Data Mining and the Criminalization of Health Care

As pernicious as it is, influencing consumer behavior is now one of the least harmful uses of our data. The implications of big tech’s data mining are vast, but one area of particular salience is reproductive rights in the wake of Dobbs v. Jackson Women’s Health Organization. Without federal protections, the potential for criminal and civil liability related to abortions is vast. As the advocacy organization Pregnancy Justice notes, “women have been arrested, prosecuted, convicted, detained, or forced to undergo medical interventions because of their pregnancy status or outcomes” in more than 1,700 cases while Roe v. Wade was good law. 

Since the Supreme Court overturned Roe, abolishing the right to abortion as an extension of the right to privacy, a hodgepodge of state laws—some predating the 1973 ruling in Roe—have gone into effect. Varying state laws, ongoing litigation, and continuing legislative efforts to further restrict abortion access make it difficult to understand the legal ramifications for performing or obtaining an abortion. As of October 2, 2022, 15 states had entirely banned abortion or instituted six-week gestational bans. Going beyond denying pregnant people the right to control their own bodies, Texas’s SB-8 and its progeny in Oklahoma and Idaho deputize private citizens to sue anyone who assists with an abortion, including the provider. The maneuver, invented by anti-choice zealots to evade judicial review, has been incorporated into copycat bans introduced in 15 states. Further, Missouri lawmakers are exploring legislation that would allow Missouri residents to sue out-of-state providers for assisting a Missouri resident in procuring an abortion even across state lines. Inconsistent state laws coupled and the lack of real data privacy place pregnant people and those who would help them access health care in real danger. 

Inconsistent state laws coupled and the lack of real data privacy place pregnant people and those who would help them access health care in real danger. 

Going forward, data will play an important role in prosecuting providers—and possibly patients. Warrants are required to obtain cell phone location data without a user’s consent, but two types of warrant for digital information—geofence and keyword—make it possible to essentially set traps for people seeking and providing abortions. Geofence warrants allow law enforcement to extract information about all devices that come within a defined perimeter (for example, an abortion clinic) during a specified timeframe. Using keyword warrants, law enforcement can obtain information about every person who searches specified terms (for example “abortion near me”). 

Among abortion advocates, the consensus is that dragnet traps of the type set by geofence and keyword warrants do not pose the greatest danger for prosecution. Law enforcement can access troves of data with few obstacles, and will likely do so in individual cases. For example, when a person comes into an emergency room post-miscarriage, the state may investigate whether she self-administered an abortion. Evidence may come from period tracking apps like Flo or other health monitoring and medical records apps, to which federal privacy protections do not apply, or from search history, messages, purchases, and more. Prosecutors can then use that data to bolster criminal allegations. 

For example, in 2017, an Oklahoma mother who experienced an at-home stillbirth was indicted for second degree murder after a jury inferred criminal intent from her internet search history, which included terms like “buy abortion pills,” “mifepristone online,” “misoprostol online.” Though the charges were ultimately dropped, similar indictments and even convictions could stem from data mining and other digital evidence. Prosecution will not be limited to miscarriages and abortions; already, women have been prosecuted for conduct that would be legal but their pregnancy, such as consumption of alcohol.

Moving forward, data from health apps can be used to a establish that a pregnant person knew of the pregnancy and took steps to end it, or to otherwise endanger the fetus in violation of post-Roe laws. If Flo’s data reveals a woman accused of drinking while pregnant had a regular menstrual cycle until her most recent period, which she missed, prosecutors can argue that she knew she was pregnant and had intent to harm the fetus. The same is true if she searched “abortion,” “abortion pills,” or any number of pregnancy-related terms, if she bought a pregnancy test or if she bought fewer menstrual hygiene products than usual in the preceding months, or if she visited a Planned Parenthood clinic, all innocuous and legal activities in and of themselves. 

Procuring incriminating data is not difficult. Law enforcement can access most data by merely requesting it from the platform (or the user) or purchasing it from a data broker, though in some cases, they may need a subpoena, court order, or warrant, depending on the service’s privacy policy and encryption practices, and whether constitutional protections apply. In August 2022, police brought charges against a Nebraska woman and her teenage daughter under the state’s pre-Dobbs 20-week gestational ban. In April 2022, the woman bought her daughter medication abortion pills online and used Facebook Messenger to instruct the teen on the use of the pills: “The 1 pill stops the hormones and then you gotta wait 24 hour 2 take the other.” A friend of the girl, who allegedly saw her take the first pill, reported the mother and daughter after police began investigating whether the teen had experienced a premature stillbirth. Because of the friend’s tip-off, police obtained a warrant for the pair’s Facebook messages and used them to charge the mother with two additional felonies related to her procurement of the abortion pills. 

As early as 2012, statisticians employed by Target’s marketing team were using data to accurately predict whether customers were pregnant and when they were due to give birth.

The nature of machine learning and AI’s predictive capabilities means complete information—like a purchase or a single incriminating search—may not be necessary to determine whether a person was pregnant or had an abortion. Tech companies are increasingly involved in health care, with Facebook, Google, Amazon, Microsoft, IBM and more working with hospitals and providers to harness personal data and records to make health predictions that are used in care decisions. In addition to invading patients’ privacy, health care algorithms exhibit systematic racial discrimination against Black people. Yet hospitals like the University of Chicago’s are all too eager to hand over barely-anonymized patient data to companies like Google for use in predictive analytics, indicating that AI is likely to pervade health care for the foreseeable future. As early as 2012, statisticians employed by Target’s marketing team were using data to accurately predict whether customers were pregnant and when they were due to give birth. The data available then was far more limited than the vast behavioral surplus now available: the eerily accurate 2012 predictions were made based on limited demographic and personal characteristics combined with purchasing history on just 25 items. 

Big Tech’s Hypocritical Messaging 

People are rightfully scared about the implications of data mining in a post-Roe world. In the wake of the Dobbs decision, companies like Google and Meta (which owns Facebook and Instagram) issued feel-good PR statements promising to cover the expenses of employees who need to travel out of state to obtain an abortion. Bolstering its pro-choice messaging, Google also implemented a policy allowing employees to apply to transfer between offices “without justification,” ostensibly so that they can proactively move to states where abortion remains legal. Finally, Google also pledged to delete the location data of users who visit abortion clinics (but not those users’ search histories, including those conducted on private browsing modes). It’s not difficult to extrapolate a patient’s location from a gap in location data corresponding with the moment they entered the “protected” perimeter surrounding a clinic. Indeed, Google prides itself on allowing advertisers to target potential customers within specific buildings (others can target within roughly a city block), so blank location data is far from the norm. And given the amount of data Google retains from each search and click, users will be at risk even if authorities cannot pinpoint their location to an abortion clinic. 

Though they were eager to showcase their “progressive values” immediately following Dobbs, tech companies funded those responsible for the decision. Since 2016, Google has given over half a million dollars to anti-choice dark money groups, and both Google and Facebook have contributed to the anti-choice Federalist Society each year since 2015. Both have likewise contributed to the Heritage Foundation (which has long pushed to overturn Roe), the Cato Institute, and the American Enterprise Institute.

This spending has prompted accusations that the companies are pursuing financial gains at the expense of the values they purport to espouse. For its part, Facebook has a history of profiting from medically inaccurate anti-choice ads and suppressing medically accurate corrections as violative of their community guidelines. In the days following the May 2, 2022 leak of the Dobbs draft opinion, Facebook and YouTube (owned by Alphabet, the parent company of Google), along with TikTok and Twitter saw a spike in abortion-related misinformation that went uncorrected and was widely amplified.  Thus, in practice, neither Google nor Facebook has taken steps to protect or facilitate non-employees’ reproductive choices within hostile states. 

People are not oblivious to the dangers of data mining, or to tech’s duplicity. Google is already facing criticism from employees, who acknowledge that the steps the tech giant has taken are inadequate to protect those most at risk from abortion bans. Indeed, the people who will bear the cost of big tech’s hypocrisy are those who are already disproportionately impacted by the systemic injustices of our society, especially when it comes to abortion. Controlling reproduction has long been a tactic of white supremacy in the U.S. and can be traced through antebellum slavery and Jim Crow-era eugenics to today’s antiabortion policies. It is well-established that people of color are subjected to disproportionate surveillance, arrest, and incarceration. Women of color are disproportionately arrested for negative pregnancy outcomes including miscarriages and stillbirths. This is the case despite the fact that women of color—particularly Black women—are far more likely to experience preventable pregnancy related complications and death due to structural racism, implicit biases, and disparities in the quality of care, according to the CDC. Unless proactive steps are taken to eliminate the structural racial biases in reproductive health care and to prevent the prosecution of patients for pregnancy outcomes, data mining by big tech will only exacerbate race- and class-based injustices. 

Seeking an Abortion in Under Surveillance Capitalism Post-Roe: A Hypothetical 

Imagine Mary, a single mom of two who works as a receptionist earning $26,000 annually (in 2018, the median Facebook employee earned nearly ten times as much, taking home $240,430). Her income puts her family just above the federal poverty line. But Mary is also just over 11 weeks pregnant. And if she chooses to go through with this pregnancy, she and her children will plummet into abject poverty.

Photo taken by Gayatri Malhotra found at https://unsplash.com/photos/yJbWnuL_sF4

In the post-Dobbs landscape, Mary’s options are extremely limited: she lives in Georgia, where abortion is banned at six weeks.    

To understand how draconian such bans are, it’s important to understand “pregnancy math.” Pregnancies are dated to the first day of a pregnant person’s last menstrual period (LMP), up to two weeks before conception and implementation actually occur. The average or “regular” menstrual cycle lasts 28 days, but cycles from 21-40 days are considered normal. By the time a person with a “regular” cycle misses their first period, they are already considered four weeks pregnant. Someone with a reliable 40 day cycle will have no reason to suspect pregnancy until they are two days shy of six weeks along. But cycles are rarely exact, and two days is hardly enough time to confirm a pregnancy and seek an abortion, if desired.

In a best case scenario, Mary would have suspected pregnancy 29 days after her last period and taken a pregnancy test to confirm. This would give her two weeks  to find an available clinic appointment or other means of accessing abortion pills, and to coordinate child care and sick leave. However, very few people have such predictable cycles or track their menstruation so carefully. And, many pregnant people do not produce detectable levels of human chorionic gonadotropin (hCG), the hormone pregnancy tests measure, until five weeks after their last period. Thus, if Mary had taken a pregnancy test on day 29, her chances of getting a false negative were high.

As it stands, Mary—who is modeled on the average abortion seeker in the U.S.—is not experiencing the best case. She gave birth 8 months ago and has been breastfeeding since, so like many new mothers, she hasn’t had a regular menstrual period in nearly a year and a half. In such circumstances, which are by no means anomalous—it’s unsurprising that Mary didn’t learn she was pregnant until Georgia’s six-week window had long since elapsed. In in the days since her positive at-home test, bought using cash from a nearby convenience store, Mary has told no one she’s pregnant. She has not even spoken the words aloud to herself, but Google and the untold number advertisers to whom Google has sold Mary’s data already know.

Unbeknownst to Mary, her Android phone has been tracking her location data, giving Google insight into her normal routine and, today, her visit to the abortion clinic just over the state line in Florida. Even if she had known she was being tracked, there would be nothing she could do: Android’s software makes it impossible for the user to override its tracking capabilities by opting out.

Google uses its search and browsing data to allow advertisers, including anti-choice crisis pregnancy centers who target pregnant people with misinformation to dissuade them from making their own choices about abortion, to guide their ads towards specific people. To advance its mission “to organize the world’s information and make it universally accessible and useful,” Google has been tracking Mary’s frantic searches: 

“late period breastfeeding” 

“5 days late pregnant?” 

“child care for three per year” 

“abortion Georgia”

“abortion pill 11 weeks” 

“cost to give birth without insurance” 

“paid maturity leave laws?” 

“abortion clinic Florida”

After weighing her options, Mary has chosen to seek an abortion in Florida (since she is more than 11 weeks pregnant, she will need a procedural abortion at a clinic). Given its predictive capabilities, Google may have known Mary’s choice before she herself was certain of it. Moving forward, Google’s tracking will prove even more dangerous. Unbeknownst to Mary, her Android phone has been tracking her location data, giving Google insight into her normal routine and, today, her visit to the abortion clinic just over the state line in Florida. Even if she had known she was being tracked, there would be nothing she could do: Android’s software makes it impossible for the user to override its tracking capabilities by opting out. If she does have the abortion and law enforcement in her home state suspect she was ever pregnant, they may be able to use this data to prosecute Mary’s health care providers. 

 Corporate Capture of the Right to Privacy

How is it possible that we have so little privacy? The continued legality of data mining is due to fortuitous timing in technological development and government priorities combined with corporations’ concerted efforts to deregulate industry by pivoting to “self-regulation.” Surveillance capitalism rose to prominence in the post-9/11 era: sacrificing privacy for security was a given. By 2006, it was reported that the U.S.’s intelligence agencies spent as much as $40 billion annually to purchase data mining and analysis technology from the private sector. Through these exchanges, and due to their financial power, tech executives built relationships with politicians and regulators that have allowed them to shape the data privacy sphere for over 20 years. 

Tech executives have tied the origins of data mining—the behavioral value reinvestment cycle, through which early search engines improved the user experience—to the familiar narrative of consumer choice and market freedom. Regulation is unnecessary, they say, because all they want is to provide a better product to the consumer. Of course, this rhetoric is accompanied by fierce lobbying and campaign spending to defeat regulations that would hamper their unlimited access to behavioral surplus from which the true product (predictive capabilities for targeted ads) is generated. 

“I’ve got nothing to hide,” is a common refrain among digital natives who know no other world than the one in which our devices predict our wants, delivering videos and ads that algorithms have already determined we’ll enjoy. But beneath the surface of viral Tweets and inane TikTok videos the sinister and deeply harmful world of data mining. The most privileged among us may indeed have nothing to hide, but most of us have everything to lose. 

From The [F]law:

• Amer Mneimneh, Surveillance Advertising
• Amy Robinson, The Digital Divide is No Accident
• Haley Florsheim, The Salience of Emotional Harms

From The Systemic Justice Journal:

• Sasha Peters, Surveillance Capitalism: How Targeted Ads Solidify the Gender Binary
• Emily Marcus, Pro-Choice or No-Choice?: How Corporations Shape Employees’ Reproductive Decision-Making
• Jack Lubin, Stream a Little Dream: How Netflix Turned Our Culture into Content
• Daniel Zaleznik, Facebook and Genocide: How Facebook contributed to genocide in Myanmar and why it will not be held accountable