous. We often

engage in them with little conscious aware-

ness. We lower our voice during intimate

conversations and raise it when we want a

large audience to hear us; we cover a docu-

ment that we are reading to protect it from

prying eyes, or raise it up and emphatically

show it when we want to make a point. In-

stead, online, the evolutionary mechanisms

that Westin ( 4 ) and Klopfer and Rubenstein

( 6 ) identified cannot help us: We do not

see Google leaning over our shoulders to

track our sensitive searches; we do not hear

the US National Security Agency stepping

closer to listen to our videoconferences.

Yet, humans may be wired to rely in part

on those very sensorial cues to assess the

privacy implications of their behaviors. This

discrepancy may create a gap, or mismatch,

in our ability to manage privacy-sensitive

scenarios in digital realms relative to the

physical world—a hypothesis we have de-

veloped ( 11 ) and that Sharif, Green, and Jet-

tinghoff have also proposed ( 12 ).

This privacy mismatch has distinctive

features vis à vis other evolutionary mis-

matches of modernity. For example, where

the mismatch between human-evolved

physiology and the modern diet arises en-

tirely within physical systems, the very

boundaries of privacy have evolved with

the digital transition—from predominantly

physical to predominantly informational.

Through that transition, privacy’s role and

relevance in society have evolved as well—

as has the value that individuals and societ-

ies can harness from personal data.

This mismatch does not imply that online

disclosures are inherently damaging. Nor

should we fall into a naturalistic fallacy of

elevating senses to be sole arbiters of pri-

vacy decision-making, even online. Rather, a

privacy mismatch implies that, if the ability

to regulate privacy boundaries is in an indi-

vidual’s best interest, once that evolution-

arily rooted ability is impaired, individuals

become vulnerable, online, to new threats.

We refer, in particular, to ubiquitous in-

stances of microlevel influence on decision-

making which, in the aggregate, shape

portentous macrolevel dynamics. Although

offline and online surveillance risks do dif-

fer along numerous dimensions (includ-

ing the likelihood of material harm), the

privacy mismatch is, in fact, happening

at the same time as ramifications of in-

formational privacy issues become conse-

quential not merely at the individual but

at the societal level. Privacy externalities

(13)—the collective ramifications of indi-

vidual disclosures—are becoming evident,

as algorithmic personalization spawns filter

bubbles, amplification of disinformation,

and political polarization, with implications

ranging from public health to the safety of

democratic elections. And though the pri-

vacy mismatch arises from our transition

to digital interactions, the consequences

are no longer merely digital, but physical

as well: Consider the material threats asso-

ciated with “doxxing,” or even episodes of

genocidal violence fostered by data-driven

personalization algorithms ( 14 ).

An evolutionary account of privacy pro-

vides the underlying reason why a domi-

nant approach to privacy management in

the United States—notice and consent—has

failed to address these problems. Though

popular with industry, overreliance on no-

tice and consent mechanisms, disjointed

from baseline privacy safeguards, is ineffec-

tual and can backfire—because we are wired

to react to privacy invasions viscerally, not

just deliberatively. Those mechanisms bur-

den individuals with “responsibilization”

for problems they did not create and cannot

truly control. Responsibilization also cre-

ates unequal burdens, as it disadvantages

certain groups more than others: To get the

most out of notice and consent, for example,

people need resources—time, education, or

the economic leverage to not consent to

unfair policies—which are not equally dis-

tributed across the population. Even con-

siderable regulatory interventions such as

the General Data Protection Regulation

(GDPR) in the European Union resulted in

a proliferation of consent mechanisms that

burden users with manipulative interfaces

and implied consent ( 15 ). Notice and con-

sent approaches fail to recognize that much

of our privacy behavior is intuitive and vis-

ceral, rendering them archaic solutions for

a modern problem.

USING POLICY TO FOSTER

AND EMBED PRIVACY TECHNOLOGIES

So, what should be done? If, as a society,

we determine that privacy is still valuable

to us, then to maintain it we should embed

privacy by default into the fabric of our

digital systems. We cannot demand that

people overcome an evolved sense of pri-

vacy strongly reliant on sensorial cues un-

available online. Any approach that places

not just the ability to choose, but ultimately

the responsibility to protect, on individu-

als themselves, will—according to this ac-

count—fail. Privacy mismatches will keep

rising in frequency and importance with

the growth of the information economy,

well past the point where any approach re-

lying upon individual responsibility alone

could contain them ( 12 ). Individuals’ per-

sonal information will keep being collected

across too many instances and vectors for

our bounded cognitive resources to manage

efficiently or effectively. And humans will

keep tuning out even visible signs of elec-

tronic surveillance.

Attempts to reproduce online the visceral

cues of the physical world are therefore un-

likely, alone, to solve the problem: An evolu-

tionary mechanism of explanation does not

necessarily demand an evolutionary mecha-

nism of change. For those same reasons,

even well-meaning proposals [such as data

propertization from economics, nudges from

behavioral research, and simplified notices

from usability research ( 13 )], though appeal-

ing, may hardly make a difference absent a

combination of technology and policy inter-

21 JANUARY 2022 • VOL 375 ISSUE 6578 271