This came up recently:
[Compared to politicians who garble the facts, Niall] Ferguson, though, is in a more difficult position. As a professional historian, he cares about the truth. He’s been known to garble the facts from time to time, but, hey, he’s a busy guy, and he’s effectively just trading off accuracy for speed, in the same way that we learned about in typing class so many years ago. I don’t think Ferguson wants to make errors; I think he wants to write more articles and books (not a bad thing! He has ideas he wants to share with the world) and doesn’t want to slow that output by stopping to check for errors or correct them when they appear.
At first glance it might seem that, for a successful academic historian, the expense of checking for, acknowledging, and correcting errors is small compared to the reputational hit of making these high-profile mistakes. Ferguson could just hire a research assistant, some Stanford student who could check everything he writes and flag the mistaken statistics and erroneous claims. The real cost would not be paying the student, however. Rather, the real cost is that, if Ferguson was restricted to only stating true facts, it would reduce his flexibility in making the larger claims he wants to make. Being willing to stretch the truth—not by flat-out lying, I think, but rather by following a general practice of not checking his statistical and historical claims—gives him extra “researcher degrees of freedom” (in the words of the famous Simmons, Nelson, and Simonsohn paper) in his theorizing. Fact-checking would reduce Ferguson’s effectiveness as a theorist and as a big-picture historian by constraining the sorts of things he could say.
This was just an offhand comment, but it got me thinking that this is maybe a special case of a general principle.
Sometimes when people get things wrong, the motivation is clear: getting the facts wrong helps them tell a story. We’ve seen this over and over with New York Times columnist David Brooks: whether he’s disseminating made-up statistics from other sources or lying about restaurant prices, it’s all in the service of his prepared narratives.
Other times people can just get misled by the literature, as with those researchers who thought that peak fertility occurred during days 6-14 of the monthly cycle. It’s tough to catch that sort of mistake before your article is published, because if you’d thought to check, you wouldn’t have made the error in the first place, and it’s easy for this sort of factual error to not get caught in the reviewing process.
Other cases, though, it doesn’t seem so hard to check your facts, or your statistical analyses, and one thing that’s been bugging me awhile is, why aren’t people more careful? I make mistakes too—that’s why I’m always checking my calculations in different ways, and that’s why I appreciate and acknowledge the corrections I receive from eagle-eyed readers.
This Ferguson thing, though, it got me thinking that being sloppy isn’t just laziness, and not acknowledging or correcting errors isn’t just a fear of embarrassment or a perceived need to appear tough. Rather, sloppiness and a refusal to engage with errors can be seen as an active strategy for success as a scholar and pundit, in that it gives the executor of this strategy valuable “researcher degrees of freedom.”
Don’t get me wrong. I’m not saying that Ferguson, or Brooks, or the Nudge guys, are sitting in some room, rubbing their hands and cackling, Snidely-Whiplash-style, about their nefarious strategy of being sloppy with the facts. My take is that being sloppy and not coming to terms with error are advantageous strategies, and when researchers or pundits of low scruple drift into that approach—perhaps because they’re on deadline or are busy and don’t feel they have time to check those too-good-to-check stories—that it’s natural for them to stay there. There are positive advantages to not restricting yourself to the truth.
What’s the big deal?
This is all obvious, you might say, and maybe it is. So why am I so excited about this—excited enough to dedicate one of our 600 or so blog posts of the year to the topic?
I’m excited because I think this latest insight, obvious as it may be to you, can be the key to resolving one of the replication crisis’s key issues—what we might call the “knave or fool” question.
Recall Clarke’s Third Law: Any sufficiently crappy research is indistinguishable from fraud.
When writing about Clarke’s Third Law in the past, my point is that intention doesn’t matter so much. For example: “Was he softly cheating, by purposely not looking into the possibility that his method might be wrong, just looking away so he could continue to use the method? Or was he just incompetent, try to do the scientific equivalent of repairing a watch using garden shears? I’m not accusing any of them of fraud! Who knows what was going through their mind when they were doing what they were doing. . . . it doesn’t matter.”
I’ve also emphasized that honesty and transparency are not enough. From one direction, just cos we point out a fatal and avoidable error in a published paper, that doesn’t mean we’re saying the authors did it on purpose. From the other direction, don’t think that just because you’re pure of heart, that your research can’t be wrong.
The common thread here has been our attempt to avoid the “knave or fool” question by saying that intention doesn’t matter.
But one of the most common modes of being a fool is to make sloppy mistakes and then to dodge and avoid accounting for criticism (see here for more examples).
So there you have it. An effective strategy for getting ahead is to be sloppy and never look back. I feel like we still need to walk a fine line here, because I don’t want to say that people who do sloppy research are bad people; it’s more that they’ve (implicitly) weighed the options and decided that this is the best way for them to advance and get their message across.
This is a version of my grifter point about Silver, Gladwell, and the Freakonomics guys. They’ve found something to sell to an audience and they’re not willing to let other factors stand too much in the way of selling that product.
With historians, the tells are: 1) too much produced too quickly. Good history takes time and the historians producing a book a year or every other year are cutting corners *somewhere,” 2) Writing larger overarching narratives of history that explain everything, 3) wandering outside their area of expertise without any sense of caution. Ferguson fits those pretty well.
While I am pure of heart, I try to do everything like Snidely Whiplash. The Dudleys of this world expect it.
+1 This reply cracked me up.
I think the word “research” is doing a lot of work here. Lately, research often = doing some Google searches. Does David Brooks do research?
The definition of research – or at least how many people have come to definite it – has become an issue of concern for me.
Sorry – somehow I nested that in the wrong location.
All part of my dastardly plan….
Where’s Dudley Do-Right when you need him?
Sean Lock apparently agrees with you:
https://www.youtube.com/watch?v=0UGuPvrsG3E
I think much of the research in question should not be called “sloppy” or “crappy.” A couple of examples come to mind: the Chinese laborer/boatman story misreported as fact; a reference I can’t quite recall where Tufte had a great example that wasn’t quite accurate (it might have been something about the Challenger disaster, but I may not be recalling the right example). I think many of the examples (certainly not all) are cases where a neat story with some public appeal fits what the researcher was trying to say. If the example had been researched more thoroughly, it would have turned out that things were not so clear – as most things are not so clear. But there is little incentive to do so because the story is appealing, fits the narrative, and isn’t too far off from possibly being accurate. I am not sure I want to denigrate all such practices – aren’t stories by their very nature somewhat exaggerated or inaccurate?
Of course, the harm is that this is a slippery slope. Exaggeration quickly becomes misinformation and provides models for others to get careless with the truth. On the other hand, insisting that every reference be carefully researched and accurately represented in all details, seems like a bit of an unproductive digression from the story. Find the line between where this is damaging and where it is innocent seems like a difficult thing to do. It’s easy in such situations to call for absolute truth-telling, but I’m not sure we’d really be happy with that outcome. Few examples (for instance, historical events) are so clear cut that they can easily be accurately reported.
With statistical analysis I think this becomes a somewhat easier line to draw. If the code and data are both provided, then that protects against potentially dangerous misrepresentations. Without the data and code, much more detailed reporting is required – all those forking paths would need to be described. I guess there would still need to be a line drawn regarding what is meant by “all,” and that line might depend on the intended audience. When viewed by similarly trained researchers, we might take for granted a certain understanding of things that don’t require more intense scrutiny – but for a less expert audience, any short-cuts become dangerous.
I am inclined to think that it is unwise to insist on complete accuracy. But that means that criticism should be doubly welcome, acknowledged, and appreciated. Some stories that are not thoroughly researched are dangerous – hopefully, these get noted and disseminated (though we have seen there are many problems standing in the way). Perhaps the Chinese laborer story was not a serious misrepresentation – but when/if it is, post-publication review is an appropriate response. Burdening authors with ensuring all details are accurate doesn’t seem like a good idea to me (although in that particular case, accurate attribution would seem to be an easy standard to insist on – I don’t see how it helps the story to misattribute – your post, https://statmodeling.stat.columbia.edu/2022/02/27/did-chinese-laborers-on-the-yangtze-pay-someone-to-whip-them-and-why-cant-political-scientists-and-economists-resist-telling-this-evidence-free-story/, explored this quite thoroughly).
I agree with your take in the first paragraph, except that I think a lot of the cases Andrew refers to are ones in which the denigration is justified. For me the reason for this is that the person telling the ‘story’ is not carrying weight with their audience/commanding an authoritative influence in the public eye on their supposed merits as a storyteller per se. Rather, the audience believes that what they are hearing is actual facts or pieces of scientific information crafted very neatly into a nice story as legitimate examples that make the point – the person telling the story is carrying gravitas on account of their expertise or knowledge and the stories are supposed to be some legitimate expression of that and what is expected to come with it (e.g., a behavioral scientist won’t just be making up stories or telling ones that they know to be false unless signposting that they are doing this). It’s precisely because a scientist or historian is seen to be being more rigorous and factual than someone who is merely telling a story to convey their opinion that they stand above a simple talking head. It’s something about that mixture of ‘infotainment’ – where part of the value comes from believing you are also acquiring legitimate knowledge as well as being just entertained.
This is very interesting. How about this? A story can do two things for a piece of exposition: (1) It can provide a vivid, concrete example of a concept that might otherwise be abstract, abstruse or ab-something else. (2) It can provide factual evidence for a general claim, especially if the story pertains to a large-sample statistical result, but even anecdotal evidence is something.
The problem is that stories that may be less than factual for #2 can serve the #1 purpose brilliantly. Ideally, we would keep these two functions in mind as distinct, so we can qualify likely factuality without impinging on the heuristic dimension, but believing a story to be factual makes its communicative power that much greater. A little slippage here is understandable, but yes, there has to be a line. (And it’s often the case that arguments that need factually suspect support are themselves suspect.)
Peter yes, that is a more articulate distillation of what I was trying to drive at!
I think also, when a story is actually true and not simply intended to be thought provoking/allegorical or demonstrative as per purpose 1, it can be more convincing because people have to update their model of the world to encapsulate something that supposedly actually happened, rather than merely thinking ‘interesting idea but not sure I agree’
In statistical science, the related practice is using small sample sizes and cookie-cutter methods to get a minimum publishable unit. Lots of researchers know these methods are not good but they are caught on the treadmill of needing publications to get the next grant or postdoc. And any one researcher who is slower and more thoughtful will be naturally selected out of the researcher pool.
And yes Total, trade presses and the media absolutely want people to publish books faster than most people can write substantive nonfiction.
The idea that Ferguson prefers a good story to careful truth is an important insight.
But it’s also important to remember that his entire “brand” is a certain kind of conservative. Any political stance like that will make it hard to be honest. The comment about Keynes was, as he acknowledged, silly, but attacking Keynes as the origins of all evil plays well to the kind of audience Ferguson wants to cultivate.
Best blog title ever
This post reminds me of an email thread Andrew and I were both on recently, so I’ll just repeat something I said there:
Given that there can often be a real advantage if one is willing to oversell or fudge the details, there is always room to tell oneself that these transgressions are just going to lead to more good done or higher impact science in the long run. Ultimately it seems how much refusing to lie or exaggerate hurts versus helps a person’s reputation is not that predictable at the individual level, and so lots of people are willing to chase popularity over rigor. I think of this a lot in the form of the more generic trade-off between how willing one is to do superficial work that sells versus being committed to doing deeper work.
Jessica:
A complicating factor is that “rigor” means different things to different people. For example, someone might say that Aki and I chased popularity over rigor when writing Regression and Other Stories, a statistics textbook that contains no theorems. In turn, I would say that traditional regression textbooks with their careful derivations of least squares etc. are low on rigor regarding the interpretation of regression coefficients and p-values.
Recall our discussion last year about purported “rigor-enhancing practices” in psychology research: What they thought was the key to rigor was not what I thought was the key to rigor.
From Andrew’s post: “[…] what we might call the “knave or fool” question”. In my experience, this is a bit of a false dichotomy. I have met very few researchers in the social sciences that are pure ‘knaves’, in the sense of having full understanding of their misdoings and pursuing them anyway. Rather, I think there is a pervasive chosen and socially reinforced foolishness about one’s own knavishness.
Most people start their careers with a healthy dose of motivation to really do good research. But incentives and a culture that are poorly aligned to research integrity create a context wherein there are much higher benefits than costs of not engaging all too critically with research methods. Those that can’t deal with the cognitive dissonance between integrity and success, tend to leave academia eventually. Most others develop individual and collective strategies to deal with it, and often that implies a great deal of rationalization and avoidance.
Changing incentives is possible, but I think it requires top-down decisions to fundamentally change how research activities are rewarded. These are decisions that I think are unlikely to materialize in the short term. Those who can make them, if able to recognize the problem at all, are often those that build a successful career within the above-described context – and might therefore find it difficult to address a problem that reflects poorly on their own success.
I think changing the academic culture can be a more immediately effective bottom-up strategy. But it does involve be willing to critically scrutinize our own and each other’s work in a much more critical manner than is the current social norm. And it will involve resisting old and contemporary trends to insulate academic work from rational criticism. That will always be an uncomfortable activity, given that implies ramping up the experienced cognitive dissonance.
You know that brilliant exchange between Carmella and the psychiatrist in the Sopranos (S3E7)? “One thing you can never say. That you haven’t been told.”
>> excited enough to one of our 400 or so blog posts
Missing word?
excited enough to *dedicate* one of our 400 or so blog posts
Jm:
Typo fixed; thanks. Also I changed 400 to 600 as that’s more accurate. (We’ve published 11,592 posts in just over 20 years.)
A correction and some praise. You’ll probably denigrate the latter as your sophisticated views and willingness to forgive are more greater and nuanced than mine.
Now I’m feeling a lack of confidence about the correction! Okay, here goes. Unsure about this word, perhaps is ‘because’ not ‘cos’?
“From one direction, just cos we point out a fatal and avoidable error in a published paper, that doesn’t mean we’re saying the authors did it on purpose. From the other direction, don’t think that just because you’re pure of heart, that your research can’t be wrong.”
Now for the praise which ties in to the typo (thus why I included such a lengthy excerpt above): I love Clarke’s Third Law of Research! So true! Any sufficiently crappy research is indistinguishable from fraud.
Final, excellent observation that goes well with the theme of (detecting) bad science and bad scholarship in most fields, not solely history, is Total’s criteria number 1:
“too much produced too quickly. Good history takes time and the historians producing a book a year or every other year are cutting corners somewhere.”
See, I do it too, and acknowledge unlike David Brooks. (Paul Krugman is often guilty too, speaking of NYT columnists, but I haven’t read either in years, so maybe no longer?) I made multiple typos.
I didn’t intend to write
“your sophisticated views and willingness to forgive are more greater and nuanced than mine.”
That’s cringe.
*your sophisticated views and willingness to forgive are more nuanced and greater than mine” (respectively)
I think this is a classic externality problem (the tragedy of the commons). These strategies seem to pay off for a lot of people, but at the same time undermine the credibility of their scientific field.
Moreover, the incentives to combat this are misaligned too because of externalities: significant personal cost for an important social gain.
I don’t have much hope for the medical or social sciences at this point, and I understand why so many people are skeptical of their results.
Perhaps in the long (long) term they might hurt the credibility of the field, but I think in the short term they may actually boost the credibility of the field. Consider something like Nudge Theory, after enough time people may stop taking Behavioural Economists/Psychologists seriously, but in the short term they get more attention, praise, and funding. This is the case even if you’re not one of the seminal figures in the field. I think this makes it even tougher to combat this issue, because you’re essentially spoiling the pot for everyone in the field, not just the individuals at the center of it
I was a history major at Wisconsin back when ideological diversity prevailed. The rise of Hitler was a hot topic. The Rise of the Great Man theory didn’t make sense. Brilliant and creative cultural historians like George Mosse (and Fritz Stern at Columbia) put forward explanations that examined the cultural and ideological precursors of Nazism in German culture. Marxist historians looked at the alliance between fascism and big business. Conservative histories like Stan theorized about fascism, and were more sympathetic to Franco’s variety in Spain. Niall Ferguson’s War of the World, published in 2006, is a masterpiece which obsoleted (or superseded) these earlier explanations. One of the great history books of all time. Read it and decide for yourself.
Joey:
I’ve not read that book, but I was a big fan of an earlier book by Ferguson. One reason it saddened me that he moved toward becoming a political hack is that it was such a step down from his earlier work. Don’t get me wrong—I can understand his motivations for moving in that direction, which is a motivation for political hackery in general: he has strong political views and is willing to cut intellectual corners to make his points. This is the argument of political activists everywhere, and I can’t say it’s wrong, even if it’s not the direction that makes me comfortable.
The title of that Ferguson book you liked reminds me of Dakin’s “subjunctive history” in The History Boys.
Ferguson’s _War of the World_ is a hastily written book (likely mostly by his research assistants) that (as frequently with Ferguson) so in love with overturning received wisdom that it often gets things quite wrong indeed. As Gerhard Weinberg, the eminent historian of WWII, said in his review of the book “there is not enough space to list even a portion of the blatant errors.”
As an organizing shorthand, one might look at `academic reputation x publication quality lower bound`:
Fools = (low, low)
Misunderstood = (low, high)
Knaves = (high, low)
Aspirational = (high, high)
The thing about finding a knave in the non-moral sense of “just published something too bad to be consistent with his reputation” is that it implies they have successfully broken a field’s reputational mechanism, so you either ignore everything they say or incur the cognitive cost of evaluating every statement as if it came from an unknown person with at least some history of dishonesty.
I’m tempted to express this as an indefensibly stylized inference model for a joint distribution over `(quality of next publication, reputation, honesty, quality of past publications)` with `honesty` as a latent variable, but I fear that’d be foolish.
> I’m tempted to express this as an indefensibly stylized inference model for a joint distribution over `(quality of next publication, reputation, honesty, quality of past publications)` with `honesty` as a latent variable, but I fear that’d be foolish.
Alright, I lie – its something I’ve been looking into doing as a content-agnostic approach to sociopolitical disinformation (hopefully with a less naive approach), but it’s not quite yet at the top of the Infinite To-Do List.
Marcelo:
That’s interesting. I don’t quite follow the details of what you’re saying, but I like the idea of connecting this to reputation. As has been discussed by other commenters, the fool/knave distinction is tricky because one way to be a knave is to actively avoid realizing you’re a fool.
Scientific production (of a scientist) may be influenced by:
1. Motivated Reasoning Doesn’t Apply to Me. (Naivette.) Ideology and confirmation bias may skew both how much due diligence they do in what they cite but also what they produce. Of course, it also likely affects what they read, etc., etc. The other consequence that you wouldn’t see PS experiments convincing people against legal abortion. I don’t even know if IRB will approve it. (That would be a nice experiment to do about what the IRB approves and not.)
2. Interest in persuasion. Even if scientists know some of the examples are wrong, they still use them because they think they can convince others of their viewpoint.
3. Playing to the crowd. Optimal strategy for publishing may be to speak to the priors of your audience. So researchers may knowingly or unknowingly frame things in ways that cater to the priors.
4. Storytelling bias. There is a coherence bias when publishing papers. There is some implied pressure that you must wrap the findings in a nice bow. It can’t be that experiment 1 showed X and 2 showed Y and hey, I don’t know what’s up.
…..
———-
“So there you have it. An effective strategy for getting ahead is to be sloppy and never look back. I feel like we still need to walk a fine line here, because I don’t want to say that people who do sloppy research are bad people; it’s more that they’ve (implicitly) weighed the options and decided that this is the best way for them to advance and get their message across.”
so these are not ‘bad people’ who choose the ‘wrong thing’ (my phrase) because they have ‘weighed the options and decided that this is the best way for them to advance.’ depends on what you mean by ‘bad.’ are you talking about bad in a consequentialist way in that few things in social science matter so …