Joshua Mozersky
____
Let me begin with a model. Suppose that there are 1000 men in the country and that 200 of them have committed ‘minor crimes’, things like shoplifting or turnstile jumping. Next, suppose that 20 of the 1000 have committed ‘major crimes’: murder, kidnapping, assault. While not everyone who commits a major crime also commits a minor crime, it is very common: 19 of the 20 major criminals are also minor criminals. Accordingly:
[1] The probability that a man commits a minor crime is 200/1000 = 20%.
[2] The probability that a man has committed a major crime is 20/1000 = 2%
[3] Since, of 20 men who committed a major crime 19 have also committed a minor crime, the probability that a man has committed a minor crime given that he has committed a major crime is 19/20 = 95%.
[4] Since, of 200 men who committed a minor crime, 19 have also committed a major crime, the probability that a man has committed a major crime given that he has committed a minor crime is 19/200 = 9.5%.
From these we can see that the probability that a man commits a major crime given that he commits a minor crime is much higher than the probability of major crime in the general population: 9.5% to 2%, or 4.75 times higher. The inverse is also true. The probability that a man commits a minor crime given that he commits a major crime is much higher than the probability of minor crime in general: 95% to 20%, or 4.75 times higher, again.
Suppose an investigator, A, discovers that B has committed a major crime. It is reasonable for A to conclude that B has likely committed a minor crime: the probability of that is 95%. On the other hand, if A discovers that B has committed a minor crime, it is not reasonable for A to conclude that B has probably committed a major crime. In this case, there is just a 9.5% probability that B has committed such a crime.
Note that A’s second inference remains unjustified even if A points out that the percentage of minor criminals who are also major criminals is much higher than the percentage of major criminals in general. In fact, one would expect A to find that the percentage is 4.75 times higher. It would, nonetheless, remain highly unlikely, just 9.5%, that a minor criminal is also a major criminal.
I will go one step further, for reasons that will become apparent. Suppose that psychologists discover that the disposition that inclines someone to minor crime is the same one that inclines one toward major crime (the differences in rates are, say, attributed to circumstances). This is consistent with the story above because the probability of each kind of criminal activity is higher given the other than either is in the general population. This still doesn’t change the fact that it is improbable that a minor criminal is also a major criminal.
So, in sum, even if:
[I] P and Q raise each other’s probability (i.e. the probability of P given Q is greater than the probability of P, and the probability of Q given P is greater than the probability of Q);
and:
[II] The probability of P given Q is high (perhaps very high);
it does not follow that:
[III] The probability of Q given P is high.
One simply cannot infer [III] from [I] and [II].
I bring all of this up in light of a recent essay by Kate Manne on Tara Reade’s accusation of sexual assault against Joe Biden. I will not comment on the specifics of the case as I am only concerned with the structure of Manne’s reasoning. She writes that people doubt Reade because of “an unwillingness to believe that Biden is ‘the type’” to commit sexual assault even though “[w]e know, alas, that Biden is the type”. How do we know this? Because, Manne states, he “sniffed and kissed” a politician’s hair, six other women have said that he touched and kissed them in ways that made them uncomfortable, and he made jokes about being permitted to hug and touch people onstage. Let us grant that these are bad behaviours and label them “creepy”. Now, one might think that someone could be creepy without committing sexual assault, but according to Manne:
Such sentiments betray a failure to understand that Biden’s demonstrably inappropriate behavior emanates from the same sense of privileged male entitlement that often underlies more serious sexual breaches, including sexual assault of the kind Reade alleges.
Accordingly, she concludes that we ought to believe Reade.
The problem with this line of reasoning is that it is logically identical to the hypothetical case above in which the investigator, upon noticing a minor crime, concludes that there is a high probability that the perpetrator has also committed a major crime, because each raises the probability of the other due to a common type of cause. If we equate creepiness with minor crime and sexual assault with major crime, then even if 95% of sexual assault perpetrators are also creeps, and the same “sense of privileged male entitlement” is causally implicated in both, we still do not know that a majority, or even a significant minority, of creeps are also perpetrators of sexual assault. Being a creep does not entail that the probability that one has committed more serious transgressions is high. This is what is shown by the invalidity of moving from [I] and [II] to [III].†
Of course, it is possible that the situation is different in real life: that most creeps assault even if most assaulters are non-creepy, or even that the two categories are co-extensive. While this is implausible, because serious crime is in fact much less common than minor crime‡, it could be true. However, to believe this we would need to see empirical evidence that it is true, and Manne provides none. My point is not that her conclusion is false just that her argument for it is invalid.
As a matter of fact, however, Manne believes there is empirical evidence against Biden, writing: “Reade’s testimony is evidence that the sexual assault occurred”. This is, to put it mildly, unpersuasive. The mere existence of the assertion that P is not evidence that P. For evidence of this, I submit all the perfectly sincere declarations, some in book form, in favour of astrology, palm reading, healing crystals, guardian angels, past lives, UFO abductions, world Jewish conspiracies, homeopathy, Obama’s foreign birth, Elvis’s ongoing life, the government response to the COVID-19 pandemic, Hitler’s survival, and millions of other examples. People are people, so we sometimes lie, forget things, make mistakes, follow the party line, suffer from delusions, seek attention, carry out vendettas, get swept up in movements, engage in wishful thinking, etc. The mere existence of an uttered sequence of syllables is evidence for the existence of the utterance but, on its own, not much else.
Much of Manne’s essay attempts to convince the reader that Reade’s testimony should not be doubted even though: it has gaps; she waited years to reveal it; she has added to it over time; and there are no other accusations against Biden. Manne is right that all of these could be the case even if Reade is telling the truth, but this demonstrates only that certain considerations fail to be conclusive as counter evidence, not that the existence of an utterance is evidence of its content.
If testimony were evidence, then so much the worse for innocent until proven guilty. Manne tries to soften the blow:
After all, we’re not contemplating convicting this man or taking away his civil liberties. We’re contemplating not believing his story—knowing, moreover, that he has lied many times before—and potentially withholding from him the chance to run for our highest office on this basis. Although this would undoubtedly be a very serious matter, the accusations he is facing are yet more so.
This is neither comforting nor convincing. First, if testimony is evidence, then Biden’s denial of assault is evidence in his favor, so why not believe his story? It cannot be just that he has lied before, because we all have, including Reade. Is it because of the existence of Reade’s counter testimony? That just leaves a stalemate.
If we were, secondly, to break the stalemate by refusing to believe Biden because he is a powerful man, then this would be an instance of “what the philosopher Miranda Fricker calls ‘testimonial injustice,’ wherein someone is not believed because of her social position”, decried by Manne earlier in her essay.
Thirdly, denying someone the opportunity to contend for a job because of the existence of an assertion may not be as serious as imprisonment, but it is serious. Imagine a philosophy department refusing to interview a candidate just because someone she worked with accused her of something; do we really want it to be that easy to sabotage, or be sabotaged by, one’s competitors?
Finally, pace the final sentence in the quotation above, what is serious is what Biden is accused of, not the accusation itself. The latter is not evidence, so it is not even in the same league of seriousness as the former. If it were, then it would be appropriate to consider conviction or the removal of civil rights, which Manne rightly rejects.
While I do not – to overstate the obvious – know whether Biden is guilty, I do know that Manne has provided no reason to believe that he is. I can only hope that we do not lose sight of the wisdom of suspending judgement until the evidence is in; genuine evidence, that is, not just an accusation or invalid reasoning.
Joshua Mozersky is Professor of Philosophy at Queen’s University (Kingston). His primary interests are in the philosophy of science, philosophy of language, and metaphysics. Some of his essays have appeared in Philosophical Studies, Synthese, International Studies in the Philosophy of Science, The Oxford Handbook of Time, and The Blackwell Companion to Time. His book, Time, Language, and Ontology was published in 2015 (Oxford).
† See these statistics, for example, which show that rates of property crime are much higher than rates of violent crime in the U.S.; hardly a surprise.
‡ Another problem: whether creepiness and sexual assault stem from a “single sense of male privilege” is an empirical proposition that requires evidence. A philosopher who claims to know which sense is responsible for which actions in an individual is engaged in a priori psychology.
Comments
75 responses to “Probability, Creeps, and Accusations: On Kate Manne on Tara Reade”
Everybody was on board with Believe All Women until a liberal was accused.
I mostly agree with your post, but I’d love to see someone apply this same level of scrutiny to a conservative man accused. When it’s a conservative, nobody thinks to this level. Everybody just cries “Believe All Women!” and calls you a sexist if you believe in due process.
I’d also love to see some liberals talk about how Biden was anti-integration back when black children started being bused into white schools.
This has nothing to do with the essay. Please keep comments on topic.
If I read you correctly, then I think we agree. It is dispiriting to see people accept a line of reasoning when it amounts to an accusation against someone they dislike, then discount the same reasoning when applied to someone they approve of. In the case of this essay, I wrote it with no intention of either defending *Biden* or accusing *Manne* but simply as an analysis of an argument, though I also hope that it stands of an example of impartial application of good reasoning.
The incident occurred almost 30 years ago, with no witnesses and so there is no way anyone can present evidence either in favor or against Biden. You have to decide who you believe, Biden or Reade. You can also simply not
have an idea which of them is telling the truth. The decision on which one of them to believe is generally going to be based on political criteria, whether you’re a Biden fan or not, whether you’re a committed feminist or not, and the same people who were sure that Brett Kavanaugh had attempted to rape a woman with no more evidence against him than now is presented against Biden are now sure that Tara Reade is lying or so mentally unstable that she cannot be believed.
Kate Manne here plays the role of attorney rather than that of a philosopher. Like a good attorney she finds reasons to support her client’s narrative, although I am not claiming that she is being paid by Tara Reade. I don’t see any reason why philosophers cannot play the role of attorney outside of the classroom.
We are never going to know whether Biden attempted to rape Reade or not. Biden as a human being turns me off immensely and my first reaction to believe anything that is said against him, but when I reason a bit more, I’ll just simply say that I have no conclusion in this case as to who to believe.
They shouldn’t be dishonest, manipulative liars, whether inside or outside the classroom
It’s an interesting argument, but there’s a weakness. It depends on the numbers you use.
As long as we don’t know how prevalent creepy behavior and sexual assault are in political circles, we don’t know how probable sexual assault is, given that the accused previously exhibited creepy behavior.
First of all, there’s no evidence that she is deliberately lying. She just might be reasoning badly. Philosophers, like everyone else, make mistakes when political passions are awakened.
Second, they have as much right to lie as anyone else does. People lie all the time, especially in politics. I agree that in their professional role they should present philosophy honestly to students just as doctors should be honest with patients and plumbers honest with people whose plumbing they are fixing, etc.
I agree: it is entirely fair to play devil’s advocate; no problem. All the same, it seems fair to point out when the advocate runs afoul of the basic principles of inductive logic. The argument I read, in a prominent magazine, is that we should believe someone did something very seriously wrong because he is kind of creepy and someone else said he did it. I guess you could say that I write in the insistence that standards of legal or any other evidence never reach a point where that would stand a chance.
This really isn’t that difficult. Reade changed her story multiple times and has multiple folks with stories that add up to her being a grifter. There are no other accusations of that nature with Biden. Her Medium article on Putin is best described as deranged.
Ford’s story is believable because it fits with the life Kavanaugh led in HS and university. There are character questions concerning his public life after university. Ford on the other hand seems to have led an admirable life and gave creditable testimony.
One recalls the old legal maxim: When the law is on your side pound the law. When the facts on on your side pound the facts. When neither the law nor the facts are on your side, pound the table. All I saw was K pounding the table followed by an inconvenient question at which point the person asking the questions disappeared and Graham was pounding the table.
Not liking somebody should lead one to more closely mark matters to market when out of character accusations appear. This is relevant because guys who do what Reade accused him of doing tend to be serial offenders.
jyvurentropy accuses the left of hypocrisy in the Biden case, and the reasons are clear. The article adds helpful clarity to the “believe all women” debate through the use of probability.
I’m curious whether a purely practical approach to thinking about the narrow case of politicians accused of sexual assault could be developed with the aid of the probability framework.
So say you are a devoted member of the left, and you’re faced with the Biden accusation. Could you take that 9.5% probability that he’s committed assault, given the minor “creepiness” of the man, and figure that in to your thinking? The simple answer is, “Yes. Obviously. That’s what we do with most of our political thinking. Weigh possibilities of one proposal working vs another, the probable actions of a politician given past decisions, etc.”
But perhaps this could be a de-escalating move during political deliberations. The high emotions in these cases come from the ethical implication of either actual assault or defamation. Maybe, as a kind of rhetorical move, these probabilistic considerations could be introduced.
I know it’s far-fetched as a real-world way to talk about in-the-moment allegations. But why not actually lay out the case? “I’m a left-of-center voter. I’ve got maybe a 78% confidence level in Biden implementing policies I approve of. This 9.5% probability knocks me down to 70% maybe a bit lower.”
I’m just thinking on the fly here, and it may seem more like early stages brainstorming for coding a (very crude) AI voter. But it could be a kind of response to jyvurentropy’s accusation of hypocrisy. I know in a purely ethical way there’s a clear hypocrisy in believing the woman against the conservative but not against the progressive. But could the specific charge of hypocrisy be addressed by admitting some asymmetry but then “quantifying” a reason for balancing the 9.5% likelihood against the much higher probability of the politician “being the right one”?
“Everybody was on board with Believe All Women until a liberal was accused.”
Al Franken.
This is a little unfair to Manne seeing that she backed off the claims in this article, but there is something going on here that is worth looking at. As I see it, Manne’s mistake was in making explicit a very popular kind of bad reasoning that afflicts many who want to use one big idea to explain complex and varied social phenomena. In Manne’s case its patriarchy or male privilege, but you see the same thing in every corner of the political world—institutional racism, ‘cultural marxism’, religious belief, atheism, fascism. All sorts of things tempt people to make this move.
It goes like this. You want X to explain a huge chunk of the world’s badness. But, frustratingly, it either only plausibly explains things that are prevalent enough but not terribly bad in the grand scheme of things, or it explains something truly bad but comparatively rare. To make things interesting then, it’s very tempting to take the badness of the one and tie to the prevalence of the other by declaring them two points on a continuum. Suddenly Biden’s not all that far from a rapist, liberals might as well be communists, and Baptists are practically the Taliban. It works great until you try to spell the logic out.
Once again, I disagree entirely. And I will continue to oppose sleaze and smear merchants.
Unfair to Manne. Lol. That’s the funniest thing I’ve heard this week.
Agreed but, as I mention in the article, Manne doesn’t know these things either, or at least she doesn’t present them, so her argument is run independently of such considerations. Under such conditions, belief should be suspended.
Since none of us have any evidence as to whether Manne is deliberately lying or just reasoning badly, you’re smearing her when you label her as a “sleaze and smear merchant”.
I won’t quibble with my level of fairness to Manne, though I certainly made the effort to be as just as I could be. You have an interesting diagnosis of a kind of bad reasoning that seems worth thinking about, and we seem to agree that a good way to inoculate oneself against such problems is to keep a firm eye on the underlying logic.
I dislike hypocrisy as much as the next guy, but what really irks my philosophical sensibilities is the double standard of picking and choosing between logically analogous arguments based on whether one likes or dislikes the conclusion. So, my argument above would and must apply just as much to a similar accusation against Donald Trump, Bill O’Reilly, or anyone else. Everyone stands to gain by mastering as much probability and inductive logic as possible (paraphrasing Ludwig Boltzmann: probability is the logic of nature), so I applaud your attempt to apply these principles to political and other domains. There are famous pitfalls, inductive mistakes that are extraordinarily tempting, so one needs to think things through quite carefully. Conditional probabilities often mess people up, which is part of what I think is going on in Manne’s essay.
Thank you so much for this. Very clear, accessible analysis of her argument.
My evidence is all of her behavior on social media and the fact that she does this sort of smearing as a matter of routine.
So, yes, she is a sleaze and smear merchant.
For the record, I think that what was done with Kavanaugh was outrageous too.
I haven’t followed her career in detail, but all of us have a tendency to attribute nefarious motives to those we politically disagree with and to see those we agree with as making human all-too human mistakes and you may be doing that in this case.
Nah. I’ve tried to engage with her in good faith many times. She is part of the Woke Grifter set that currently is doing so much damage to professional philosophy.
Doesn’t it make sense to say that testimony *does* (or can) count evidence, but that it’s not necessarily very strong evidence? As in, if dozens of women testify that a particular man assaulted them (e.g. Weinstein), then that looks like much stronger evidence that assaults occurred.
You seem to be claiming that people who have led respectable lives like Ford should be trusted and people who have led unstable or precarious lives like Reade shouldn’t be trusted. I’ve known enough respectable liars and enough precarious truth-tellers to not accept that entirely.
I agree with you that not liking someone should lead one to closely examine oneself when character accusations appear and that it is why I made it clear above that although I dislike Biden, I will not accuse him of sexual abuse in this case.
I can imagine contexts in which it would make me less confident in the accusation.
I think both are a disgrace, honestly.
That’s an important point. The utterance that P can count as evidence that P in certain circumstances, but in others it could count against P. What might be going on in Manne’s essay is the assumption that the right context holds for Reade but no case is made for that, and the overall point remains: the mere existence of the utterance/accusation is not evidence. If you imagine a scenario in which you know someone well, trust him/her, have a good read on his/her character, know something about the situation that prompted the utterance, etc., then you may be inclined to take the assertion that P as evidence that P. It would be an error to suppose that absent all that context, or in a different one, it would still count as such.
S.W: If I know someone is a serial liar and they have defrauded and exploited multiple folks then I wouldn’t consider them respectable under any circumstances. There was enough evidence, not including Ford’s character, and not limited to Ford’s testimony, to do an actual investigation. A single, unsubstantiated accusation isn’t a “case.” Your mileage seems to vary.
Dan: You use the charges you have, e.g. Capone and taxes. K is a bad actor with suspicious financial issues in the present. I doubt he could have withstood a thorough (as opposed to the limited ones actually done) FBI background check. As that was under the Trump Administration’s control options were limited. I would have preferred a real vetting but that wasn’t going to happen. I heard enough to convince me that K has a lot to hide.
Please tell us that you didn’t fall for his blubbering and posturing “defense.”
I don’t think it should have even received a hearing of any sort. Same with this case.
I think most of us recognize that the use of the Tara Reade legend by certain leftists was not really an attempt to further #MeToo feminism, but a strategy to somehow cancel the primary process and pressure the DNC to ‘get back to Bernie,’ or even Elizabeth Warren. But there was Ms. Reade and what she said, and there were a handful of friends and family members who reported ‘well, this was what she said,’ and that was all the evidence brought forth. As Mozersky notes, “This is, to put it mildly, unpersuasive. The mere existence of the assertion that P is not evidence that P.” If Reade had made her claim public at the time of the alleged occurrence, or, really more to the point, had she initiated a suit at the time, or even initiated criminal investigation for assault, the matter might be different. Instead, the accusation seems to be a complement to her resume. Her public presentations since then have also been a rather confusing waffling – from praise for Vladimir Putin, to praise for – Joe Biden. This waffling really raises questions of credibility that the #MeToo Bernie (‘better, Elizabeth!’) leftists have trenchantly avoided. What they really want to say is that the US is irredeemably misogynistic, and unless we submit to the feminist revolution, we might as well go to hell and enjoy the ride – thus their willingness to lose to Donald Trump. ‘It won’t get better until it gets worse!’ Well, the problem with that model of course is that it can *always* get worse. And even when things look like they’re getting better, say, in a real revolution, they can still get worse. Emma Goldman went to the new Soviet Union under Lenin, thinking that a new utopia with a unified left was in the offing. She found oppressed women, gays under arrest, the beginnings of a secret police, and anarchists – her own leftist team – slaughtered in the Ukraine. She got out of there; and the next act was Stalin’s. She didn’t lose her political beliefs; but like many on the left at the time, she gave up her hopes for any utopia.
It can always get worse. Even a little better – and maybe Biden is just that, a little better – is preferable to the continuing worsening that is Trump on a daily basis. I don’t want a revolution, not even Bernie’s ‘revolution.’ I just want us to be able to discuss policy again, and not need to comment on an orange faced baboon’s disgusting tweets and anti-American “dominance” anymore.
Apologies, I wasn’t very clear.
My reaction wasn’t about Biden or Manne, but about the general form this type of statistical argument has. Its persuasiveness depends – to my taste – too much on the numbers used.
“The mere existence of the assertion that P is not evidence that P”:
https://www.nsvrc.org/sites/default/files/publications/2018-10/Lisak-False-Reports-Moving-beyond.pdf
“…when more methodologically rigorous research has been conducted, estimates for the percentage of false reports [of sexual assault] begin to converge around 2-8%.”
I just want to register the fact that this conversation is taking place in apparent isolation from the large philosophical literature on this question. But this isn’t something you guys need to theorize from the ground up!
“…is testimony an autonomous source of epistemic authority? Reductionists answer negatively. They are opposed by anti-reductionists who hold, characteristically on a priori grounds, that testimony is a source of warrant in itself, not reducible to warrant derived from these other sources, even if empirically dependent on them.” –SEP, “Epistemological Problems of Testimony”
So Manne is just echoing the anti-reductionist view, which has lots of defenders. Her claim about evidence is not obviously wrong or absurd; the view is that the evidence is defeasible, i.e. can be defeated by situational factors, but if no such factors exist in a case then testimony can be a reason to believe all by itself.
That Tara Reade’s accusations have been utilized by the left against Biden says nothing about whether they are true or false. As I stated above, I don’t believe nor do I disbelieve her.
That Reade waffles indicates that she is an unstable and psychologically precarious person. I would imagine that sexual predators look for unstable and precarious people to prey on, so that does not indicate that she is necessarily lying about Biden abusing her nor do her stupid comments about Putin. Reade is a loser, but losers tell the truth as often as winners do.
For the record, I dislike Biden and my dislike for him is more than just political. In fact, I don’t necessarily dislike people with whom I disagree with politically. Biden is the kind of person I’ve disliked since high school. However, he won the primaries, he beats Trump in the polls and he does seem like the best bet in the real world to beat Trump, which is, I believe we agree, the main political goal at present.
It’s really simple to me. You don’t go out in public and accuse someone of something as serious as sexual assault, when you have no evidence. That someone said so is not evidence. That’s why “say so’s” are interrogated in courts. And that someone “is the type” is no evidence either.
It requires neither philosophy nor probability theory to grasp this. Only a sense of basic decency and maybe a dash of personal humility.
See my reply above.
To clarify a bit more: your argument looks convincing because the number of people who commit minor crimes is much higher than the number of people who commit serious crimes.
It would be far less convincing if these numbers where close to each other (it could be argued that Manne implicitly uses this hypothesis).
In the absence of reliable, empirical data on the behavior in political circles, I find both Manne’s argument and your argument that she’s wrong, equally unconvincing. The only thing you show is that one needs to be careful with assumptions in the absence of data. But I don’t think this comes as a surprise, and it’s equally true for your and Manne’s argument.
Statistics may be the logic of nature, but as Ludwig Boltzmann knew very well, you need data on nature before you can start to do statistics. Without data, it’s often “garbage in, garbage out.”
Don’t get me wrong: intuitively I think that in this case, serious crimes indeed are much less frequent than minor crimes. But I have solid proof that my intuition often is wrong.
Since you closed comments on the previous post about the protests and covid 19 among other topics, I’m going to send this link here. It’s an article by Glenn Greenwald from the Intercept about how public health criteria suddenly changed when anti-racist and leftwing protesters took to the streets without social distancing. What was previously considered irresponsible and dangerous to public health, getting close to others in the streets, is now seen as the epitome of civic responsibility, points out Greenwald.
I know that others have said the same thing, but Greenwald is a noted figure from the radical left and so I thought that his intellectual honesty and moral courage, his willingness to challenge radical left common sense, is worthy of note and of praise.
https://theintercept.com/2020/06/11/the-abrupt-radical-reversal-in-how-public-health-experts-now-speak-about-the-coronavirus-and-mass-gatherings/?utm_medium=email&utm_source=The%20Intercept%20Newsletter
Okay, but Manne’s argument remains invalid no matte what numbers you plug in, so it plays no role in establishing its conclusion, even if the conclusion is true. For example, “All dogs are mammals; Spot is a mammal; therefore Spot is a dog”. Even if the conclusion is true, the argument doesn’t establish it.
I have reasons for my stance here (though, to be fair, I don’t present them in the foregoing). Burge’s anti-reductionism, as well as Davidson’s principle of charity with which it is compared in the SEP, both depend on (versions of) the principle of sufficient reason: as Burge writes, “reason is a guide to truth”. I would point out that – as Kant, Leibniz, Descartes, Plato, Aristotle, Putnam, Nietzsche, Wittgenstein, and others all saw – one needs a metaphysical system of a rather particular character to support this principle, and contemporary science by and large rules out the possible candidates (I am just stating a view here, but I do have arguments). Further, I think enough evidence has come in from cognitive science to cast doubt on a default stance of self-trust (because we don’t seem to be self-transparent in the right way). Finally, I don’t see good reasons for the “assurance” view. In sum, I don’t find anti-reductionism persuasive. Now, this is all presented without the arguments, so it may look like I am just “echoing the reductionist view”, but then again, isn’t that what Manne is doing, for the other side, so if it’s good enough for her…?
More seriously, it is true that there is substantive philosophy to address here (another example: I have doubts about the way ‘content’ is employed by many participants in this debate), so I do appreciate your point, though no matter where we fall on testimony, the article contains an invalid probabilistic argument, so at best it amounts to: there is an accusation against Biden, so we should believe he is guilty of something serious because anti-reductionism in testimony is true. I do not find that convincing.
Agreed. I am dismayed by the extent to which humanities professors and students tend to be sanguine about revolution and the threat of subsequent totalitarianism, as though they’ve never been taught about Robespierre, Stalin, Pol Pot, etc. It is odd.
An accusation in conjunction with all of this research backing it up suggests one thing. Then comes evidence that Reade fabricated her story, and something else suggests itself. Then, imagine, someone presents emails indicating that the evidence of fabrication was itself planted by friends of Biden. So back and forth it goes, all around the same accusation. This suggests to me that a statement depends on context for its evidential status, as Dan pointed out earlier.
Dear Dan. The OP has it that “the structure of Manne’s reasoning” is the topic. I am saying that Bayesian reasoning takes a prior, and I present one reasonable prior. Given Manne’s publication record, she would be aware of the academic literature on sexual assault prosecution rates. Her article is called “I believe Tara Reade…”, not “I believe a case against Biden would be successful in court”.
Yeah, my point stands. She has no good reason to think Biden did anything. The accusation against Biden is just an expression of the views she already held about men and women. And revenge for Warren. Her previous antics involved asserting that Warren’s loss was due to misogyny.
This might be a bit boring for some readers, but Mozersky’s example is as follows, when expressed as a diagnostic test,
see for example here
Major+ Major-
Minor+ 20 180
Minor- 0 800
Likelihood Ratio for Major+, given Minor+ = (20/20) / (180/980) = 5.44
Likelihood Ratio for Major+, given Minor- = (0/20) / (800/980) = 0
By Bayes Theorem, posterior odds = (test likelihood ratio) x prior odds
If prior odds for person X being a major criminal are 50:50 (indifference),
then posterior odds are 5.4:1 = 84% chance of guilt
If prior odds for person X being a major criminal being 90:10, as per the reference,
then posterior odds are 49:1 = 98% chance of guilt
Obviously, to decrease the utility of this test, we need more major criminals who have never committed a minor offence.
Depending on the numbers you plug in, Manne’s argument can be quite persuasive.
Take the following assumptions:
– 200 = number of people who commit minor crimes;
– 180 = number of people who commit serious crimes;
– Of the people who commit serious crimes, 170 commit minor crimes (about 95 %).
In other words: among the 200 people who commit minor crimes, there are 170 who commit serious crimes. That’s an 85 % probability.
In the absence of reliable empirical data, this argument is just as good as yours.
I think Dan nails it when he writes that she starts from assumptions she already holds.
But the statistical argument of prof. Mozersky also starts from certain assumptions that – as far as I can see – aren’t based on reliable empirical data, valid for this particular situation. As your example implicitly shows, the output of a statistical model can be anything you want, as long as you use the right numbers as input.
Statistics is famous for being worse than “damn lies” when it works with data. In the absence of data it’s even worse than that. Just leave statistics out of it when you don’t have a validated statistical model.
I thought the statistical argument was interesting, but it’s not the one I would make. My objection to what Manne did — and does — is expressible in ordinary language, in about a paragraph.
Hi David, if I were to put my model in Bayesian form, it would be as follows (where MA = major crime, MI = minor crime):
Pr(MA/MI) = Pr(MI/MA) x Pr(MA)/[Pr(MA) x Pr(MI/MA) + Pr(not-MA) x Pr(MI/not-MA)].
The prior is Pr(MA) = 0.02. The likelihoods are:
Pr(MI/MA) = 0.95;
Pr(MI/not-MA) = approx. 0.185 (181/980).
Plugging these in, I get Pr(MA/MI) = 0.095.
So this is, I argue, a model in which Manne’s premises (Biden is a minor transgressor and a common disposition contributes to minor and major crime) are true but her conclusion (we should believe Biden is guilty) is false, assuming 9.5% probability of truth is not sufficient for belief. So I guess I am not sure where we disagree.
“My objection to what Manne did — and does — is expressible in ordinary language, in about a paragraph.”
I entirely agree.
What annoyed me from the beginning in this part of the argument of prof. Mozersky, is the appeal to “mathematical authority”. As a physicist, I’m quite sensitive to the misuse of mathematical authority.
Seems to me like a harmless and interesting way of looking at the issue.
For you, it may be harmless, but many people think “it’s mathematics, it must be right”.
My argument is as follows: Manne’s argument has the form P; Q; therefore R, but here is a model in which (propositions of the same logical form as) P and Q are true but (a proposition with the same logical form as) R is false; therefore, her argument is invalid. I made no claims about the plausibility of my model as a representation of reality, just its logical consistency. Nor did I claim that Manne’s conclusion is in fact false. Now, I obviously made assumptions because that is what building a counter-model requires but I maintain that if an argument has a model in which its premises are true and its conclusion false, then it should be rejected. The fact that one can *change* the form of the premises and thereby end up with a better argument isn’t relevant because that can be done for any invalid argument.
While “radical left” is in the eye of the beholder I followed the link and found it interesting. Recently there was a case in Springfield, Missouri in which a hairdresser with Covid19 served well over a hundred clients. All wore masks and none were infected. Note the pic with the article. All seem to be wearing masks. Glenn the “radical leftist” seems to never miss a chance to miss the point. His take is what one would expect from a radical libertarian.
What will be interesting will be comparing rates between jurisdiction where the police used tear gas and violence against those where the police remained civil.
In my view, you argued that conclusions depend on premises. But we already knew that, didn’t we?
If we assume that the model is correct, the question obviously is: how good are the premises?
We don’t know.
If, in this part of your argument, you really wanted to show that Manne is wrong, you should have shown that her (implicit) premises are wrong. But you don’t. You just pick a few numbers, and calculate that Manne’s argument doesn’t work IF your numbers are a valid description of reality. But what IF your numbers aren’t a valid description of reality? Your argument becomes quite a bit less convincing.
Anyhow, I don’t understand what you mean when you write “I maintain that if an argument has a model in which its premises are true and its conclusion false, then it should be rejected.”
If there’s a solid argument you’ll die with 99,9 % probability when climbing K2 – it’s hard to believe you’ll try to climb it. There is a model in which the premise is true and you survive, but few people will bet on it. Do you reject the danger of the climb because there’s a 1/1000 probability you’ll survive; in other words because there’s not a logically solid relation between climbing and dying?
My view: statistics is NOT logic, and NEVER do statistics without solid, reliable empirical data.
If you’re interested, there’s a link to a podcast in the article and there Glenn talks about the effects of tear gas on contagion from Covid and then dialogues with a medical expert who is in favor of the protests (as is Glenn) and who also goes into the effects of tear gas and of jailing protesters on infection from Covid. They also discuss the possibility of a double standard among some medical experts: who condemn protests by white nationalists as being unsafe during a pandemic and who support protesters by Black Lives Matters as somehow being safe during the same pandemic
In any case, I’ve seen many pictures and videos of protesters without masks and who are not keeping the proper social distance.
I’ll put this down the bottom:
couvent2104 writes “statistics is NOT logic, and NEVER do statistics without solid, reliable empirical data”. This is completely wrong, Statistics is about rigorously reasoning under uncertainty. It is well known that physicists prefer to generate sufficient data that they don’t need to use statistical methods at all 😉 The “Bayesian Revolution” in statistics has been driven by availability of adequate computing power to model correctly when there are lots of unknowns with vague prior information. There are lots of philosophical arguments that it is irrational not to use Bayesian reasoning in all settings. The logics involved are non-classical eg Pearl’s Bayesian causal models can be partially expressed as McCain and Turner’s causal calculus. Obviously, there are many situations where the correct model says “can’t know one way or the other”. I should point out I’m a frequentist, but will use objective/empirical Bayesian methods or very vague priors.
Josh Mozersky points out his prior is “Pr(MA) = 0.02”. If one multiplies these odds (1:49 against) by 5.4, one does get 9.9%. In Manne’s setting, we are doing a series of tests, each of which has a particular sensitivity and specificity (or equivalently test+ and test- likelihood ratios). Each test increases or decreases one’s certainty given one’s initial prior in a chain-wise fashion. So, starting with the prior probability that a man taken at random from the population has, say, a 1% of being a rapist (this is probably a little low), we first assess the test characteristics of an accusation by estimating the four probabilities in the two-by-two table, including how many false-negatives there are, ie accusations that are never made after an actual assault.
So, US estimates of cumulative risk of sexual assault for white women 10-20%, and of those 15% are reported to police. There are roughly 10% false accusations. Of 1000 women,
Assaulted Not-assaulted
Accuse 22 3
No accuse 128 847
LR+ = 41
This gives us our first posterior probability of 30%, . We then use this probability as a prior for the second test “that creepy overpersonal men are (slightly) more likely to assault women” [1], and generate a new posterior probability. And so on.We can put formal confidence intervals around these numbers, Most people are doing these kind of calculations using heuristics, eg her ex-husband mentioned she had told told him about sexual harrasment at work, “yeah that increases my certainty a bit”, but “she likes Putin, so I’ll increase the probability of a false-positive” etc etc.
[1] No, I’m not going to do this one. However I think Manne is allowed to make this move, arguing just that she thinks the LR+ > 1. An LR of 2 is a pretty weak test, but it will still bump up probability of guilt from 30% to 45%.
reply to couvent2014
Bayes estimates 9% for the original numbers and 50% for couvent2104, mostly because 180 out of 1000 increases the prior while the likelihood, as noted, is similar.
“couvent2104 writes “statistics is NOT logic, and NEVER do statistics without solid, reliable empirical data”. This is completely wrong.”
No, it’s not completely wrong. “Never do statistics without solid, reliable empirical data” actually is completely right if you want to say something sensible about reality (and you know that very well).
With that quote “statistics is NOT logic” I obviously pointed out that probabilities are not certainties. Equally obviously, it’s your privilege to misunderstand.
Sorry, my previous reaction was a bit too harsh.
To say something sensible about reality, statistics need good data, but I agree that in this particular case it’s more – much more – about beliefs, priors and dubious data than about reality.
“You don’t go out in public and accuse someone of something as serious as sexual assault, when you have no evidence”
What do you mean by “evidence”? Tara Read has given an account of an event, allegedly observed by Tara Read herself, which she asserts occurred while she was in her sober, waking state, which she says she now remembers as it happened. That is evidence under any common sense denotation of the word. Taking into consideration all that can be said to call into question her credibility and her reliability the result may be that this evidence is assessed by you as having very little weight, but I can tell you from personal experience (here follows some evidence of the state of affairs I shall describe) that in a court of law the testimony under oath of one witness is some evidence, and may be sufficient evidence, for conviction on a standard of proof of beyond reasonable doubt if the testimony is adjudged to be true. That is so even if there is no corroborating evidence whatsoever. Again, that testimony is evidence of what the witness asserts, even if the Jury finds it to be of insufficient weight to warrant conviction. Now, Tara Read has not, so far as I know, testified under oath; so perhaps you will say that her utterances on this topic are for that reason not “evidence”. But that would be special pleading, I submit.
Nah, anyone can accuse anyone of anything. The accusation in itself is not evidence.
I don’t dispute your data, but that is not the argument I encountered in the essay in The Nation and to which I am replying, so I don’t see any need to dig in here. I think what I wrote is compatible with what you say.
I agree with you on climbing K2, but I interpret you as providing an argument based on cited statistics and considerations like the limits of the human body. There wasn’t anything like that in Manne’s article. Perhaps you, and others in the comments, believe that such data are implied in her argument and that I should have given her credit for that, but I guess I can only retreat to the line professors give students, “I am only able evaluate what she wrote, not what she meant to write”. What I read wasn’t a presentation of statistics but, instead, mention of a correlation between serious criminality and a sense of entitlement, on the one hand, and between creepiness and the same sense of entitlement, on the other; then the conclusion that an observation that someone engaged creepiness gives us sufficient reason to believe he also engaged in serious crime. To me, that’s like arguing that because those who climb K2 are daredevils and those who race NASCAR are also daredevils, then if you learn someone is a NASCAR driver you can be confident they also climb K2. Even if most NASCAR drivers do in fact climb K2, that argument doesn’t establish it, in my opinion at any rate. Anyhow, I would love to discuss the conceptual foundations of probability theory with a physicist, but this probably isn’t the place. I appreciate your engagement.
Thanks for the kind words!
Dear Josh. Sorry, I’m afraid you’re conflating couvent2104 and me in this thread. I’m not that interested in pursuing this further, but I did look at Chapters 6 (“Exonerating Men”) and 7 (“Suspecting Victims”) of Manne’s Down Girl: The Logic of Misogyny (DG) where she discusses the fact that
So, I read the informal argument in her essay as:
1) The accusation has some degree of plausibility (paras 1-3)
2) Why would one discount the accusation? a) “an unwillingness to believe that Biden is ‘the type’” and b) “sheer political inconvenience” (para 4)
3) Block first objection via two arguments a) “Biden [actually] is the type” eg Biden has a record of low level behaviours that are actual “sexual misconduct” (paras 5 and 6) b) and anyway “many women’s monsters can seem like nice guys to the rest of us” (para 7), a thesis expanded in Ch 6 of DG.
4) Block second objection as morally unworthy (paras 9 and 10).
therefore:
5) “such conclusions [that an accusation is very likely to be false] do a profound injustice to women, amounting to what the philosopher Miranda Fricker calls ‘testimonial injustice,’ [Ch 6 of DF] …Reade’s testimony is evidence that the sexual assault occurred, though there remains room to disagree on its strength or probative value.”
Manne does not explicitly mention objection 2c) that Reade is exactly the type to make a false allegation ie a low status “woman in a historically patriarchal society, in which powerful and privileged men have long been deemed more credible”. As a card carrying feminist, she doesn’t have to.
Cheers, David,
Actually, I’ll add the opening para of “Suspecting Victims”, since it’s so apposite
Prof. Mozersky,
I think Dan’s head is exploding from frustration by now, but one last reply.
No need for a discussion about the foundations of probability.
I made a perfectly banal and totally uncontroversial observation:
if you plug in the right numbers in your model, you can get any result you want, any probability from 0 to 1 that a person has committed a serious crime, given he has committed a minor crime.
Let’s call that probability P(S/M).
If you wanted to point out that probability is not certainty (unless P(S/M) = 0 or 1) – well, again, that’s perfectly banal. In this discussion, I implicitly have given Manne the benefit of doubt, thinking even she knows that (*).
If you wanted to point out that under certain assumptions P(S/M) is close to 0 – well, yes. Just plug in the right numbers. I can take your model, plug in other numbers and the result will be that P(S/M) is close to 1.
So what does your statistical argument show? That the results it produces depend on the input. But where did you get your numbers, if I may ask? Do you have reliable empirical data on inappropriate behavior that is valid in this situation, i.e. in political circles? As far as I can see, you don’t.
And neither does Manne. In a certain sense, you’re both guilty of the same thing: starting from assumptions that produce the desired result: in your case, a low P(S/M), in her case a high P(S/M).
In itself that doesn’t bother me too much. What does bother me though, is the fact that you’re framing it as a mathematical argument. I always fear that mathematics comes with “mathematical authority”. In this case, this authority is absent because we’re not dealing with a mathematical question (**). We’re dealing with the question: how good is the input? I’m not certain people are mathematically savvy enough to see that. Manne’s assumption is quite clear from the beginning, while yours are hidden behind the numbers you plug in.
(*) I may be wrong here, perhaps Manne is certain that Biden is guilty because he committed minor crimes. You disagree (and I disagree with you) but your model is not a good counterargument. Plug in the right numbers and you get a probability P(S/M) inspiring high confidence in Manne’s point of view.
(**) Actually, it’s also a mathematical question. This discussion has shown that other, more Bayesian approaches are possible, but for argument’s sake I stay within the framework of your model.
I hope this clarifies things a bit.
If Kate Mann was in another situation not so strongly influenced by prior belief she would follow the correct path of probability and refrain from judgment. This sort of swerve into irrationality is commonplace and is analogous to the memory deficit confabulation. In this case it is causal ‘confabulation’. We saw it with those who denied that covid was anything more than a normal flu. Their rational mind then demanded that they deny excess deaths even though the evidence for such was overwhelming unless ice rinks and refrigerator trucks turned into morgues were merely theatrical ebullience. They had to believe this given their prior commitment to big governments’ plan to take away the liberty to do as they pleased of free-born Britons and Americans. Quite intelligent and philosophically trained persons in Terf/Tran disputes take the same confabulatory route. The BLM riots/demonstrations have a number of such causal blackouts.
No, I’m not frustrated at all. A little bit bemused (and amused) by how ideologically captured some very smart people are, and how much they remind me of Trump supporters, only pointed in the opposite direction. (I actually could hear some Trumpkin making arguments about “black criminality” and “records of low level criminal behavior” and certain people being “the type” and how “victims are afraid to come forward” or it being “politically inconvenient” that would be virtually identical to the bizarre exercise in special pleading on behalf of Manne that D.D. engages in above.
“No, I’m not frustrated at all. A little bit bemused (and amused) by how ideologically captured some very smart people are, and how much they remind me of Trump supporters, only pointed in the opposite direction.”
This is off topic, so feel free to delete it, but this is something I have noticed too. “Progressives” and “conservatives” nowadays often use tricks from the same rhetorical toolbox.
Perhaps I’m getting old.
I’m seeing signs of change, though. We have a young left-wing mayor (his family is from Morocco), but he didn’t permit anti-racist demonstrations – public health was more important in these corona-times.
In general, I have the impression that the left – not the same thing as “the progressives” – is slowly moving back to its socio-economical roots, focusing more on things like income inequality and less on things dear to the gauche caviar.
One factor may be that there’s a generational turnover in politics. Young left-wing politicians “grew up with the rhetorical toolbox” and perhaps dislike it as much as I disliked the political climate I grew up in.
Hi David, thanks. Where I depart from Manne is at your 3c. I don’t see an argument for it *as the essay is written* that is valid.
Thanks. I agree but for one thing. The difference is that I am not arguing that Biden is innocent because it is possible that he is – that would be silly. Manne, however, is arguing that he is guilty because of his “type”, i. e. entitled creep. If, as you point out, this depends on the numbers, then her argument is incomplete and invalid as written because it doesn’t include the numbers. Compare this argument: (1) Jamal stole a case of beer; (2) theft indicates an anti-authoritarian rebellious nature; (3) membership in a terrorist organization also indicates an anti-authoritarian rebellious nature; (4) so Jamal is probably a terrorist because he is the type, i.e. anti-authoritarian rebel. *As written*, this argument is no good and would likely be taken to indicate racial or religious bias. Now, if I add (3a) “we know from studies x, y, z that 99% of beer thefts are committed by terrorists”, then sure, the argument becomes convincing. But without (3a), or something like it, no go. I argue that Manne’s essay *as written* is logically analogous to (1) – (4) without (3a). So I agree with what you say about numbers but see it as telling against Manne’s argument. Perhaps this doesn’t move you but it is how I see the structure of her argument.