The John Doe Fiction Continues

Last November, Rolling Stone magazine published a story describing a University of Virginia student’s account of her alleged gang rape at a campus fraternity house. The story was quickly embraced by many, particularly on the left. Just as quickly it fell apart when details in the story didn’t match reality. The magazine then asked Steve Coll, dean of the Columbia School of Journalism, to investigate what went wrong. Rolling Stone published Coll’s full report online. The magazine’s editor, describing the report as “painful reading” apologized to his readers.

Reading Coll’s account, it is evident the Rolling Stone writer allowed herself to be blackmailed by her source. When she pushed the source for the names of people who could confirm the accusations, the source stopped communicating. The writer developed a fear of losing the story entirely if she pushed too hard.

When the National Review was caught in a lie, its reaction was very different. The National Review and other conservative media have published articles purporting to describe what happened during alleged “raids” aimed at collecting evidence during the two John Doe investigations involving Scott Walker by the Milwaukee County District Attorney. These reports, as I’ve previously written, had remarkably similar elements. Four serious charges appear again and again, both in the articles and in the court filings from the lawyer David Rivkin: (1) an abusive chief investigator, who shouts, threatens, gets angry at any reasonable request, and uses intimidation to get the resident talking; (2) denial of a request for a lawyer; (3) not reading–or not allowing the resident to read–the search warrant, and (4) not reading the Miranda rights. No corroborating evidence is offered in any these articles.

But the release of an audio tape and other evidence from one of these search operations—that of Cindy Archer’s house as part of the first John Doe–makes it possible to check the veracity of these claims. Between an article in the National Review by David French and Archer’s complaint to the court, all four of the major accusations are made:

1. “The officer or agent in charge demanded that Cindy sit on the couch, but she wanted to get up and get a cup of coffee. ‘I told him this was my house and I could do what I wanted.’ Wrong thing to say. ‘This made the agent in charge furious. He towered over me with his finger in my face and yelled like a drill sergeant that I either do it his way or he would handcuff me.’ “ (French article)
2. “They wouldn’t let her speak to a lawyer” (French article)
3. “When Archer opened the door, officers and others flooded in, throwing the warrant at her without giving her an opportunity to read it.” (Archer’s complaint)
4. They “detained Archer in her home for hours of questioning without giving her a Miranda warning.” “Out of fear and intimidation, Archer agreed to answer their questions and to cooperate with their wishes. No one informed her that she had a constitutional right to remain silent and the right to an attorney.” (Archer’s complaint)

Which of these major charges turns out to be true? The answer is none. The tape contradicts all four.

The steady stream of articles in National Review, The Wall Street Journal and others make the same four allegations regarding other search operations by investigators But no corroboration is ever offered for these either. How should we regard these claims in light of the Archer results? In statistical terms, when a sample (the Archer search) is taken from a larger population (all the searches that happened in both John Doe investigations), the expected value of the population is the average value of the sample. Therefore the expected percentage of claims that would be found true if the whole population were examined is zero, often called the “point estimate.” Perhaps information on other searches will emerge in the future and the point estimate will have to change, but for now it is zero percent. And if the Wisconsin Supreme Court has its way and all the evidence is destroyed, we may never get that additional information.

Ideally, the single value is accompanied by a “confidence interval.” Unfortunately, with a sample size of four, no confidence interval can be calculated.

When caught in a lie, the press has two choices. It can come clean and frankly explore why the process went wrong. Rolling Stone, to its credit, made this choice on its faulty fraternity rape story.

Or it can choose to double down and claim the new information doesn’t disprove its story. National Review made this choice when faced with the evidence from the Archer tape. In a follow-up article, David French concedes that all four of the claims were false:

1. “Weiss [the investigator from the Milwaukee DA] is indeed unfailingly polite on the tape.”
2. “Weiss did in fact advise her of her right to an attorney”
3. “Beginning at the 4:00 mark, Weiss reads the extraordinarily expansive search warrant and then — at 9:20 — he reads the John Doe secrecy order.”
4. French also concedes that the Miranda warnings were read.

Yet in every case French’s concession that these four claims are untrue is grudging and accompanied by attempts to spin them to his advantage. Here is how he words the Miranda warning concession:

In fact, the delivery of the Miranda warnings — at an indeterminate time after the raid began — is the one tiny ray of good news on the tape. On a day when an innocent was subjected to the terrifying experience of an early-morning armed raid, was exposed naked in front of the police while her partner’s shower was interrupted by law enforcement, and received a gag order even as the press and public gathered around her besieged home, she was at least eventually told she was entitled to a lawyer and had the right to remain silent.

The proper reaction to his complaint about the timing is “so what.” The warning was read before the interview was started, which is when it should be. As for some of French’s other claims here, they relate more to what Archer did and felt than what the agents did (that she panicked and opened the door while naked), or are not provable one way or another by the available evidence (that the press was tipped off by the DA’s office).

To make mistakes is human. We all do. Is French an intentional liar or the innocent victim of mistakes, perhaps Archer’s faulty memory?

This is a question I face often when I set out to write a Data Wonk column. The column is often dependent on research done by others. Can I trust their integrity? In many cases their data may have taken months or years to develop and there is no practical way to check their calculations.  I find that one useful tool in assessing integrity is to look for invisible elephants, relevant information that goes unmentioned because it contradicts the source’s agenda.  

French’s invisible elephant is the FBI. The collection of evidence at Archer’s house, while mounted at the request of the Milwaukee DA, was managed by the FBI. Six of the ten people involved were from the FBI. The Team Leader was an FBI agent. An FBI agent joined Weiss in interviewing Archer. Yet the FBI is strangely absent from both French’s articles and the Archer complaint. To mention the FBI would have undermined French’s agenda of vilifying the Milwaukee DA.

French also doesn’t mention the two other agencies, the Dane County Sheriff and the Dane County DA, which, between them, contributed three people. Weiss was the only employee of the Milwaukee DA office present. French tries to explain away the lack of evidence for his abusive investigator claim by saying it must have happened before the recording was started, that is, before Weiss entered the house. In other words it was due to actions by employees of one of these other agencies. In his attempt to slime the Milwaukee DA office, French succeeds only in sliming the agencies he does not name, including the FBI.

Besides invisible elephants, I look for a candid acknowledgment of uncertainty. Honest researchers report confidence intervals as a measure of uncertainty. For French there is no uncertainty. Things happened exactly as he said, even though he wasn’t there. When his version of events turns out to be false, he creates an alternative version. Even after blaming Archer’s faulty recollection for the “divergence” between his claims and the recording, he remains absolutely certain that his revised version of events is the absolute truth. He quotes Archer’s lawyer, David Rivkin, as saying that “given the trauma of the day,” it was “not surprising” that she “may not remember every detail with perfect clarity.”

The available evidence suggests that the untrue assertions result far more from French and Rivkin’s political agenda than Archer’s faulty memory. The woman on the tape comes through as extremely loyal to Walker, describing him as “squeaky clean.” She is clearly devastated by the Walker administration’s decision to put her on medical leave and then demote her. She desperately wants her old job back—or something of equal importance. She comes through as being extremely eager to please, confiding things to Weiss unlikely to make her bosses happy, such as expressing her frustration that Walker found it hard to make decisions or that Walker wanted her “to get Hiller off his back.”

In examining why the campus rape story was insufficiently fact-checked, Coll identified confirmation bias. This is the tendency of people to be trapped by their pre-existing assumptions and to select facts that support their own views while overlooking contradictory ones. The writer, he said “believed the university was obstructing justice. She felt she had been blocked. Like many other universities, UVA had a flawed record of managing sexual assault cases.” Further, the alleged victim’s “experience seemed to confirm this larger pattern. Her story seemed well established on campus, repeated and accepted.”

Confirmation bias also helped explain the uncritical acceptance by many who read the campus rape story. For many on the left, in particular, the story confirmed what they already believed.

Confirmation bias is not unique to the left. Of the over 2,700 comments posted on the original David French story, only a handful express any doubts about the “facts” in the account. Confirmation bias could help explain the willingness of a federal judge and four Wisconsin Supreme Court justices to proclaim that the lies spread by French, Rivkin, and others were true despite the lack of any confirming evidence and the failure to even hold a hearing.

What would William F. Buckley, the founder of National Review, think of it handling of this issue? Would Buckley be shocked to see the magazine he founded become an instrument of untrue propaganda? Buckley is often credited as the mastermind behind the modern conservative revival, perhaps ranking second in the conservative pantheon after Ronald Reagan. But it turns out there was a good Bill and a bad Bill.

The good Bill loved intellectual ferment. He was the friend to many prominent liberals whom he invited to spar with him on his show. In starting his magazine he praised “non-licensed nonconformists” and famously described the National Review as standing “athwart history, yelling Stop.” He opposed some of the most unpleasant characters calling themselves conservatives, including the John Birch Society, George Wallace, anti-Semites, and racists. After initially defending segregation, he reversed course and turned against it. I think the good Bill would be dismayed at how the non-licensed nonconformists he hoped would lead the conservative revolution have turned into ditto-heads, and dismayed at the willingness of the readers of the magazine he founded to suspend their critical faculties.

At the same time there was a bad Bill, who embraced dictators such as Franco because they were anti-communists and justified lying in the service of a good cause. The bad Bill appears most clearly in a book he coauthored with his brother-in-law defending Wisconsin’s Senator Joseph McCarthy. Although concluding that “some of his specific charges were exaggerated; a few had no apparent foundation whatever,” they praise McCarthy for alerting the nation about vulnerabilities in its security.  Perhaps the bad Bill injected the idea of lying for a good cause into National Review’s DNA.

There are many thoughtful conservatives in this country today. Unfortunately they have been marginalized by people like Rivkin and French who apparently take the bad Bill as their role model, and who continue to spin a story that has no apparent foundation.

End Notes (For those who crave even more details…)

1 French also claims that “the police searched her house, making a mess, and — according to Cindy — leaving her ‘dead mother’s belongings strewn across the basement floor in a most disrespectful way.’” The FBI log lists 20 exit photos of the rooms in the house, which, according the prosecutors’ response, show this charge was untrue. I don’t include this charge in my count, partly because it seems less serious than the other four and partly because I did not view the disk with the photos. In any case, the fact that the FBI saw fit to document the condition of the house after the search suggests that this is a common charge after FBI searches and, for that very reason, agents are expected to take pains to make sure it is untrue.

French also states, “Then they left, carrying with them only a cellphone and a laptop.” The use of “only” suggests this is also a criticism, suggesting that all this activity yielded very little. In any case it is also untrue. The FBI report list seven items, for which a receipt was given to Archer.

There are other claims made in the article and/or complaint, but they should be discounted because they relate more to what Archer did and felt or are not provable one way or another by the available evidence.

2 Readers fortunate enough to have taken statistics may recall that the formula for proportions is where is the proportion in the sample found true (0% in this case), n is the sample size (4), and z is 1.96 for a 95% confidence interval. This gives a point estimate of   Unfortunately this formula only gives an accurate estimate with large sample sizes (usually over 30) and where n times  is larger than 5. Our sample fails on both counts.

Acceptance sampling often deals with small samples. It involves taking a sample from a shipment and using the sample to decide whether or not to accept the shipment. There are dangers: accepting a bad shipment (called the consumer’s error) and rejecting a good shipment (producer’s error). For this to work, a first step is to agree on what makes a shipment good and what makes it bad. For example, a purchaser of grain might find that it gets satisfactory results so long as the defective kernels are under 20 percent.

The ancient Greek Diogenes is said to have wandered the marketplace carrying a lamp in broad daylight. When asked why, he answered that he was looking for an “honest man.” So far as we know he never found one. Perhaps his problem was that he never sufficiently defined what he meant by an “honest man.” To help out Diogenes, let’s specify that an “honest man” is someone who tells the truth at least 80% of the time (to err is human, after all). Clearly Diogenes doesn’t have the time or resources to test a person’s every statement. Instead, he takes four of the statements and tests them. If all four are false, what is the chance that Diogenes will reject a true honest man? Acceptance sampling puts that risk at well under 1% (0.16% to be exact).

One possible argument against the acceptance sampling approach is that it assumes a randomly-selected sample. That is true. However, Rivkin took the initiative in bringing the lawsuit. It strains credibility that he would choose his weakest sample. 3 For example, there is a group that has massaged census data so that it can be used to calculate population values based on how far people live from the center of a metropolitan area, rather than what political jurisdiction they live in. This allows a much more realistic comparison between cities. Boston itself, for instance, occupies only a small fraction of its metropolis; whereas Houston has managed to annex most of its suburbs. 4 For instance, any report from the Wisconsin Manufacturers and Commerce is likely to avoid mentioning any research that goes against whatever the WMC is currently arguing. 5 She also comes across as somewhat naïve, at the end asking Weiss for advice on whether she should get an attorney. 6 This is an oddly immoral book, identifying some of McCarthy’s excesses, distancing themselves from them, but often on the basis that they are ineffective: “For one thing, in calling Drew Pearson and the Milwaukee Journal pro-Communist, he loses the opportunity to make the far more telling indictment of them that their performance invites; that, whatever their intentions, they do act in a way that advances the fortunes of Communism.”