(This article was originally published at Statistical Modeling, Causal Inference, and Social Science, and syndicated at StatsBlogs.)
1. The king and his memory
There’s this stunning passage near the end of Josephine Tey’s classic The Daughter of Time. Most of the book is taken up with the main characters laboriously discovering the evidence that Richard III was not really a bad guy, he didn’t really kill those little princes, etc. Having made this discovery, the characters are planning to write a book and break the big news to the world.
But, then, some hours later:
And Brent came in.
But it was not the Brent who had last gone out.
Gone was the jubilation. Gone was his newly acquired breadth.
He was no longer Carradine the pioneer, the blazer of trails.
He was just a thin boy in a very long, very large overcoat. He looked young, and shocked, and bereaved.
Grant watched him in dismay as he crossed the room with his listless unco-ordinated walk. There was no bundle of paper sticking out of his mail-sack of a pocket today. . . .
‘What is it? Have you discovered that there was a general rumour about the boys before Richard’s death, after all?’
‘Oh, much worse than that.’
‘Oh. Something in print? A letter?’
‘No, it isn’t that sort of thing at all. It’s something much worse. Something quite—quite fundamental. I don’t know how to tell you.’ He glowered at the quarrelling sparrows. ‘These damned birds. I’ll never write that book now, Mr Grant.’
‘Why not, Brent?’
‘Because it isn’t news to anyone. Everyone has known all about those things all along.’
‘Known? About what?’
‘About Richard not having killed the boys at all, and all that.’
‘They’ve known? Since when!’
‘Oh, hundreds and hundreds of years.’
‘Pull yourself together, chum. It’s only four hundred years altogether since the thing happened.’
‘I know. But it doesn’t make any difference. People have known about Richard’s not doing it for hundreds and hundreds—’
‘Will you stop that keening and talk sense. When did this—this rehabilitation first begin?’
‘Begin? Oh, at the first available moment.’
‘When was that?’
‘As soon as the Tudors were gone and it was safe to talk.’
‘In Stuart times, you mean?’
‘Yes, I suppose—yes. A man Buck wrote a vindication in the seventeenth century. And Horace Walpole in the eighteenth. And someone called Markham in the nineteenth.’ . . .
Yup, it was known all along. But nobody wanted to hear it. Nobody wanted to spoil the pretty story.
OK, I don’t know anything about Richard III. And The Daughter of Time is fiction. So I’m expressing no opinion on what happened to the princes in the tower. Rather, I’m drawing attention to the compelling idea that an error in historical memory can remain in plain sight, that certain mistakes seem to need to be rediscovered to be believed.
I’ve remembered the above scene (not the exact words, but the general idea) ever since I read that book, nearly thirty years ago.
2. The discredited researcher
During the past year we’ve been hearing a lot about the misdeeds of Brian Wansink, the Cornell University business school professor who’s received contracts with major corporations, millions in government research grants, and tons of largely uncritical media coverage over the past decade or so based on his “nudge”-style research on eating behavior. It appears that that Wansink repeatedly was manipulating research materials, equipment, or processes, or changing or omitting data or results such that the research is not accurately represented in the research record. (I use that particular phrasing because it aligns with the NIH’s definition of research misconduct, from a webpage that was linked to by the Cornell Media Relations Office.)
The main discoveries of misconduct occurred in 2016 and 2017, but Wansink, for people in his lab, have a track record of ducking criticism, acting like they take it seriously but then continuing with sloppy scientific procedures; here’s a story that goes back to 2012.
The ratty sweater that is Wansink’s research record continues to unravel. Here for example is a post by Bettina Elias Siegel, “New Study Casts More Doubt on ‘Smarter Lunchrooms’ Data,” reporting on a debunking by Eric Robinson on some of the most influential work from that lab.
What I want to point to here, though, is something different. Linked from Siegel’s blog is this post of hers from 2014, “Moms, ‘Food Fears’ and the Power of the Internet.” Key quotes:
Dr. Brian Wansink, a professor of consumer behavior at Cornell University and director of the Cornell Food and Brand Lab, has published a new study in the journal Food Quality and Preference entitled “Ingredient-Based Food Fears and Avoidance: Antecedents and Antidotes.” This study, co-authored by Aner Tal and Adam Brumberg, seeks to determine why people – mothers in particular — develop so-called “food fears” about certain ingredients (such as sodium, fat, sugar, high fructose corn syrup, MSG and lean finely textured beef) and what the food industry and government can do about it.
The study’s ultimate conclusion, that “food fears” can be addressed by “providing information regarding an ingredient’s history or the other products in which it is used,” is hardly controversial. But some other things about this study raise red flags . . . in Wansink’s own YouTube video created to promote the study, he tells us that people with “really bad ingredient food fears have three things in common:”
First of all, they tend to hate the foods the product’s in, almost more than the [unintelligible] ingredient itself, meaning they tend to hate potato chips or candy or soft drinks almost more than the ingredients themselves.
Second of all, they get most of their information . . . from the Internet, they look at their favorite websites, they don’t get it from mainstream media and they certainly don’t get it from health care professionals.
The third thing they have in common is that they are much more likely to need social approval.
The problem is, Wansink’s study simply does not support these characterizations of individuals who get their food information from the Internet, and Wansink’s own recap of his study is in some ways as grossly inaccurate as the media reports I cite. Here’s why. . . .
The Study Did Not Address Social Media At All
In a survey of 1,008 women who had two or more children, the question posed to respondents about where they obtained information about food ingredients did not include the words “Facebook,” “Twitter,” “newsfeed” or even the more general term “social media.” Instead, respondents were simply asked if they obtained such information from “Internet/Online,” an incredibly broad descriptor which could include anything from the sketchiest of blogs to the website of the Institute of Medicine.
So, in fact, the study has nothing at all to say about the role of Facebook, Twitter or other social media per se in stoking “food fears.”
The Study Failed to Distinguish Between Types of Online Media
Wansink also contrasts what he sees as the largely biased Internet with more trustworthy “mainstream media,” but without acknowledging that almost every local and national news outlet operating in traditional media now also has its own website. (In fact, where did I find Wansink’s own study, the news coverage about it, and every other citation in this post? Online, of course.) . . .
So when Wansink says in his video that people with “food fears” “look at their favorite websites, they don’t get [food news] from mainstream media,” he has no basis at all on which to make this key distinction. . . .
Conclusions About Sharing “Food Fears” on Social Media Are Entirely Unsupported
As noted, Today reports that moms with “food fears” “feel strongly about sharing these opinions on social media or their own blogs,” and Wansink notes that “they have a higher need to tell other people about their opinion.” In his video, he says such people “are much more likely to need social approval.” (Emphasis mine.)
But while the study did find that “some individuals who avoid ingredients may have a greater need for social approval among their reference group than those with a more moderate view,” the study’s authors were forced to admit that “such effects were small in our sample.” So Wansink’s “much more likely” characterization is patently false.
And even if this finding were significant, the supposed need for social approval was not measured by respondents’ use of social media or blogs. Rather, it was measured using a standard “social desirability” assessment tool that has nothing to do with social media, and also by asking respondents if they agreed with two statements (“It is important to me that my friends know that I buy Organic Foods and Beverages” and “It is important to me that my friends know that I buy Natural Foods and Beverages”), neither of which have anything to do with “food fears” or social media. . . .
Those With “Food Fears” Are Not “Haters” of Junk Food
Wansink tells us in his YouTube video that those with “food fears” actually “hate” the product in which the feared ingredient is found more than the ingredient itself. Specifically, he tells us, the study found that they “tend to hate potato chips or candy or soft drinks almost more than the ingredients themselves.”
Now here’s what Wansink’s study actually found. Participants were asked to rate the healthfulness of four foods (yogurt, granola, pre-sweetened cereal and cookies). Some participants were then told that these four products contained HFCS and among that subset, the “healthy” rating went down for yogurt, granola and pre-sweetened cereal, but not for cookies (presumably because cookies are not thought to be healthful in the first place.)
And that’s it. Not a word in the study about “potato chips, candy or soft drinks.” Not a word about “hating.” But Wansink apparently likes this fictional finding so much he mentions it in his video not once, but twice.
People keep rediscovering that Wansink does not tell the truth about his experimental procedures, or his data, or his results. But, even all these years later, his claims are driving policy. Here’s Bettina Elias Siegel from 2017:
To date, the U.S. Department of Agriculture has spent $8.4 million to directly fund the Smarter Lunchrooms research and implementation, and another $10 million in grants to help schools put the Smarter Lunchroom principles into practice. The agency also now requires schools to implement some Smarter Lunchroom techniques in order to qualify as “HealtherUS Schools,” and the Obama-era federal rule on local wellness policies specifically informs school districts that
at a minimum, FNS [the USDA’s Food and Nutrition Service] expects [districts] to review “Smarter Lunchroom” tools and strategies, which are evidence-based, simple, low-cost or no-cost changes that are shown to improve student participation in the school meals program while encouraging consumption of more whole grains, fruits, vegetables, and legumes, and decreasing plate waste.
At the very least, perhaps the president of Cornell University could call up the USDA and apologize for all these wasted tax dollars, not to mention all the kids who had to put up with all this b.s. in their lunchrooms for so many years.
What’s striking about this story is that the relevant information has been in the public record all along. I guess that’s one reason Wansink thought he could keep doing it: he’d got away with it for years and nobody cared, so why expect anything to change. You think they’ll ever take him off the Ted webpage? I wonder what it takes to get de-listed by Ted.
P.S. Siegel did not start out as a Wansink skeptic; here’s an interview from 2011 where she fell for his whole shtick (and where Wansink, ironically in retrospect, is labeled as a “Master of Lunchroom Trickery”). The interview’s good: Wansink’s a good talker. I guess it’s easy to be persuasive, at least in the short term, if you’re not constrained by the truth.
P.P.S. In the months since I wrote the above post, lots more has appeared on Wansink; see for example here, here, here, and here.
That last link is particularly amazing—it’s the story of a $10,000 Kickstarter campaign “guaranteed to help users lose 2+ lbs each month,” which never got started:
Those who gave at least $30 were supposed to receive three months’ worth of the service, which would start right after the campaign ended in September 2014, according to the Kickstarter page. . . .
A few weeks after the campaign met its goal, Wansink emailed three dozen contributors . . . “We anticipate this will go live in January 2015 . . . Really exciting results so far, now we’re smoothing out the interface.” . . .
In February 2015, Star Li, who was listed as the campaign’s manager, wrote in response to Ruth, the donor in Israel who had kicked in $30: “This monthly method plan should be available within the next month or two. Our apologies for the delay.”
When Ruth later tried to follow up again, Wansink wrote in January 2017, “We’re still working on this.” . . .
In his [February 2018] email to donors, Wansink . . . said his team had worked with programmers for a year to create the service, but it didn’t materialize by the time he left for sabbatical in 2015. It was then that he decided to put it “on the back-burner” and develop a substitute service on a website and app. These were launched in beta-testing form, but have since been taken down for adjustments, he wrote.
It really is a lot easier to say you’re gonna do something, or to even say you’ve done something, than to actually do it.
Again, what’s stunning about this whole story is that so many of the problems were hiding in plain sight, just sitting out there for all to see, who chose to look. The strategy of affable denial worked so well for so long.
The post Concerns about Brian Wansink’s claims and research methods have been known for years appeared first on Statistical Modeling, Causal Inference, and Social Science.
Please comment on the article here: Statistical Modeling, Causal Inference, and Social Science
