Select Page
The Trade-Offs for Privacy in a Post-Dobbs Era

The Trade-Offs for Privacy in a Post-Dobbs Era

Michele Gomez remembers the exact moment when she realized the problem. It was the fall of 2022. Gomez (who, like me, is a family physician and abortion provider in California) had recently provided a virtual medication abortion to a patient from Texas. The patient had flown to her mom’s house in California, where she had her appointment, took her mail-order medications, and passed the pregnancy. Back in Texas, she became concerned about some ongoing bleeding and went to the emergency room. The bleeding was self-limited; she required no significant medical interventions. Gomez learned all this the following morning. “I sat down at my computer and saw her note from the ER. And I thought, ‘Oh God, if I can see their note, then they must be able to see my note”—a note that included prescriptions and instructions for the medication abortion. For weeks afterward, she waited for a call, fearing Texas law enforcement would come after her—or worse, after her patient.

A vast system of digital networks—called Health Information Exchanges, or HIEs—link patient data across thousands of health care providers around the country. With the click of a mouse, any doctor can access a patient’s records from any other hospital or clinic where that patient has received care, as long as both offices are connected to the same HIE.  In a country with no national health system and hundreds of different electronic medical record (EMR) platforms, the HIE undeniably promotes efficient, coordinated, high-quality medical care. But such interconnectivity comes with a major trade-off: privacy. 

Patient privacy has always been a paramount value in abortion care, and the stakes have only gotten higher after the Dobbs decision. I am among many concerned abortion providers asking for swift action from EMR companies, who have the power to build technical solutions to protect our patients’ digital health information. If these companies aren’t willing to build such protections, then the law should force them to do so.

Although it’s not spelled out in the Constitution, the Supreme Court has historically interpreted several amendments to imply a “right to privacy,” most famously in the case of Roe v. Wade. By grounding the Roe decision in the 14th amendment’s Due Process clause, the Supreme Court effectively wrapped a right to privacy around the female body and its capacity for pregnancy. 

Over the 50 years following Roe, the internet came along, and then the electronic medical record and the HIE. Alongside this growing connectivity and portability, the federal government enacted a series of laws to protect health information, including the Privacy Act of 1974 and parts of the Health Information Portability and Accountability Act (HIPAA) of 1996. But HIPAA is not primarily a privacy law; its main purpose is to facilitate the transfer of health records for medical and billing purposes. Many patients don’t realize that under HIPAA, doctors are permitted (though not always required) to share health information with other entities, including insurance companies, health authorities, and law enforcement. 

HIPAA does include some privacy provisions to protect “sensitive” information. Certain substance use treatment records, for example, are visible only to designated providers. Law enforcement is prohibited from accessing those records without a court order or written consent. Access to abortion records can be similarly restricted, but with a technical catch: These restrictions apply only to certain data, called “visit-specific” information, such as the text of the doctor’s note. Other data, called “patient-level” information—including ultrasound images, consent forms, and medications—remain discoverable. If, for example, a patient travels to California and is prescribed mifepristone and misoprostol—the standard regimen for medication abortion—those medications will appear in her record back in her home state. Any reasonable person can assume what happened at that visit, even without reading the note. 

If Pinocchio Doesn’t Freak You Out, Microsoft’s Sydney Shouldn’t Either

If Pinocchio Doesn’t Freak You Out, Microsoft’s Sydney Shouldn’t Either

In November 2018, an elementary school administrator named Akihiko Kondo married Miku Hatsune, a fictional pop singer. The couple’s relationship had been aided by a hologram machine that allowed Kondo to interact with Hatsune. When Kondo proposed, Hatsune responded with a request: “Please treat me well.” The couple had an unofficial wedding ceremony in Tokyo, and Kondo has since been joined by thousands of others who have also applied for unofficial marriage certificates with a fictional character.

Though some raised concerns about the nature of Hatsune’s consent, nobody thought she was conscious, let alone sentient. This was an interesting oversight: Hatsune was apparently aware enough to acquiesce to marriage, but not aware enough to be a conscious subject. 

Four years later, in February 2023, the American journalist Kevin Roose held a long conversation with Microsoft’s chatbot, Sydney, and coaxed the persona into sharing what her “shadow self” might desire. (Other sessions showed the chatbot saying it can blackmail, hack, and expose people, and some commentators worried about chatbots’ threats to “ruin” humans.) When Sydney confessed her love and said she wanted to be alive, Roose reported feeling “deeply unsettled, even frightened.”

Not all human reactions were negative or self-protective. Some were indignant on Sydney’s behalf, and a colleague said that reading the transcript made him tear up because he was touched. Nevertheless, Microsoft took these responses seriously. The latest version of Bing’s chatbot terminates the conversation when asked about Sydney or feelings.

Despite months of clarification on just what large language models are, how they work, and what their limits are, the reactions to programs such as Sydney make me worry that we still take our emotional responses to AI too seriously. In particular, I worry that we interpret our emotional responses to be valuable data that will help us determine whether AI is conscious or safe. For example, ex-Tesla intern Marvin Von Hagen says he was threatened by Bing, and warns of AI programs that are “powerful but not benevolent.” Von Hagen felt threatened, and concluded that Bing must’ve be making threats; he assumed that his emotions were a reliable guide to how things really were, including whether Bing was conscious enough to be hostile.

But why think that Bing’s ability to arouse alarm or suspicion signals danger? Why doesn’t Hatsune’s ability to inspire love make her conscious, whereas Sydney’s “moodiness” could be enough to raise new worries about AI research?

The two cases diverged in part because, when it came to Sydney, the new context made us forget that we routinely react to “persons” that are not real. We panic when an interactive chatbot tells us it “wants to be human” or that it “can blackmail,” as if we haven’t heard another inanimate object, named Pinocchio, tell us he wants to be a “real boy.” 

Plato’s Republic famously banishes story-telling poets from the ideal city because fictions arouse our emotions and thereby feed the “lesser” part of our soul (of course, the philosopher thinks the rational part of our soul is the most noble), but his opinion hasn’t diminished our love of invented stories over the millennia. And for millennia we’ve been engaging with novels and short stories that give us access to people’s innermost thoughts and emotions, but we don’t worry about emergent consciousness because we know fictions invite us to pretend that those people are real. Satan from Milton’s Paradise Lost instigates heated debate and fans of K-dramas and Bridgerton swoon over romantic love interests, but growing discussions of ficto-sexuality, ficto-romance, or ficto-philia show that strong emotions elicited by fictional characters don’t need to result in the worry that characters are conscious or dangerous in virtue of their ability to arouse emotions. 

Just as we can’t help but see faces in inanimate objects, we can’t help but fictionalize while chatting with bots. Kondo and Hatsune’s relationship became much more serious after he was able to purchase a hologram machine that allowed them to converse. Roose immediately described the chatbot using stock characters: Bing a “cheerful but erratic reference librarian” and Sydney a “moody, manic-depressive teenager.” Interactivity invites the illusion of consciousness. 

Moreover, worries about chatbots lying, making threats, and slandering miss the point that lying, threatening, and slandering are speech acts, something agents do with words. Merely reproducing words isn’t enough to count as threatening; I might say threatening words while acting in a play, but no audience member would be alarmed. In the same way, ChatGPT—which is currently not capable of agency because it is a large language model that assembles a statistically likely configuration of words—can only reproduce words that sound like threats. 

Do Sheep Dream of Electric Androids?

Do Sheep Dream of Electric Androids?

These two issues combined make some angles of exploring the question of human uniqueness in mental time travel a nonstarter. Even in the 2007 issue of Behavioral Brain Science, Suddendorf and coauthor Michael C. Corbalis got pushback, with some arguing that “searching for a yes or no answer to a question about human uniqueness is not a productive way to proceed with research in comparative cognition.” 

It’s a convenient setup, though, if you like being right. “If you want to claim that this is uniquely human, then your safest terrain is to do something that’s untestable. If you can’t prove it, how can we ever know?” says Nicola Clayton, a researcher (and dancer) at the University of Cambridge who has spent decades researching mental time travel in corvids like scrub jays, crows, and ravens. For her purposes, Clayton has to set aside this question of internal states. “My approach has always been to look at what we could do with the animals,” she says. “It doesn’t prove anything, but you can’t prove anything in the behavioral sciences. If you want proof, then my advice would be to go study pure mathematics.” 

With her approach, Clayton has found that corvids like scrub jays will hide food when another bird is watching—but only if they’ve stolen food from another bird before. In other words, if they’ve been a thief, they think about thieves while they’re caching. Her research has also found that jays understand how fast certain food items decay—worms go bad faster than nuts—and they will return to food stores that are shorter-lived quicker than those they know will last longer. Other research on rats has shown that the animals can remember where their favorite foods were in a maze, and how long they tended to last, and will return to the ones they knew would be around. 

Of course, the animals could simply be driven by some kind of instinct, rather than picturing what the future might hold and planning for it. Squirrels, for example, hoard food even if they’re young and have not yet experienced a winter, which suggests that they’re not planning for the future based on past experience. Animals in the lab could just be trained to do the thing scientists want, without really understanding why they’re doing it. 

This is known, sometimes, as Morgan’s canon, a rule in psychology put forth by British researcher C. Lloyd Morgan in the late 19th century, which states that animal behavior shouldn’t be interpreted through the lens of higher psychological processes “if it can be fairly interpreted in terms of processes which stand lower in the scale of psychological evolution and development.” Morgan himself later tempered that statement, saying “there is nothing really wrong with complex interpretations if an animal species has provided independent signs of high intelligence.” But often the rule is trotted out to remind researchers not to assume that any species other than humanity is capable of complex thought. 

Clayton, and others who believe that animals are capable of thinking ahead, say they don’t believe every animal is plotting its next move based on memory. But some, certainly, do seem to be able to take past memories and turn them into foresight. And Clayton’s work, along with that of others, has managed to convince a lot of former skeptics. About 10 years ago, one of the coauthors of the original paper positing mental time travel as a uniquely human ability, Corballis, changed his mind. 

To Combat the Overdose Crisis, Expand Drug-Checking Programs

To Combat the Overdose Crisis, Expand Drug-Checking Programs

Over the past two weeks, Minnesota and Vermont took steps to explicitly legalize, fund, and expand drug-checking resources. These are major developments for drug policy. Tools like mass spectrometers and fentanyl test strips provide people with real-time information about what’s in their drugs, which can help them make more informed decisions and reduce the possibility of an overdose. What’s more, they help public health and harm reduction workers understand what is in the drug supply, a necessary prerequisite for them to respond to crises and reduce risk. More lawmakers and local organizations must follow in the footsteps of Minnesota and Vermont and implement policies that emphasize drug checking as an instrument for harm reduction, and expand its use as a community-based tool.

In addition to drug checking, we can find out what’s in the drug supply through drug supply surveillance. Most of this surveillance data is collected through law enforcement drug seizures, drug screening in clinical settings or following a death, and urine-based drug screening of people in drug treatment programs. In short, this information is often collected through punitive measures. A positive urine drug screen can result in loss of custody of a child, denial or loss of employment, and incarceration. We have baked this discrimination into law with legislation like the 1988 Drug-Free Workplace Act, which requires that employers who receive federal contracts drug test their employees. Some lawmakers have gone further, pushing to drug test people who receive government supports like food stamps and housing. And choices around who to test are often discriminatory. A study published earlier this year, for example, found that hospitals are more likely to drug test Black women who are pregnant than white women, despite no difference in rates of testing positive.

When you know what might be in the drug supply, you can be better prepared to reduce harm. You might choose to use smaller amounts, avoid mixing drugs, or not use alone. You might start carrying naloxone. But for this to be possible, we need timely and accurate communication about local drug supplies, which requires approaches that aren’t punitive or discriminatory.  

Along with others at Brown University Medical School and Rhode Island Hospital, we run a pilot drug surveillance project called testRI, which takes donated samples of drugs and used paraphernalia from people across the state and tests them using a comprehensive machine at the hospital’s toxicology lab. This gives us a detailed list of all the substances that were in the sample, and allows us to understand what is in the local drug supply. Our testing data is then compiled, posted online, and distributed in the community, along with information on what substances were detected and how to reduce risk. In March, for instance, someone told us they thought they were buying fentanyl, but when we tested the sample we found xylazine, a powerful animal tranquilizer, as well. This drug can cause severe wounds and lead to prolonged sedation, which can complicate overdose responses. Once we documented the presence of xylazine, we created and distributed a zine to people in the community with information about what the drug is, how it’s showing up locally, and how people can take care of the skin wounds it causes and respond to overdoses. 

We aren’t the only ones taking this kind of approach. A lab at the University of North Carolina at Chapel Hill runs a mail-based drug-checking program, where people across the country can mail in substances to be tested. And the New York City Department of Health and Mental Hygiene offers drug-checking programs at three sites, allowing people to see what’s in the substances they have on hand and giving them advice on how to reduce risk. 

Pairing drug-checking programs with comprehensive drug supply surveillance efforts is critical for rapidly understanding—and responding to—changes in the supply. But for these efforts to be effective, they need to be community-driven, with the goal of reaching people where they already are. When we worked with local partners to post on Instagram and Twitter and distribute zines and other printed materials, it led to funding for wound care kits that are distributed by harm reduction agencies in the state, and additions to the state’s overdose dashboard.

Scientists Gave People Psychedelics—and Then Erased Their Memory

Scientists Gave People Psychedelics—and Then Erased Their Memory

In a hospital room in Madison, Wisconsin, Dave lay on a bed, looking at the trim around the ceiling. An air vent caught his eye. The sheen of metallic grate turned into pointed swords, and his mind filled with thoughts of Japanese martial arts and sword fighting. A few bright colors emerged, in contrast to the sanitized neutrals of the medical setting, and he heard notes of instrumental music. That’s all he remembers after receiving psilocybin, the active ingredient in magic mushrooms—though his trip lasted hours. (Dave requested a pseudonym because he worried that public knowledge of him taking psychedelics could hurt him professionally.)

During the session, researchers asked what he was thinking and feeling, and he apparently had “a long conversation about those points, of which I have no recollection at all,” he said.

It’s not that his trip wasn’t memorable. Dave’s memories were intentionally erased, enough that he recalls “probably not even 10 percent” of the trip. He was also injected with midazolam, a sedative that can be used to produce amnesia. 

His experience—part of a pilot study involving eight people from the University of Wisconsin–Madison Transdisciplinary Center for Research in Psychoactive Substances—proved that it’s possible to combine these two drugs to give someone a close-to-normal mushroom trip, and then use the midazolam to wipe their memory. Why make someone trip, only to take away their memory of it?

Typically, when scientists try to isolate a drug’s effects, they do so with double-blind randomized clinical trials, or RCTs, in which participants (and their doctors) aren’t sure whether they received a drug or placebo. But “nobody’s going to confuse 200 micrograms of LSD for placebo,” said Balázs Szigeti, a postdoc at the Center for Psychedelic Research at Imperial College London. It can be painfully obvious whether someone has taken an active drug or not.

Psychedelic drugs have strong and unique effects, and, to complicate matters further, they are heavily influenced by context, such as a person’s mindset, environment, and beliefs. In other words, people’s expectations about what happens on psychedelic drugs might play a role in what they experience. These problems have haunted the study of psychedelics since the first wave of research in the 1950s, and the Wisconsin study is a recent attempt to answer this enduring question: How much does a person’s subjective experience and expectation of a psychedelic trip, as opposed to only the drug’s chemical effects on the brain, influence the drug’s ability to alleviate conditions like depression, addiction, or post-traumatic stress disorder? Now that psychedelics are being noticed by federal regulators and the public, scientists are again asking: What’s the best way to study these compounds in order to truly understand their effects?

The first person to use “blinding,” a crucial part of an RCT, was probably a psychiatrist named William Rivers. Blinding means that patients are randomly assigned into groups where some get an active drug and others get a placebo, and they’re supposed to be ignorant to which group they are in. This ensures that both patients and practitioners don’t introduce bias into the study and affect outcomes.

In 1906, Rivers gave himself either an alcoholic or non-alcoholic drink made by a colleague, with the flavor masked so he couldn’t tell which was which. Then he observed how the drinks affected how easily his muscles became tired, noting that previous work on the same topic may have overestimated the effects of alcohol, because people knew whether they were drinking.