Abortion Myths, Part 3
Misinformation, misinformation misinformation, and the problem of non-neutrality in science
In the previous posts (1, 2), I looked at why “correcting for confounds” in abortion science is a bad idea, and how the other usual methods mistakes of the so-called “science crisis” afflict this area, too. Now onto a brief overview of abortion misinformation discourse, which is itself full of misinformation. The over-arching problem here is that perfect neutrality in science on human beings is impossible, and no one wants to accept that.
Both Sides Claim Their Science is the Only Science (This Is Silly)
So why does pro-choice science that makes common methodological mistakes win the day, generally being accepted as the consensus science when it comes to institutional reports, standard abortion care, and policy in many places? One possible explanation is that it doesn’t appear that anyone has previously noted the collider bias problem in the relevant models. Correcting for confounds and calling it a day is still common practice in many areas, even though it’s bad science. The causal revolution is still filtering into scientific practice. Progress takes time.
Another possible explanation is that the information environment is so hyperpolarized that you can’t critique abortion research, or question abortion’s safety or benefit to women’s health, without being called pro-life and shut out of the conversation. This feeds a “spiral of silence,” German political scientist Elisabeth Noelle-Neumann’s term for the process typically ignited by emotionally and morally weighty issues wherein loud public opinions on one side stoke fear of social isolation if one violates apparent norms by expressing different sentiments. One interesting feature of the spiral of silence in the abortion discourse is that it seems to be reinforced by rhetoric of neutral science on both sides, which may contribute to hyperpolarization.
Experts on both sides often talk as if their side’s research is neutral science, and there is no legitimate scientific debate over abortion risks. It’s not clear if this is a rhetorical strategy or a genuine belief. Assuming it’s the latter makes sense; rhetorical strategies can employ lies, and still be effective and difficult to counter, but at least you can reason with people who believe something that’s wrong. So it’s important to counter this sort of claim with the reality that we don’t know what’s cause and effect when it comes to correlates of abortion like substantially increased suicide risk. Instead of juxtaposing pure science with the realm of the unknown and the political, we should have and encourage a recognition that all science in this realm (as in other human realms) is going to contain uncertainty and value-laden choices from start to finish.
Example 1, canonical pro-life book Forbidden Grief: The Unspoken Pain of Abortion, by Theresa Burke, with David C. Reardon
Pro-life researchers have tried to frame their contributions to the literature contesting abortion’s benefits for women’s well-being as being beyond politics. Forbidden Grief, a review of psychotherapist Theresa Burke’s experience as a post-abortion counselor, says on the back cover “It’s time to move beyond the politics of abortion.”
Evidence that abortion may substantially harm women’s mental health deserves our full consideration as scientists. But it is not and can not be beyond or outside the political realm. Science is part of culture; no exit. One might even go so far as to claim that the apolitical is a realm beyond man, and that to claim the ability to see and act in that realm is hubris. I think the intent here, though, is to represent evidence itself as being beyond bias and dishonesty. This misses the limits of neutrality given the man-made nature of evidence and its uses.
Example 2, leading pro-choice expert Julia Steinberg talking to CNN
In a back-and-forth pattern typical of this literature, pro-life abortion expert Priscilla Coleman published a Sept. 2011 Br J Psychiatry meta-analysis showing “Women who had undergone an abortion experienced an 81% increased risk of mental health problems, and nearly 10% of the incidence of mental health problems was shown to be attributable to abortion.” The Royal College of Psychiatrists then in Dec. 2011 published its review of the same literature finding no abortion-mental health risk link. Leading pro-choice expert Julia Steinberg then joined this discourse with a critique of Coleman’s meta-analysis. Steinberg considered and still considers her critique decisive, and recently explained to CNN that many studies showing harm were egregiously bad and there is not really a scientific debate about abortion risks.
Putting aside whether CNN should have let a scientist apparently engaged in a debate be the expert on whether she was right and her opponents were wrong, so there was no debate…
The main substantive argument here, as voiced by Brenda Major — lead author on the 2008 APA report — in the same article, is that you have to control for confounds. But as we’ve seen, controlling for confounds risks introducing more bias than it corrects for (collider-stratification bias). Raw data show substantial links between abortion and risks including suicide; modeling away the links and saying there is no possible causality in them is wrong.
Example 3, Misinformation misinformation
It’s also standard in pro-choice science and abortion care contexts. Possible abortion-suicide effects are massive (at least 2x ballpark increased risk), but most abortion providers don’t tell patients about most possible risks including this one. In fact, many tell them the opposite: That abortion minimizes health risks to women as compared with continuing pregnancy.
When some states have responded by legislating that abortion providers must inform patients about associated risks, some scientists have studied this as an example of misinformation. But their science itself contains misinformation. For example, Table 1 in that Berglas et al Women’s Health Issues 2017 paper lists “Childbirth in the United States is safer than abortion” as a myth; but evidence suggests that childbirth may be considerably safer than abortion in terms of numerous short-term mortality risks (suicide, homicide, and accidental death). Women who give birth in developed countries consistently appear to not die as much as women who abort, and we can’t rule out abortion contributing causally to some of these deaths. Or abortion not causally contributing to them. Probably important to tell women both parts.
This is a huge area of contention. According to the NSF-funded Embryo Project Encyclopedia, “As of 2020, twenty-eight US states have laws that resemble AUL’s [pro-life] Women’s Right to Know Act model legislation.” There’s been plenty of legal wrangling about this. For example, Planned Parenthood argued we don’t know causality in the abortion-suicide link (true), and the appellate court upheld the law because the states retained “ ‘wide discretion to pass legislation in areas where there is medical and scientific uncertainty’ ” (Daniels et al quoting Rounds quoting Gonzales v. Carhart, 550 U.S. 124, 163 S. Ct. 1610 (2007)). The court found that talking about risk doesn’t necessitate proving causation.
That consensus scientists and journalists tend to dismiss the abortion-suicide link as a product of bias reflects a double standard when it comes to evaluating information critically. For example, commentators like British psychiatrist and pseudoscience critic Ben Goldacre and American psychologist and skeptic James C. Coyne have denounced Coleman and Reardon as biased, and challenged Coleman’s declaration of no conflict of interest in her BJP meta-analysis. The evidence for these charges seems to be that Coleman’s read of the evidence differs from theirs. For instance, Coleman allegedly presented a PowerPoint saying “The rapidly accumulating literature on the negative effects of abortion is rarely made available to practitioners and to women considering abortion as professional organizations, including the APA and the AMA, along with the liberal press expend incredible amounts of energy to hide the now scientifically verified truths.” There’s no classic conflict of interest here like there is for abortion providers who publish research and are widely quoted as experts without being denounced as biased in mainstream press.
Standard abortion care really is myth-based. Organizations like the APA and NAS have indeed misinterpreted the evidence. Yet there’s been more press on abortion restrictions and suicides, then on the associations between abortion and deaths including suicides.
It’s as if everyone wants to be neutral because it’s coded morally good, bias is coded morally bad, and the consensus liberal position is coded as neutral. Bias is synonymous with prejudice, even though cognitive bias is something else, and conflict of interest is something different still. Is this a problem of language? A problem of limitation in seeing our own limitations? We’re all biased. Let’s just try to be less stupid and not evil about it.
Bottom line, there is ongoing debate in abortion science about the suicide link (among other things). It’s mostly between pro-choice and pro-life researchers. The evidence contains uncertainty, politics pervade, and we don’t know what’s cause and effect. Sure, science is about figuring that out. But not entirely or perfectly, and not claiming some kind of magical value-neutrality that you really wouldn’t want scientists to have when it comes to the central issue of right to life versus right to liberty, anyway.
In the abortion discourse, everyone is right within their teams, confirmation bias and hyperpolarization abound, and no one stands back and goes: Guys, maybe it’s “turtles all the way down.” Maybe you have to give people information with as little interpretation (spin) as you can put on it at some point, and still admit that there are other possible interpretations, and that even your seeing of the different interpretive possibilities is itself biased, and you’re really sorry that we’re such a pathetic species with such cool, grand visions of omniscience and ominpotence as God and Science, but we don’t really swing that way, so have a cookie and sign on the dotted line.
Abortion may substantially, causally increase your suicide risk. Maybe not. We don’t know. There’s a link and we’re not sure why. Maybe you should think or talk more about some of the factors that might matter. We can give you a list! Do this especially if you’ve been traumatized, you don’t want an abortion, you think it’s unethical, you’re suicidal, and you’re younger or older than most childbearing women. You might think these would be basic things to screen for in abortion care, but they’re not. It would be hard to do that well, anyway.
No one wants to be in the position of responding to politicization with more potentially politically useful confusion, understandably. But that is still, I think, the correct empirical position here — which is also, by definition, the correct ethical position in science. When we don’t know, the correct scientific answer is that we don’t know, and you can tell people that.
Why is it so hard to not lie with statistics here? As Harvard epidemiologist Tyler VanderWeele notes, abortion research tends to be weaponized on both sides, as pro-lifers who believe abortion is murder butt heads with pro-choicers who advocate for and in some cases provide abortion access to women whose autonomy they privilege as a value over whatever value they give the embryo/fetus. This polarization perpetuates an information environment in which pregnant women can’t get neutral information from anyone, because no one is neutral — abortion clinics make money performing abortions, abortion researchers tend to be explicitly pro-life or pro-choice, and even scientific institutions generally respected as neutral have long gotten abortion science wrong, conforming to pro-choice narratives by making demonstrable methods mistakes.
Abortion is a political conflict, a clash of values. Women’s health is collateral damage in this conflict as scientists and doctors try to do the right thing, and misrepresent the evidence in that effort. Women’s healthcare providers are compounding their patients’ losses by (unknowingly) systematically misinforming them.
So why is misinforming pregnant women about abortion risks standard, and why is it likely to stay that way? It’s threatening (financially, politically, psychologically) for abortion providers to recognize that their work may be substantially harmful to some women’s mental health. So they systematically don't see that interpretive possibility. Women seeking abortions themselves might not trust anyone giving them this information, due to the highly politicized issue environment; “abortion is seriously risky” could be discounted as propaganda, even if it’s a truthful possibility. And the research literature reassures practitioners and patients alike (at least, those few who read it) that these substantial possible risk increases in no way suggest causation — even though the evidence does not demonstrate this. Nothing necessarily nefarious here; it’s so common to reject ambiguity, making certainty out of uncertainty for our comfort and convenience.
By another potentially benign mechanism with a terrible output: It’s also way more complex ethically to make a sensible legal regime for abortion if it jeopardizes the lives of a non-negligible number of the women whose rights it ostensibly protects and promotes. So the women’s rights discourse avoids facing these difficult issues, which could be harnessed for potential evil (i.e., usefully destructive conflict) in the culture war.
Mainly the blame for this falls, I would argue, on people with liberal leanings trying to “do the right thing” (which they read as protecting women’s and thus abortion rights) instead of doing good science and calling the culture war crap when it interferes. There are parallels in other cases where do-gooders are willing to throw some people under the bus (or the trolley, as the case may be) to save others — but they haven’t really done the math to know how many people, or whether they for sure net save lives. Global infant feeding policy is another such case. This presents obvious moral problems, especially when these calculations are non-transparent, the do-gooders are disproportionately white and Western, and the people being sacrificed are not — as in the case of exclusive breastfeeding advocates arguing that their preferred infant feeding regime might kill or maim some children, but alternative (unspecified, untested) regimes involving formula supplementation would kill or maim more.
More generally, scientists have to get better at explicitly saying so when our arguments take place on a different plane from how they’re likely to be read (e.g., orthogonal to pro-life/pro-choice). What does that really mean? Is it a hedge, a placeholder, a fantasy? What does it look like?
These are not rhetorical questions. I don’t know the answers. It’s difficult to envision an information source that would or should be credible as neutral to pregnant women considering abortion. It’s easy, in fact, to imagine that patients might dismiss negative information about abortion risk even from an abortion doctor, because the issue is so polarized — there might be a heavy cognitive discount on such information as likely to be biased or false. Mistrust of authority may seem like the conservative cognitive strategy in this context…
Accepting Imperfection
It would be great if we could break science out of the culture war to make women’s lives better. That’s what abortion science on both sides falsely promises that it does. But we’re all stuck in the swamp of imperfect science within meaning-making culture, unable to redo our work perfectly or attain neutrality when we do it. Even when we’re drawing DAGs.
We can’t do perfect science, but we can do better (McElreath). We can not and should not leave our values at the door when we do science, but we can rigorously adhere to advanced methodological standards. Here, as often, I think those standards require accepting uncertainty and being open to the possibility that myth-based medicine endangers rather than improving many lives.
Further Readings
“Transparency and disclosure, neutrality and balance: shared values or just shared words?” Sander Greenland, 2012, J Epidemiol Community Health
“Invited Commentary: A Critical Look at Some Popular Meta-Analytic Methods,” Sander Greenland, American Journal of Epidemiology, Vol. 140, No. 3, 1994, p. 290-6
“…it seems underappreciated that regression curves fit to such [scatter]plots are closer to subjective impressions than to objective data properties. Even when the pattern of study results is not in dispute, there is always considerable room for dispute about the source of the pattern” (p. 296)
“The Need for Critical Appraisal of Expert Witnesses in Epidemiology and Statistics,” Sander Greenland, Wake Forest Law Review, Vol. 39, 2004, No. 2
“The expert failed to note that, logically, ‘being uncertain’ not only means one cannot reject the hypothesis that there is no effect (the null hypothesis); it also means one cannot reject the hypothesis that there is an effect” (p. 297).
According to Greenland, one solution to the problem of non-representative experts is to make references “written and reviewed by a broad panel of experts recruited from different universities, as has been done in consensus-development conferences for special methodologic topics,” citing Working Group on Meta-Analysis in Environmental Epidemiology, 1995 (p. 308).
He continues: “I would recommend that such workshops should emphasize the fallacies in expert testimony that stem from scientific and statistical methods that privilege the null hypothesis; in particular, the workshop should show how such methods induce a false sense of certainty that the null is correct” (p. 309).
“Arguing for a precise conclusion based on indecisive evidence could thus be seen as an attempt to usurp the decision-making role of the court” (p. 309).
And closes with Richard Feynman on physics: “ ‘What we call scientific knowledge is a body of statements of varying degrees of certainty… None is absolutely certain’ ” (p. 310).