“Welcome to You:” 23andMe’s Bioconstitutional Moment ...! 1! “Welcome to You:” 23andMe’s Bioconstitutional Moment Joseph Vitti I. Introduction: The Personal Genomics Market

  1  

“Welcome to You:” 23andMe’s Bioconstitutional Moment Joseph Vitti I . Introduction: The Personal Genomics Market and Bioconstitutionalism

The development of genomic sequencing technologies in the late 1990s and early

2000s was heralded not only as a major breakthrough for scientific research, but also as

the harbinger of a new era of biological self-understanding. By examining the human

genome in its (staggeringly large) totality, researchers believed, we could gain novel

insights into all aspects of our biology, from our evolutionary history to the determinants

of our body’s reactions to disease and beyond. Researchers at that time had already long

understood the basic mechanics of DNA, the primary genetic molecule: Watson and

Crick’s 1953 paper in Nature describing its double helical structure had paved the way

for several decades of genetic research at the molecular level. This included the all-

important unraveling of the ‘genetic code,’ the rules that dictated how genetic sequences

were translated into specific proteins. While the advancements made during this early

period were foundational for genetics, many suspected that a more global approach

would be necessary in order to understand the relation between a complex biological

system and its genetic underpinnings (McElheny, 2010; see Wynne, 2005 for dissenting

argument). Thus, as technology to probe the genome came of age, it catalyzed a shift in

biological thinking from genetics to genomics,1 and created possibilities for asking new

questions about the biological basis of human life.

In the US and elsewhere, entrepreneurs as well as researchers were excited about

the development of genomic technologies and, in particular, their application to humans.

While a close interrelationship with the market (or markets) is understood to be a largely

universal characteristic of technology (Bijker et al., 1987), technologies to extract and

                                                                                                               1 These two related fields are distinguished by their scope: where ‘genetics’ typically involves the analysis of one or a few genes or genetic regions, ‘genomics’ typically involves analysis of many regions or the entire genome of an organism or group of organisms.

  2  

analyze genomic information in particular lent themselves readily to commercialization.

On the one hand such information was practically impossible for the layperson to access,

and similarly difficult to interpret, ensuring a monopoly on such information-generating

technology for well-resourced centers of calculation (Latour, 1987). But the information

these centers produced was not just esoteric but at the same time intensely personal, able

to give consumers insight into their own biological makeup and, as spokespeople for

scientific and corporate institutions alike would allege, their fate. Of the new products

and markets that genetic and genomic technologies enabled – from the manipulation of

genetically modified organisms (GMOs) to the development of prenatal screening –

‘personal genomics’ stands out because of this relationship of intimacy to the consumer.

Put another way, much of the allure of personal genomic technology stemmed from its

purported ability to tell its consumers who they were.

This unique relationship between the consumer and the technology – rendered

even more complicated by their interrelationships with other involved entities, such as the

companies that design and provide the technology, or the health care providers whose

input and expertise could be alternatively supplemented or undermined by information

generated by that technology – put companies aiming to provide personal genomics

services in a powerful position. The way that a company such as 23andMe, which forms

the subject of my investigation in this paper, positions itself in relationship to its

consumers can deeply influence those consumer’s conceptions of and attitudes towards

their own biological identity. In this way, companies providing personal genomics

services in the early 2000s stood at a bioconstitutional moment – an opportunity for a

radical reconfiguration of the ways that individuals relate to themselves and to other

entities with respect to biological identity and its regulation (Jasanoff, 2011, pp. 10-11;

cf. also 2004). Would personal genomics be a precursor to personalized medicine, in

  3  

which an individual’s genetic indicators of disease susceptibility and drug efficacy are

used in prescriptive clinical decisions? Is it within an individual’s rights to access his or

her own genetic data, and what sorts of qualifications is this right subject to? Does the

generation of that data give the company the right to perform research and/or

development based on that data, and how can appropriate consent be ensured? These

questions were among the many that would be decided, and continue to be decided, by

the interplay between government, companies and consumers in the US and abroad2, and

define the emergence of personal genomics services as ‘bioconstitutional’.

Importantly, the way that answers to these questions came into formation was

deeply influenced by competition among various business models of personal genomics

companies. Although many companies have offered personal genomics services for

various purposes (reviewed in Wright & MacArthur, 2012), 23andMe emerged as the

most commercially successful of these in the late 2000s, with companies like Navigenics

and deCODEme enjoying similar albeit diminished success. Corporate politics invariably

played a highly significant role in 23andMe’s success. Indeed, the alignment of 23andMe

with Google, solidified on the one hand with the 2007 marriage of 23andMe’s Anne

Wocjicki and Google’s Sergey Brin, and on the other with Google’s $3.9M investment in

23andMe in the same year, was in many ways definitive for securing 23andMe’s

hegemony in the personal genomics market (Kelleher, 2007). These factors

notwithstanding, 23andMe deployed a unique business model – one that sought to

cultivate an online community around their services, and emphasized the role of its

consumers as active participants in not just their own health care, but also in research –

                                                                                                               2 In many ways, the interrelationships of these agents and others in shaping these policies exemplify the concept of ‘co-production’ (Jasanoff, 2004). In this piece, particularly in section IV, I demonstrate the influence of a consumer identity and epistemology – or, more accurately, of 23andMe’s conception of their consumers’ identity and epistemology – on their business model and modes of operating. More generally, however, it is worth attending to the fact that this element of influence forms part of a larger, multidimensional network of mutual influence and coproduction. With this in mind, sections II and III attempt to situate the company in its broader scientific and sociocultural contexts.

  4  

that is informative for understanding its commercial success. The deployment and

success of this business model highlight, moreover, its influence on the questions posed

by the aforementioned ‘bioconstitutional moment’ that the advent of personal genomics

services engendered.

In what follows, I analyze 23andMe and its corporate success from an STS

perspective (Jasanoff et al., 1995; Hackett et al., 2008). To orient the reader, I begin with

a brief description of personal genomics technology and its use by 23andMe. Because the

company and its interactions with consumers and government agencies have attracted

much scholarly attention, I then outline some of the major ongoing debates surrounding

the company. Given the depth and breadth of the many criticisms provoking these

debates, it is remarkable that the company has continued to enjoy relative freedom from

government oversight, as well as continued support from a growing consumer base. To

understand their success in spite of major dissent, I turn my attention to their business

model, using the idiom of STS to identify the ways that 23andMe distinguishes itself

from competitor services and the ways that these business strategies appear to target a

predominantly American audience. I conclude by considering how these choices and the

commercial-cultural success they facilitated is redefining ideas of biological identity and

genomic research.

II. Genotyping Technology and 23andMe

Sequencing a genome means determining the order of nucleotide bases (the

molecules that form the ‘rungs’ of the double helix ladder) contained within the totality

of that organism’s DNA. There are four molecules that serve this purpose in DNA, which

are commonly represented by the letters A, G, C, and T (adenine, guanine, cytosine, and

thymine). Each nucleotide is always paired with its complement on the other strand of the

helix to form a ‘base pair;’ A-G or C-T. Although technologies have been improving at a

  5  

remarkable rate, determining an organism’s full genome sequence is no small task even

today: the human genome contains about six billion base pairs, spread across 23 pairs of

chromosomes3 (hence the name, “23andMe”).

Although sequencing technology was a necessary prerequisite to personal

genomics services, it is extremely uncommon at this time for a consumer to have his or

her genome sequenced in full. First, many parts of the genome (~8%) appear to be

structural rather than functional (‘heterochromatin’), and these regions are generally left

unsequenced even by the Human Genome Project because of the methodological

difficulties they present. This qualification aside, it is still uncommon for individuals to

have their genome sequenced in full. The vast majority of the genome is shared among

humans; only one in one thousand nucleotides appears to vary among individuals. These

regions are called single-nucleotide polymorphisms, or SNPs, and often appear at regular

locations in the genome that, thanks to full-genome sequencing of many individuals, have

been identified and can be targeted. Personal genomics technology takes advantage of

this fact and uses SNP arrays to genotype, rather than sequence, individuals at many

genetic regions.

In communicating results to consumers, 23andMe reaffirms that genetic variation

cannot explain all of human variation (see section IV.iii below). This is because the

environment plays many roles: (1) social and cultural factors are known to be influential

in the development of complex traits such as those pertaining to personality or cognition;

(2) even simpler biological processes such as drug metabolism are not purely genetically

determined, but rather are subject to contingencies of the organism’s environment such as

diet; (3) environmental factors, especially those in the uterine environment of a

                                                                                                               3  A child receives 23 chromosome from each parent for a total of 46. Thus, each parent donates roughly half of their genome to each offspring. The process of recombination, in which chromosomes exchange information before becoming sperm or egg cells, ensures that offspring always receive novel combinations of their parents’ genetic information.  

  6  

developing embryo, can have epigenetic effects – that is, they can make alterations to

chromosomes without changing the nucleotide sequence that affect when and how genes

are expressed. Further, it is worth pointing out that the genetic variation examined by

personal genomics services does not describe the totality of human genetic variation. In

addition to SNPs, individuals often differ with respect to the number of copies of a gene

they may contain (copy number variants, or CNVs). This happens as the result of

duplication or deletion of genetic regions. Thus, two individuals with the same SNP

profile at a given gene may differ with respect to the related phenotype (i.e., trait)

because they may have different numbers of the gene being expressed. In addition, some

individuals undergo rearrangements of parts of their chromosomes, such as inversions or

translocations. Chromosomal rearrangements and CNVs are often (though not always)

phenotypically meaningful, and are typically not detected by traditional genotyping

methods.

These limitations aside, 23andMe (and indeed most personal genomics services)

test for an individual’s SNP profile at a number of common sites, and interpret the

probable phenotypic import of this data. 23andMe consumers purchase a kit that must

first be activated online (including a process of consent, wherein users can determine to

what degree their genetic data will be private or public, including its use in research).

Once registered, customers deposit about 2.5 mL of saliva in a plastic tube included with

the kit. Closing the tube causes a package of enzymes to be released into the spit for

preservation and preparation. Users then send the tube to 23andMe’s headquarters in a

biological specimen bag provided with the kit. Once received, the sample is processed

using a SNP microarray designed by the company Illumina, which tests a total of

~960,000 SNPs on both copies of each chromosome. Two to three weeks later, customers

receive their results through 23andMe’s website. Results are binned by category: disease

  7  

risk, carrier status, drug response, traits, maternal ancestry, etc. Users access this

information through pages that provide interpretive guidance and other resources (e.g.,

one’s risk status for a given health disorder is presented compared to the overall average,

together with a host of relevant data and citations – see section IV.iii).

III. 23andMe: The Debates

As the most commercially successful venture of its kind, 23andMe has been

subject to critique on a number of fronts. Some of the questions and concerns raised by

theorists are applicable towards personal genomics services and direct-to-consumer

(DTC) genetic testing more generally, while others are directed towards 23andMe in

particular, concerning its specific business strategies and corporate activities – such as the

recent, unexpected patenting of polymorphisms (i.e., SNPs) associated with Parkinson’s

disease (Sterckx et al., 2012). These critiques are put forward by STS scholars as well as

commentators from many other fields, including bioethics, law, and the life sciences

themselves. In what follows, I outline the most salient of these critiques.

III.i Utility and Validity

Many commentators first question the axiom, ostensibly central to the motivations

of many 23andMe customers, that knowledge is empowering. Foster, Mulvihill and

Sharp, for example, consider the value of personal genomics data using a cost-benefit

analysis, in which they first distinguish between clinical and personal utility (2009). They

point out that from a clinical perspective, such data is at worst irrelevant, but rarely if

ever detrimental (assuming it is accurate and assuming that the relevant parties are well-

informed on its relevance to the health of the patient in question – two assumptions that,

as illustrated below, have invited skepticism). Alarmingly, however, such data can be of

negative personal utility in a number of scenarios: when the data provoke an increased

health concern accompanied by adverse psychological impact, or conversely when

  8  

misinformed consumers take a lack of risk indicators in their results to imply that they are

not at risk (whereas this may be an artifact of a limited dataset).

The possibility that an individual’s choice to receive personal genomics services

could prove detrimental is exacerbated by the possibility of misinformation (Ransohoff

and Khoury, 2010). The psychological trauma or stress induced by an indication of

heightened risk for a given condition could turn out to be in vain if that indication was

provided in error. Thus, the validity of the metrics employed by companies such as

23andMe are of paramount concern. Ensuring quality in personal genetic testing takes

place along two dimensions: on the one hand, companies must ensure that the technology

used to read an individual’s genetic information is correct (‘analytical validity’); on the

other they must ensure that their provided interpretation of the data is correct (‘clinical

validity’) (Grimaldi, 2011). A 2009 study comparing the results of two genotyping

services (23andMe and Navigenics) for thirteen diseases in five individuals found

divergent interpretations of risk. While the two companies agreed in their analysis of all

individuals for five diseases, 50% or less of the predictions for the remaining seven

diseases agree between the two companies (Ng et al., 2009). The authors speculate that

this divergence is due to clinical, rather than analytical, invalidity, and offer suggestions

for genotyping companies to improve the accuracy of their interpretive practices. More

generally, however, their results call into question the presumed adequacy of the tests

employed and the means used to interpret them.

Even under the (contested) supposition that the tests employed are of sufficient

analytical and clinical validity, some suggest that their utility remains limited (Hall &

Gartner, 2009). According to this line of thought, the information presented by such tests

is minimally, if indeed at all, ‘actionable.’ The authors point out that the correlations

identified by genome-wide association studies (i.e., those used to calculate an

  9  

individual’s risk) are typically weak, diminishing their predictive value. The reason for

this is that few health conditions (or, more generally, traits) are determined by singular

genes in Mendelian fashion; more typically many genetic regions act together to

influence an individual’s susceptibility in ways that are not necessarily additive. Given

the weakness of these associations, most customers will therefore be determined to have

‘average risk’ – which, Hall and Gartner point out, is not an ‘actionable’ piece of

information. Even in cases where an individual’s risk is significantly heightened,

moreover, the authors argue that there is no evidence suggesting that those who do make

lifestyle changes will have a better outcome.

Some have also problematized the metrics with which an individual’s risk is

calculated. The 23andMe ‘odds calculator’ and other tools used to interpret genetic data

often hinge on demographic assumptions that are not always applicable (Hart, 2011). It is

well-known that factors like ethnicity and age can significantly affect one’s susceptibility

to various conditions, but the limited availability of data means that, for many conditions,

the odds calculator will default to what numbers are known – which are often taken from

studies of European adults older than forty years old.

In summary, critics have challenged 23andMe and related services on the grounds

that their interpretation of data is minimally, if indeed at all, useful for consumers, and

can be skewed by inappropriate assumptions. An individual ultimately will or will not

contract a health condition, and any prior information about the relative likelihood of

such events is more apt to provoke undue anxiety than to spur any meaningful lifestyle

changes – indeed, it has not been proved that such changes are efficacious. For many

individuals, where personal genetic analysis is concerned, ignorance is far preferable to

knowledge – particularly where the analytical and clinical validity of that knowledge is

not well established.

  10  

III.ii Is This a Medical Device?

As is typical with the introduction of novel biological entities, the technology

23andMe develops and employs – their ‘Personal Genome Service®’ – is of contested

ontological status. This technology appears to involve both product and process: on the

one hand $99 buys customers a ‘spit kit’ and postage to send said kit to the 23andMe for

analysis, but at the same time it grants the consumer the right to access to the results of

those analysis (and the concurrent right to participate in the online 23andMe community).

The technology’s standing as product, process or hybrid becomes legally relevant

because it has bearing on the regulations it becomes subject to. If their technology is

construed simply as a device, a consumer product, then it becomes relatively free from

governmental oversight. By contrast, the fact that it involves a professional, expertise-

dependent interpretation of data rendered suggests that it is rather a service, and a

medical one at that – in which case, it becomes subject to the regulation of the FDA and

informed consent laws, as well as informal codes governing the relationships among

doctors and their patients. Thus, a certain degree of ‘ontological surgery’ has been

necessary for the company and its legal regulators to determine which rules it falls

subject to (Jasanoff, 2011).

23andMe has distanced itself from any legal classification as a medical device or

service, ostensibly because such classification would render it subject to much heavier

regulation and would therefore adversely affect its profitability and accessibility. At the

same time, however, the company has sought to employ medical rhetoric in an unofficial,

promotional capacity, emphasizing the health implications of its technology to customers

but insisting that their technology is in no way prescriptive. Thus, while the homepage of

their website puts forward the slogan “23andMe can help you manage risk and make

informed decisions,” elsewhere in their legal documentation they reaffirm that this risk

  11  

management service is not offered in a medical capacity: “what we do not and will not do

is provide medical advice to our customers. Though our service delivers personalized

data, the information it provides is tailored to genotypes, not to individuals” (23andMe,

2012a, 2012b).

From a legal standpoint, the debate over the question of 23andMe technology’s

status as medical was played out in summer 2008, when the California Health

Department sent cease-and-desist letters to 13 genetic testing firms, 23andMe and

Navigenics among them. These letters asked the companies to prove that a physician has

licensed the performance of each test, as well as that state clinical laboratory licensing

requirements were being fulfilled (Langreth, 2008). Popular response to the injunction in

support of 23andMe asserted that an individual’s genetic data is their own personal

property, and that a physician’s mandate should not be required (Goetz, 2008). By

August, 23andMe had received a state license allowing them to continue their business in

California (Pollack, 2008).

While this episode secured 23andMe’s identity as a non-medical agency within

the U.S., other countries have responded with continued dissent. In the European Union,

responses towards emerging biological technology have varied, reflecting the existence

of a plurality of publics and value systems (Jasanoff, 2005). The European response to

personal genomics services reflects this trend. While some countries, such as the UK,

have adopted permissive attitudes towards companies like 23andMe, others have adopted

and enforced stricter regulatory policies. In France and Switzerland, for example,

regulations dictate that the involvement of a medical specialist is necessary for personal

genomics services to be performed, although the claim that such services are not in and of

themselves medical have allowed some companies to circumvent this restriction

(Grimaldi, 2011). In Germany, these regulations have been more successfully enforced,

  12  

as evidenced by the passage of the Human Genetic Examination Act in 2009. This law

effectively places a ban on direct-to-consumer genetic tests (Wright, 2009). Additionally,

it prohibits the order of genetic testing by practitioners other than medical doctors (e.g.

dieticians or nutritionists).

As the differential response from European countries illustrates, the ontological

status of 23andMe’s Personal Genome Service remains contested. This is further

reflected by 23andMe’s recent submission of their first round of 510(k) documentation to

the US Food and Drug Administration (FDA) (23andMe, 2012c). This move reflects a

desire on the part of the company to receive approval at the federal government level, but

also indicates recognition of the ontologically loaded status of its technology and its

obvious biomedical implications.

III.iii Consumer Rights, Consent, and Use in Research

Another line of critique directed at 23andMe concerns its policy regarding use of

customer data for research (occurring under the aegis of the ‘23andWe’ subdivision,

further discussed in section IV.v). In recent years, the company has started publishing

association studies that implicate certain genetic variants in the development of certain

traits or conditions (e.g. Ericksson et al., 2010; Do et al., 2011; Ericksson et al., 2012).

These studies were initially controversial for three reasons (Gibson & Copenhaver,

2010). First, they had not received approval from an Institutional Review Board (IRB)

prior to the commencement of sample collection, as is required by US Human Subjects

protections regulations (i.e. 45CRF46).4 The second issue concerned consent: although all

23andMe participants were at the time required to consent broadly to the use of their data

in research, the consent form specifies that individuals will be contacted for re-consent if

                                                                                                               4  23andMe received a post hoc ruling of ‘not human subjects research’ from an independent accredited IRB on the basis of the fact that major identifying information, such as name or hometown, had been removed from data before analysis (Gibson & Copenhaver, 2010). Nonetheless, it is typically required for researchers to secure IRB approval prior to the commencement of data collection.  

  13  

their data is at any point to be shared with collaborators external to 23andMe. It was

contested at the time of the first 23andMe-backed publication whether all authors on the

paper were indeed internal to 23andMe, with some author’s official designation as

‘consultants’ to the company entering a gray area. Lastly, although it is typical for

researchers to publish raw data as a supplement to their analyses, 23andMe included only

aggregate data, as required by the wording of their consent document. Thus, academic

audiences has only filtered access to the data used to evidence the claims made in the

papers and did not have the option of verifying the claims put forward, or performing any

subsequent analysis.

The open-source journal PLoS Genetics ultimately decided to move forward with

publication in spite of these issues, on the grounds that no outright violations of the rights

of human subjects were reported (indeed the only data shared with individuals possibly

‘external’ to 23andMe was de-identified, a qualification that has now been built into the

23andMe consent document: 23andMe, 2012d) and that the above issues represented

rather issues of process that could be in this case overlooked, given that 23andMe was

piloting a new form of genetic association study, one based on self-selected participants

connected through the internet, that then-current regulations were too strict to

accommodate (Gibson & Copenhaver, 2010). Backed by 23andMe, authors have since

begun to lay out guidelines to govern the appropriate use of customers as research

participants, and 23andMe itself has since changed its policy on research to accommodate

individuals who wish to receive personal genomics services but do not wish to participate

in research (Tobin et al., 2012). Even without this modification to 23andMe policy, some

have suggested that the use of customers as research subjects is not in and of itself

problematic, but simply reflects a shift in the constitution of the research subject (Tutton

& Prainsack, 2011). On this view, participation in a 23andMe-type study is a decision

  14  

made by an enterprising and largely autonomous customer who make the informed

decision to accept participation in research studies as part-and-parcel of an economic

transaction.

These affirmations of 23andMe’s research legitimacy notwithstanding, others

have maintained that the use of customers in research is in and of itself ethically

unscrupulous. For example, Williams (2012) points out that, in conducting genetic

association studies, 23andMe utilizes not only the genetic information that they

themselves generate, but also the lifestyle and family history data that customers provide

– which, Williams argues, amounts to commercial exploitation. She also points out that,

while explicit patient consent is formally required for “research” – i.e., producing results

for academic publication – no such consent is needed for “research and development” or

“R&D” of commercial products. Thus, 23andMe recently secured a patent – their first

genetic patent, although subsequent investigation uncovered several unsuccessful prior

applications (23andMe, 2012e; Sterckx et al., 2012). The patent, entitled “polymorphisms

associated with Parkinson’s disease” (US-B-8187811), secures 23andMe’s legal claim to

the development of therapeutics stemming from this discovery, and is highly contested

because the customers whose genetic and phenotypic information were employed in

making these discoveries were unaware of its eventual use (ibid.). 23andMe is not only

charged with failing to communicate their intent to their consumers earlier (and so failing

to grant would-be dissenters the option to back out of any study that might lead to a

patent) but also with potentially unethical intent. Spokespeople for the company suggest

that the patent is a necessary step for the development of an eventual treatment, but do

not substantiate the reasons why this would be so (23andMe, 2012e). This has led many

to surmise that the patent is a means for 23andMe to ensure that it controls the

commercial benefits of the discovery, at best, while at worst it grants them certain

  15  

abusive potential: 23andMe could legally charge royalties or block performance of

genetic tests for Parkinson’s disease, for example (Sterckx et al., 2012).

IV. Strategic Representations

Given the wealth of criticisms lodged at 23andMe, many of them stemming from

powerful institutions like the California Board of Health, the company’s continued

prosperity bears some explanation. As indicated previously, material and political factors

such as the alignment of 23andMe with Google are understood to be highly influential.

Ultimately, however, 23andMe had to make strategic choices about how to leverage the

resources that it was able to amass, and it is these choices that I here investigate.

Specifically, I examine how 23andMe choose to represent itself to the public in order to

cultivate consumer interest and trust, invoking the idiom of STS to analyze 23andMe’s

business model.

IV.i Know Your Audience: The Civic Epistemologies of the US

My argument in this section is that 23andMe, born of the Silicon Valley, chose to

represent itself in ways that appeal to an American value system or, more specifically, an

American civic epistemology (Jasanoff, 2005). The term civic epistemology refers to the

culturally and historically specific ways that nations or other groups of people come to

accept certain knowledge claims. The fate of any novel technology is undetermined until

its proponents manage to propagate its use; these proponents may seek to accumulate

public support and acceptance in myriad ways, and each of these ways will succeed in

certain sociopolitical contexts and not in others.

As Jasanoff and others have argued, the prevailing civic epistemologies of various

nations can be revealed by comparative analysis of the ways in which similar

technologies and scientific advancements were received by different publics (ibid.). In

the United States, a plurality of methods are employed by proponents of novel

  16  

technologies, but these methods have certain commonalities among them. Among these

are an assumption of distrust for large institutions and a concurrent demand for

transparency, as well as a preference for de-centralized power structures. This

contentious mode of evaluating and accepting knowledge claims may relate to the

reaffirmation, whether conscious or otherwise, of a cultural heritage that holds sacred the

abstract notion of democracy and prides itself on maintaining a certain distinction from

its European forebears.

How could a personal genomics company like 23andMe take advantage of the

emphasis on democracy and mistrust of monolithic, impenetrable social institutions that

are common to American civic epistemologies to secure its own commercial success?

While some have argued that the sociopolitical valence of a given technology can be

considered inherent to that technology (Winner, 1980), I argue rather that 23andMe

deliberately constructed a social and political image around their personal genomics

technology. This image can be seen through contrast with 23andMe’s less successful

competitors, like Navigenics and deCODEme. In what follows, I highlight a few of the

decisions unique to 23andMe that helped to effect a public image that would be attractive

to an American civic epistemology – one that construed itself as essentially

democratizing. As I show below, this image was built through strategic decisions the

company made with respect to its self-representation in promotional materials, in its use

of social media, in its construction of an interactive and thoroughly didactic web-based

interface, and in its representation of its research activities.

IV . ii The Right to Know: 23andMe’s Performance of a Public Image

First and foremost, 23andMe sought to portray itself as a democratizing agent

through its self-representation in promotional materials such as its website. This mode of

  17  

deliberate and conscientious presentation is an instance of the phenomenon Hilgartner

refers to as the ‘performance’ of public authority and identity (2000).

Central to this performance and its democratic overtones is an emphasis on ‘you’

– the customer (as captured in the name, ‘23andMe’). The front page of their website

invites the audience member to “be your own best advocate,” in virtue of 23andMe’s

capacity to “help you manage risk and make informed decisions” (23andMe, 2012a).

These ideas are further elaborated on the “core values” page, in which the company

reaffirms a number of beliefs, the first of which are that access to one’s own genetic

information is a good thing, and that this information should be controlled by the

individual. Other values listed subsequently include the need for dialogue on the ethical,

legal, social and policy implications of personalized genetic understanding and “giving

everybody the opportunity to contribute to improving human understanding” (23andMe,

2012f).

While competitor agencies such as deCODEme have placed similar emphasis on

the customer and her right to know her DNA, 23andMe’s model invokes the rhetoric of

democracy in a more positive light. As deCODEme’s front page displays the motto

“YOUR DNA / YOUR HEALTH” in a drab shade of red, together with images of

customers labeled as survivors of health conditions, the message is that their personal

genomics service is meant to empower the disenfranchised (deCODEme, 2012). By

contrast, 23andMe’s model characterizes its consumer as already empowered in virtue of

being on their website, with colorful, illuminated graphics – among which the only one

depicting ‘health’ is an abstract representation of a fit woman running, rather than

someone diagnosed with an adverse health condition. 23andMe also downplays the

somber nature of their biomedical applications by putting them forward as one of three

  18  

dimensions of participation: ancestry (“connect with your past”), health (“learn for the

present”) and research (“participate for the future”) (23andMe, 2012a).

This relative de-emphasizing of the health component of personal genomics is one

of the major decisions that distinguishes 23andMe from its competitors, with companies

like Navigenics portraying themselves primarily as agencies that can supplement

information given by physicians. 23andMe managed to substantiate their emphasis on

research and ancestry in the first case by publishing novel results in peer-reviewed

journals and in the latter by finding endorsement from a prominent academic. Harvard

professor Henry Louis Gates Jr. served as the executive producer and host of Finding

Your Roots, a PBS series that used 23andMe to investigate the ancestry of celebrities like

Barbara Walters and Wanda Sykes. This mode of endorsement by recognized authorities

is one of the strategies that Hilgartner considers central to the performance of authority,

and indeed on their web-based promotional materials 23andMe uses images of Gates and

PBS (23andMe, 2012g).

IV.iii Making Up Consumers: Between the Deficit Model and the Black

Box

In order to construct a corporate image that was perceived as essentially

democratizing, 23andMe had to address the question of just who was the demos that they

were addressing – and how best to earn the respect and trust of that demos.

As a general strategy, following the example set by Silicon Valley corporate

juggernauts like Google and Apple, 23andMe first aimed to make its technology as

simple, streamlined, and user-friendly as possible. For example, while they could have

offered a number of different purchase options at multiple price gradations (i.e., with

more expensive purchases granting the user a higher resolution of SNP genotyping, or

additional interpretive features made available through the website), they opted instead to

  19  

offer one single package for all consumers. Thus, the complex and personal decision-

making process confronting a would-be customer of 23andMe becomes simplified in

virtue of a limited set of options: purchase, or don’t purchase, all-or-nothing. Once

customers have ‘bought in’, they can opt to customize their level of involvement – for

example by participating to a greater or lesser degree in the online community (discussed

below). However, the prerequisite decision that will allow customers to reach this stage is

a binary one.

In some ways, this broad strategy of user-friendliness and accessibility structured

23andMe’s response to the question of how to construe their consumer. What level of

relevant background knowledge should be assumed for 23andMe’s audience? Is the

average user knowledgeable, or else indifferent – in which case 23andMe can avoid

overcrowding its presentation with extraneous explanations of terms like “allele” and

“gene”? Or is it more prudent to assume a ‘deficit model’, whereby the user is assumed

by default to be ignorant (Miller, 2001)? On the one hand, user-friendliness necessitates

simplicity, but on the other it necessitates completeness. 23andMe’s adopted strategy

reflects a compromise between these two opposites. They aim to provide explanations for

all the topics covered, but to do so in layers, such that users who have questions after

reading a summary page can easily link through and access more detailed explanations,

as well as community pages (i.e. online discussion boards). Resources are also provided

for users with relevant expertise, including citations and links to referenced scientific

publications for each genetic trait and the option of downloading one’s own raw data.

Links to didactic materials are ubiquitous in the 23andMe interface, with a

number of short ‘Genetics 101’-style tutorials accessible to all audiences, as well as much

more specific information available for customers investigating their own results. Each

result (e.g. a user’s risk for a given condition, a user’s probability of having a certain trait,

  20  

information that can be inferred about a user’s maternal or paternal ancestry, etc.) is

presented first with a summary page detailing the result described and what is known

about its genetic underpinnings. This is always supplemented by a “Genes vs.

Environment” blurb, which reaffirms the general importance of non-genetic factors in

complex traits and explains what is currently known about the gene-environment

interplay with respect to the trait of interest. Users can also pass through a number of

other ‘tabs’ within the results report, including a more detailed breakdown of the user’s

data and its significance, a technical report, a list of resources (including tools for

locating genetic counselors and support groups), and a community page, where users can

discuss their results and experience.

The appeal to an American civic epistemology is apparent not only in 23andMe’s

layered organization of informational materials, but also in the formatting of those

materials. Figures such as relative risk or prevalence of a condition are accompanied by

graphic representations wherever possible. These graphics are interactive: one can, for

example, request to see how relative risk would change if one were of a different

ethnicity or age group. These images help to convey information as well as to effect a

sense of epistemic authority: they are outgrowths of Latour’s “immutable mobiles”

(1987), now rendered even more mobile through the use of the internet5 and rendered

quasi-mutable through their interactive nature. Reassuringly, they appear to be distinct

from Ezrahi’s concept of “outformations” (2004), in that their generation is not a simple

or inexpensive process, reflecting rather the accumulated labor of 23andMe’s research-

focused employees. Outformations are low in their knowledge content, and appear

propagandistic rather than informative – by contrast, the diagrams presented by 23andMe

are supplemented with citations and opportunities for discussion for any dissenters who

                                                                                                               5  Results pages accessed via the customer’s interface have links prominently displayed that allow users to share their information, both internally with other 23andMe customers and externally through email.  

  21  

question the validity or value of the information represented. Thus, 23andMe’s graphic

strategies assume a viewer who is capable of investigating data, but who would first have

that data presented in a streamlined, visual, and easy-to-parse format.

IV.iv Creating a ‘Gene Pool’: 23andMe and Social Networking

Catering to a democratically-oriented civic epistemology necessitates not only

careful construction of the demos, but also the provision of opportunities for that demos

to participate in the processing of data and knowledge. As indicated above, these

opportunities are realized first through the creation of a user-friendly, interactive data

visualization and exploration interface. But beyond allowing the consumer to interact

with the data, 23andMe also provides opportunities for consumers to act with each other

– for example, through the aforementioned bulletin boards for each genetic result.

Perhaps through the influence of their allies at Google (or perhaps in a response more

generally to the contemporary climate of social networking), 23andMe constructed a

social network as part-and-parcel of their personal genomics service. This emphasis on

networking is apparent from the start: in order to have a kit analyzed, one must first

register that kit online – a step which includes the creation of an online user that will be

used to access results, but can also be used to connect with other users.

The social networking component of 23andMe extends beyond results-based

bulletin boards. Users are also encouraged to interact with each other through the creation

of user-defined interest groups (e.g., for discussing a certain disease or ancestry group)

and tools that allow consenting users to share and compare their genetic data. Users can

also decide to participate in a ‘relative finder’ feature that matches individuals with other

consenting participants on the basis of similarity and inferred shared ancestry.

Lastly, participants have the option not only of interacting with their data and with

each other, but also with the company itself – albeit in extremely limited fashion.

  22  

Announcements of multiple kinds – new research initiatives, changes in governance

structure, etc. – are posted via a publicly accessible blog, where users have the option of

posting comments. It is worth noting that these arenas are subject to the moderation of

23andMe officials, and thus their utility as authentic public fora appears limited.

Nonetheless, they offer a pulpit for users to voice concerns, as indeed many did following

the announcement of 23andMe’s first patent (23andMe, 2012e). These concerns may or

may not be addressed to satisfaction by company representatives.

More typically, the interaction between customer and company is unidirectional,

as in the case of user surveys. 23andMe prominently encourages users to participate in

surveys on various topics that enable their aggregation of phenotypic data. These surveys

are further incentivized on the 23andMe mobile application, where participants earn

digital ‘badges’ for completing milestone numbers of surveys. While on the one hand the

invitation to participate in these surveys can be seen as a democratizing move, inasmuch

as it encourages active engagement rather than passive receipt of results, there are reasons

to question this relationship. In what follows, I consider 23andMe’s research model and

the arguments for and against its supposedly democratic nature.

IV.v 23andWe: A New Mode of Research Participation

Perhaps the most conspicuous piece of the 23andMe business model that

distinguishes them from competitors is their emphasis on the performance of novel

research. By supplementing the genetic data generated through analysis of customer-

donated specimens with customer-provided phenotypic information, 23andMe has

amassed enough data to begin to perform genome-wide association studies (GWAS),

which identify SNPs that are significantly associated with various traits (including

disease susceptibility). From the point of view of research, this ‘crowdsourcing’ strategy

opens up new possibilities: on the assumption that self-reported information is accurate,

  23  

and provided that corrections are made to account for an atypical, self-selected

population structure, 23andMe’s do-it-yourself approach overcomes the traditional

limitations of GWAS (small sample sizes and logistically cumbersome data collection).

There are, however, reasons to be skeptical of 23andMe’s approach. Consistent

with the aspects of their business model outlined above, 23andMe presents their approach

to research as part of a larger democratic turn in genetics: by conducting surveys, the

company “gives customers the opportunity to leverage their data by contributing it to

studies of genetics” (23andMe, 2012h). By providing this ‘opportunity’ to users,

23andMe portrays itself as catalyzing a revolution in disease research (23andMe, 2009).

Such democratizing rhetoric obscures the way that the company itself stands to benefit

from user participation in those surveys: namely, through the production of novel

research discoveries that can be leveraged as publications or patents, either of which

bolster the brand and profitability of 23andMe. With this in mind, some have suggested

that the use of surveys amounts to little more than a ploy for free labor (Harris, Wyatt &

Kelley, 2011). On this view, 23andMe’s use of rhetoric and other devices (e.g. the

aforementioned ‘badges’ one received for completing milestone numbers of surveys) are

devices used to create a sentiment of reciprocity and social ties with the aim of coercing

the user into freely giving information that can be capitalized upon.

As the above argument reminds us (and, indeed, as this paper demonstrates),

23andMe is, at heart, a corporate venture, that will aim to represent itself in the most

profitable way possible. Whether this fact completely invalidates the ideals of democracy

in medicine and research that they espouse – or, perhaps, whether it simply renders them

somewhat flimsier than they would have the consumer believe – is debatable. There are,

however, reasons to believe that their research model is not so sinister. These reasons

come from the recognition that, although there is plainly an asymmetrical distribution of

  24  

power between corporation and consumer, the average consumer is not so

disenfranchised as, for example, the Foucaldian conception of ‘biopower’ would suggest

(Jasanoff, 2005). Consumers are capable of making decisions for themselves, and

although 23andMe helps itself to slippery rhetorical and graphical devices to influence

those decisions, they still remain the consumer’s sovereign right.

To demonstrate this point, it is worth considering that ultimately consumers can

determine not to participate in online studies at all, and can also decide to withdraw their

genetic and phenotypic data at any point (23andMe, 2012d). Additionally, since

consulting with an IRB (spurred on by their early attempts at publication), 23andMe has

added an option whereby individuals can receive their genetic results in a private manner,

without those results being contributed for research purposes.

Although these options are put in place, many consumers decide to participate in

the surveys anyway. There may be many reasons why: succumbing to the social pressure

effected by 23andMe on the one hand and believing in the importance of genetic research

on the other among them. Somewhat more simply, studies have shown that people often

enjoy participating in crowdsourced efforts; presumably this pleasure is modulated by the

degree to which the company and other participants reaffirm the value of participation

(Brabham, 2010). Some have also suggested that participation in 23andMe’s surveys is

spurred on by the recognition of a social contract (Tutton & Prainsack, 2011). On this

view, enabling 23andMe to conduct genetic research is something consumers do in return

for the services rendered to them – genotyping and interpretation of data – at a much

lower price than such research services typically require.

Ultimately, a normative assessment of 23andMe’s research model must

accommodate the ambiguous intentions that undergird it. On the one hand, the research

23andMe is productive represents genuine contributions to the scientific literature.

  25  

Previously published results (e.g. Ericksson et al., 2010; Do et al., 2011; Ericksson et al.,

2012) demonstrate rigorous and useful analysis, made publicly available through open-

source journals. On the other hand, however, the lack of public availability of their

datasets represents a serious limitation to the public utility of their research; given the

amount of reconsideration their consent form has received since they first moved to

publish, one would expect that they would arrange for de-identified data to be made

public if the common good really is their first priority. Along similar lines, their recent

patent suggests that their own corporate profitability was determined to trump public

research (their unsubstantiated claim that a patent is the most secure route to improved

therapeutics notwithstanding – 23andMe, 2012e). The patent likely represents the

decision to block external agencies from capitalizing on a discovery that belongs

arguably to the consumers who donated data as much as the company that generated and

analyzed that data (Sterckx et al., 2012).

V. Conclusion: 23andMe’s Bioconstitutional Moment

In this paper, I have examined the ways that 23andMe’s representational

strategies reflect an awareness of an American civic epistemology. 23andMe has

distinguished itself from its competitors by constructing a technology targeted at a

democratically-minded constituency, and in the process, has opened up new opportunities

for simultaneously advancing research and bolstering its own profitability.

The emergence of 23andMe as the victor over its competitors comes perhaps as

no surprise in light of the strategic decisions I have discussed above. The significance of

23andMe’s victory, however, is perhaps not so obvious. The advent of personal genomics

services defined a bioconstitutional moment, with the potential to redefine the

relationships among corporations, physicians, research institutions, individuals, and our

own biological selves. How has 23andMe redefined those relationships?

  26  

Inevitably, the first major shift among these relationships is one of

commercialism. Access to one’s biological information comes with an (admittedly

declining) price tag. While self-knowledge has long been a means of profit (consider

ancestry searches, personal diet consultants, or, historically, fortune-tellers of various

sorts), the extension of this phenomenon to genomic technology represents a

biotechnological milestone. In some ways, the sharp realization that one’s biological

identity can be commoditized is softened by 23andMe’s emphasis on its ‘non-corporate’

ventures such as its online community, as well as the bona fide performance of

meaningful genetic research – but nevertheless the point remains that intimacy with one’s

genetic self has become something that can be bought or sold.

Less definitive is the effect of 23andMe upon the medical industry. While

personal genomics was meant to be the first step towards personalized medicine,

commercial services such as 23andMe are rarely if ever employed in the clinic, and

indeed many speculate that their utility will remain limited until full-sequence data can be

affordably obtained (Ng et al., 2009). While the possibility of FDA approval suggests that

23andMe’s technology may later gain greater acceptance in the medical community, at

present it appears to be at best a supplemental source of information that health-conscious

consumers may choose to employ on an individual basis.

On the other hand, the deployment of a new model of large-scale GWAS has

already begun to redefine the research community’s approach to genetic studies. It is

improbable that more traditional, local cohort-based studies will cease altogether, but the

23andMe model has provided an alternative that is gaining momentum – and will likely

continue to do so as the company expands. Accordingly, it is the research subject whose

constitution is most shaken by the success of 23andMe. In classical studies, subjects are

passive donors of information, typically recruited with some financial incentive, and

  27  

uninformed of the results of the study to which they contribute. In 23andMe’s studies,

subjects participate of their own volition (indeed, they pay for the opportunity), and are

kept connected to a centralized community where results are shared and discussed.

The research subject has shifted from a passive recipient to an active participant,

engaged both in research and in a corporate transaction. As long as commentators and

participators cultivate an awareness of the corporate dimension to 23andMe – which its

spokespeople would downplay – such transactions can continue to the (by no means

equal) benefit of the inquiring customer, of the biomedical research community, and of

the company alike. Whether the broader commercialization of the biological self is a

worthy price to pay for the reinvention of the genetic research subject will be born out in

coming years – although 23andMe’s potentially shady motives in securing a patent

suggest that there are reasons to maintain skepticism.

  28  

Cited 23andMe. 2009. ‘Announcing a do-it-yourself revolution in disease research.’ 7 July 2009. Available at http://blog.23andme.com/23andme-and-you/introducing-a-do-it-yourself-revolution-in-disease-research/; accessed 12/15/12. 23andMe. 2012a. 23andMe homepage. Available at https://www.23andme.com/; accessed 12/10/12. 23andMe. 2012b. Open letter to the medical community. Available at https://www.23andme.com/for/physicians/; accessed 12/12/12. 23andMe. 2012c. ‘23andMe takes first step toward FDA clearance,’ 30 July 2012, available at: https://www.23andme.com/about/press/fda_application/; accessed 12/12/12. 23andMe. 2012d. Consent document. Available at https://www.23andme.com/about/consent/; accessed 12/13/12. 23andMe. 2012e. ‘Announcing 23andMe’s first patent,’ 28 May 2012, available at: http://blog.23andme.com/news/announcements/announcing-23andmes-first-patent/; accessed 12/12/12. 23andMe, 2012f. Core values. Available at https://www.23andme.com/about/values/; accessed 12/15/12. 23andMe, 2012g. Understand the story of your past. Available at https://www.23andme.com/ancestry/; accessed 12/15/12. 23andMe. 2012h. 23andWe research. Available at https://www.23andme.com/research/; accessed 12/13/12. Bijker, Wiebe; Hughes, Thomas; & Pinch, Trevor. 1973. The social construction of technological systems: new directions in the sociology and history of technology. Cambridge, MA: MIT Press. Brabham, Darren. 2010. Moving the crowd at Threadless: motivations for participation in a crowdsourcing application. Information, Communication & Society, 13(8): 1122–1145. deCODEme, 2012. deCODEme homepage. Available at http://www.decodeme.com/; accessed 12/13/12. Do, Chuong; Tung, Joyce; Dorfman, Elizabeth; et al. 2011. Web-based genome-wide association study identifies two novel loci and a substantial genetic component for Parkinson’s disease. PLoS Genetics 7(6): e10002141. Ericksson, Nicholas; Macpherson, J. Michael; Tung, Joyce; et al. 2010. Web-based, participant-driven studies yield novel genetic associations for common traits. PLoS Genetics 6(6): e1000993.

  29  

Ericksson, Nicholas; Benton, Geoffrey; Do, Chuong et al. 2012. Genetic variants associated with breast size also influence breast cancer risk. BMC Medical Genetics, 13:53. Doi:10.1186/1471-2530-13-53 Ezrahi, Yaron. 2004. Science and the political imagination in contemporary democracies. In Jasanoff, Sheila, ed. States of Knowledge: The Co-Production of Science and Social Order, Routledge. Foster, Morris; Mulvihill, John; & Sharp, Richard. Evaluating the utility of personal genomic information. Genetics in Medicine 11: 570-574. Gibson, Greg & Copenhaver, Gregory. 2010. Consent and internet-enabled human genomics. PLoS Genetics 6(6): e1000965. Goetz, Thomas. 2008. Attention, California Health Department: my DNA is my data. Wired, 17 June 2008. Grimaldi, Keith; Look, Markus; Scioli, G. Antonio; et al. 2011. Personal genetics: regulatory framework in Europe from a service provider’s perspective. European Journal of Human Genetics 19(4): 382-388. Hackett, Edward; Amsterdamska, Olga; Lynch, Michael & Wajcman, Judy. 2008. The handbook of science and technology studies. Cambridge, MA: MIT Press. Hall, Wayne & Gartner, Coral. 2009. Direct-to-consumer genome-wide scans: astrologicogenomics or simple scams? The American Journal of Bioethics 9(6-7): 54-56. Harris, Anna; Wyatt, Sally; & Kelly, Susan. 2012. The gift of spit (and the obligation to return it). Information, Communication & Society. Doi:10.1080/1369118X.2012.701656 Hart, Elly. 2011. Why 23andMe genetic testing is a waste of time and money. Gizmodo, April 7. Accessed December 11, 2012. Hilgartner, Stephen. 2000. Science on stage: expert advice as public drama. Palo Alto: Stanford University Press. Jasanoff, Sheila; Markle, Gerald; Peterson, James; & Pinch, Trevor. 1995. Handbook of science and technology studies. Thousand Oaks, CA: SAGE Publications, Inc. Jasanoff, Sheila. 2005. Designs on nature: science and democracy in Europe and the United States. Princeton: Princeton University Press. Jasanoff, Sheila, ed. 2004. States of knowledge: the co-production of science and social order. London: Routledge. Jasanoff, Sheila, ed. 2011. Reframing rights: bioconstitutionalism in the genetic age. Cambridge, MA: MIT Press. Kelleher, Kevin. 2007. Google, Sergey and 23andMe: why it all makes sense. Gigaom.com, 24 May 2007.

  30  

Langreth, Robert. 2008. California orders stop to gene testing. Forbes 14 June 2008. Latour, Bruno. 1987. Science in action: how to follow scientists and engineers through society. Cambridge, MA: Harvard University Press. McElheny, Victor. 2010. Drawing the map of life: inside the human genome project. Philadelphia: Perseus Books. Miller, Steve. 2001. Public understanding of science at the crossroads. Public Understanding of Science, 10(1): 115-120. Ng, Pauline; Murray, Sarah; Levy, Samuel; & Venter, J. Craig. 2009. An agenda for personalized medicine. Nature 461: 724-726. Pollack, Andrew. 2008. California licenses 2 companies to offer gene services. New York Times, 19 August 2008. Ransohoff, David; & Khoury, Muin J. 2010. Personal genomics: information can be harmful. European Journal of Clinical Investigation 40(1): 64-68. Sterckx, Sigrid; Cockbain, Julian; Howard, Heidi; et al. 2012. “Trust is not something you can reclaim easily”: patenting in the field of direct-to-consumer genetic testing. Genetics in Medicine, doi:10.1039/gim.2012.143. Tobin, Sara et al. 2012. Customers or research participants?: Guidance for research practices in commercialization of personal genomics. Genetics in Medicine 14: 833-835. Tutton, Richard & Prainsack, Barbara. 2011. Enterprising or altruistic selves? Making up research subjects in genetics research. Sociology of Health & Illness. 33(7): 1081-1095. Williams, Patricia. 2012. How our genetic maps are being sold to the highest bidder. The Nation, June 25. Winner, Langdon. 1980. Do artifacts have politics? Daedalus, 109(1): 121-136. Wright, Caroline. 2009. Ban on direct-to-consumer genetic tests in Germany. Available http://www.phgfoundation.org/news/4562/; accessed 12/12/12. Wright, Caroline & MacArthur, Daniel. 2012. Direct-to-consumer genetic testing. In Best, Hunter & Swensen, Jeffrey, eds. Molecular Genetics and Personalized Medicine. London: Humana Press. Wynne, Brian. 2005. Reflexing complexity: post-genomic knowledge and reductionist returns in public science. Theory, Culture & Society 22(5): 67-94.