Note: while this post attempts to highlights some of the limitations I perceive with the “Schrödinger’s rapist” post, I accept most of the central claims made in the analogy (such as don’t rape, that women set their own risk tolerance, that men need to be aware of the first person cognitive evaluation made by women, respecting women and don’t harass them etc.). Therefore, this evaluation should not be understood as providing any support whatsoever for men’s rights activism (MRA) and/or those strongly opposing the “Schrödinger’s rapist” analogy. See “Summary” for a short explanation of what I think of the analogy if you don’t feel like reading through the entire post.
There is one special blog post that has become iconic for social justice vigilantes online. It got over a thousand comments in five days (on a relatively small blog) before the comment section was closed and has probably been linked and shared thousands and thousands of times. I am, of course, talking about the “Schrödinger’s Rapist” analogy, composed by Phaedra Starling on the now defunct blog called Shapely Prose.
Needless to say, the blog post caused a storm of both approval and praise, but also criticism and vitriolic hate. However, as the dust slowly settles, I think it is time to take a close, disinterested look at the advantages and drawbacks of the analogy. As we will see, it correctly captures something fundamental about the experience of women and the general message of the blog post is genuine and reasonable. However, it does not fully appreciate the complexity of the situation and also makes a couple of statistical and factual errors that should be acknowledged.
I find that the strengths of the Schrödinger’s rapist analogy primarily lie in the emotional impact of the text on the reader who genuinely did not understand the inside cognitive view of being a woman in such a social context. I have separated the strengths into four different categories: common sense, conscious-raising, empathy and realism.
Common Sense: the blog post makes a lot of common sense points that every rational person should accept. These include things like don’t rape, don’t grope, don’t threaten with or use physical force, as well as the importance of social context when trying to approach women. If everyone would understand and implement these lessons in their everyday life, the world would be a much better place.
Conscious-raising: reading and understanding the innermost thoughts and feelings of someone is a very conscious-raising experience. Things that you either did not know at all, or suspected but never fully appreciated the scope of, can become abundantly clear in an instant. The “Schrödinger’s rapist” analogy is a very good conscious-raising piece for those that did not understand the specifics of the experience of women in this context.
Empathy: because the analogy portrays the cognitive inside view of how it is sometimes like to be a women and thereby taps into empathy and compassion, the central points makes a much stronger impact than if it merely was a cold reporting of facts.
Realism: the “Schrödinger’s rapist” analogy may very well realistically explain the experience of many women around the world. It does not appear to be written in an exaggerated form just to cause controversy, but to be a genuine piece with an important message. Of course, it does not necessarily represent the view of all women but judging by the support it has garnished so far among various Internet feminists, it does not appear to be a fringe view either.
There are also a few weaknesses with the blog post. These are primarily statistical and factual mistakes.
Quantum quackery: the analogy uses the phase “Schrödinger’s rapist” to indicate that a given women does not know for a fact if the man is a rapist or not. She can only assign subjective probabilities (corresponding to degrees of belief) for the two competing hypotheses. While this is true, it is unfortunate that Starling alluded to quantum mechanics in naming the analogy. Had the women had complete knowledge of the situation (including the man’s intentions), there would probably not have been much doubt regarding whether or not the man would rape her or not. This is not the case for quantum-scale particles. The uncertainty principle does not exist due to imperfect measuring equipment, but is a fundamental aspect of nature. It does not involve subjective degrees of belief.
Two thoughtful counterarguments can be made to this objection: (1) Starling did not use it to intentionally deceive people about quantum mechanics and that (2) Starling should be able to use scientific terms with some poetic license to make the analogy memorable and drive a point home. I find these to be quite reasonable, but it is still important to correct common misconceptions about science.
Life-time prevalence is higher: Starling asserts that the life-time prevalence for rape among women is one in six (~16.7 %). This is wrong. According to The National Intimate Partner and Sexual Violence Survey (NISVS) performed by the CDC, the figure is actually a bit higher at 18.3% (Black et. al, 2011).
There are two ways one could respond to this criticism: (1) the error is small (only 1.6 percentage points) and (2) the NISVS report was published in 2011, which was after Starling wrote the blog post. This is all true, but as before, it is important to provide accurate information, especially when it comes to statistics. It is also not shameful to correct errors on already published blog posts after coming across credible and updated information.
Confusing life-time prevalence with incidence: Starling argues that you can go from the life-time prevalence of rape among women (1 in 6 according to Starling), assume that one rapist rapes 10 victims and get the concentration of rapists at any given time as 1 in 60 (a little over 1.5%). However, this figure is not the “concentration of rapists” (by which I assume Starling means the incidence of active rapists) but rather the life-time prevalence of rapists among men. So, assuming the premises put forward by Starling, 1 in 60 of men will sometime in their life commit rape. But that does not mean that they are, right now, active rapists or have any intention of committing rape.
Now, we should not single out radical feminists as perpetrators of statistical fallacies. In fact, I exposed the a similar statistical fallacy performed by a men’s rights activist (MRA) in my post A Voice for Men on Rape Statistics: Confusing Life-Time Prevalence with Incidence, where the MRA blogger Phil from Utah tried to argue that the life-time prevalence was not particularly high by citing incidence data. Starling makes the reverse fallacy, trying to argue from life-time prevalence to incidence. In the end, life-time prevalence and incidence are two different things.
However, the general method of shifting focus from proportion of rape victims to proportion of rapists is sometimes useful.
Misunderstandings of frequentist probability: Starling goes from claiming that the incidence of active rapists (see point above) is 1 in 60 to saying that this means that there are four rapists in Starling’s graduating class in high school, eleven who train at her gym and one of her coworkers. But this is not how frequentist probability works. If a Punnett square tells you that one green plant and one yellow plant produces 3/4 green offspring, this does not mean that if you make them produce 4 offspring, 3 of them will be green. It means that over a very high number of such attempts, the average green offspring per attempt is 3 out of 4. The average says nothing about the spread. It tells you very little about the proportion in any given attempt. is entirely possible that there are 5 rapists in Starling’s graduating class high school, only 1 that works out at her gym and none of her coworkers. Or any other combination.
While it is possible that Starling used this as a literary tool, it is still worth correcting.
Fallacy of transposed conditionals: this fallacy occurs when one misinterprets the probability of some data on a certain hypothesis P(D|H) with the probability of a certain hypothesis given some data P(H|D). Starling confuses these two when the hypothetical woman in the analogy checks the man, who she fears is a potential rapist for signs. Presumably, what is being modified by the presence or absence of these signs is not the base rate, but P(E|H), where E is evidence and H is the hypothesis that the man is a rapist. That is, does the signals and signs giving off match the idea of a threat (i.e. what would be expected on the hypothesis that the man is a rapist)? However, these two probabilities are not identical, but rather completely different. In fact, the probability of some evidence on some hypothesis is only a very crude predictor of the probability of a hypothesis given some evidence. what is missing is a base rate (which can have a substantial effect on the posterior probability).
Base rate neglect fallacy: most people, including Starling and women who actively apply the Schrödinger’s rapist approach, perform this fallacy by ignoring base rates. Consider the following classic example from the field of statistical cognition (Gigerenzer & Hoffrage, 1995):
1% of women aged 40 who take part in a routine screening have breast cancer. If she has breast cancer, she has an 80% probability to get a positive result. If she doesn’t have breast cancer, there is a 9.6% that she will get a positive test result. Assume that a women aged 40 who took part in a routine screen got a positive test result. What is the probability she has breast cancer?
Most people get it wrong. Even the majority of physicians and the staff at Harvard Medical School get this wrong. Most people would guess that the probability is very high. In fact, it is only a little less than 8%. Why the discrepancy between the estimates and the true value? It is because people are not correctly paying attention to the base rate of 1%. This is the anchor from which the evidence upgrades the probability of having breast cancer.
So even if it is twice times as likely for a man to behave in a certain creepy way given that he is a rapist than if he is not a rapist, this does not mean that there is a very high chance he is a rapist because the prior probability that he is a rapist (assume, for the sake of argument, 1 in 60 as given by Starling) is so low compared with him not being a rapist (presumably 59/60).
This means that it is irrational to just focus on how likely the data is on either of the two hypotheses. One has to take into account base rates.
Stranger danger myth: the analogy puts undue emphasis on the dangers of being raped by strangers approaching women. However, a woman is much more likely to be raped by a man she is in an intimate relationship with, such as a boyfriend or husband. This would mean that the prior probability for such a rape is higher, but I doubt that Starling would advocate that women should be using a Schrödinger’s rapist approach to their boyfriends and husbands. This is because it would be practically and emotionally very difficult to be on a constant, high-level alert in a serious romantic relationship. Therefore, there appears to be a contradiction here.
Confusing actual with perceived safety: although Starling does a good job of separating actual and perceived safety throughout the blog post, there are instances were Starling does not. Starling argues that when a nice guy approach her in public and expects her to trust him, he fails “to respect my reasonable caution, you are being cavalier about my personal safety.” While the former is undoubtedly true, the second part of that statement seems to equivocate the term “safety”. The more adequate phrase should be “about my perception of my personal safety”.
Humans are not completely rational: the “Schrödinger’s rapist” approach for women makes a number of dubious assumptions about human rationality. It assumes, among other things, that:
(1) people can reliably identify rapists by looking at behavior. Can they? If this ability is very unreliable, then the entire exercise may not only be less useful, it may even be harmful if it causes too much anxiety to be worth it. On the other hand, even a slight advantage with using this method may be enough to increase actual safety.
(2) people can reliably perform statistically appropriate Bayesian inference in what can be highly stressful situations. Since people cannot do this very well in calm situations, it seems unlikely that the addition of a lot of stress could improve it. Furthermore, hypervigilance tends to increase the risk of misinterpreting a non-harmful situation as threatening, thereby artificially inflate the risk assessment. Although, with proper training in the area, one may improve reasoning skills in this area.
(3) people have a rational perspective on risk. While anyone is free to have whatever level of risk one is willing to tolerate, but some levels are clearly irrational (yet completely psychologically understandable), such as “any level of risk is unacceptable”. This is because risk is unavoidable. So the question should not be “does this entail a risk?”, but rather “how big is the risk, and how does that risk compare with risks I accept without issue in other areas?”. These areas include eating food, driving a car and taking medication. A possible counterpoint is to bite the bullet and accept that it is irrational, but to argue that the psychological benefit from using an irrational risk cutoff is great enough to merit using it in practice.
The general message of the “Schrödinger’s rapist” analogy seems fairly reasonable: men should not rape, it is women themselves and not men who set their own risk tolerance, that it is very useful for men to understand the first person cognitive evaluation made by women, respecting women and not harass them.
The major strengths of the analogy is that many of the things are pretty common sense, it provides a conscious-raiser for those men who have not yet appreciated what goes on from the first-person perspective of women in such social contexts, it taps into empathy and compassion in a very useful way and may realistically reflect the way many women feel and experience such situations.
Some of the drawbacks with the analogy includes the unfortunate confusion of subjective probabilities with the Heisenberg’s uncertainty principle, an underestimation of life-time prevalence of rape, a couple of statistical fallacies and misunderstandings (such as the fallacy of transposed conditionals, base rate neglect etc.) as well as making at least three problematic assumptions about human rationality. More or less credible counterarguments to several of these drawbacks were provided and discussed. The fact that there is more written space for drawbacks than strengths should not be interpreted as the drawbacks being larger than the strengths, merely that it takes more space to explain statistical fallacies.
Despite these problems, the “Schrödinger’s rapist” analogy, I conclude that it is mostly a useful analogy and has many of its core features intact.
References and further reading
Black, M.C., Basile, K.C., Breiding, M.J., Smith, S.G., Walters, M.L., Merrick, M.T., Chen, J., & Stevens, M.R. (2011). The National Intimate Partner and Sexual Violence Survey (NISVS): 2010 Summary Report. Atlanta, GA: National Center for Injury Prevention and Control, Centers for Disease Control and Prevention.
Gigerenzer, Gerd, & Hoffrage, Ulrich. (1995). How to improve Bayesian reasoning without instruction: Frequency formats. Psychological Review, 102(4), 684-704.