Arnold Kling  

The Confirmatory Bias Diet

Reminder--Tuesday on Capitol H... I Can Drunk Just as Well Think...

John Ford writes,

Seth Roberts, a psychologist at UC Berkeley has written a book called The Shangri-La Diet. In it, Roberts described some old obesity rat data and via "self-experimentation" developed a technique for weight loss that he hopes will change millions of lives.

...However, the scientific method exists for a reason: to root out poor hypotheses and to direct research towards those more likely to be fruitful. If Roberts were truly interested in investigating his approach, he should have subjected it to the dispassionate rigor of clinical study and peer review.

I was foolish enough to buy the book and try the diet. It was a total failure. It did not work for fellow TCS writer Glenn Reynolds, either.

In fact, if you think about it, the last thing you should believe in is a diet that was found to be successful primarily by the inventor of the diet. Dieting means reducing calorie intake, and the question is how you can motivate yourself to do that in a sustained way. If you are convinced that a diet will work, that may motivate you to stick to it. Having invented the diet yourself, you may be more convinced that it will work, and you will stick to it more easily. This is a form of confirmatory bias.

So let me propound the confirmatory bias diet. It works like this:

1. Invent any diet, as long as it leads you to reduce your calorie intake by, say, 150-200 calories a day.

2. Convince yourself that this is a brilliant new diet.

3. Try the diet on yourself.

Because of confirmatory bias--the desire to seek evidence that confirms your prior beliefs, you will have success with the diet.

Do I have the next diet book?

Comments and Sharing

TRACKBACKS (3 to date)
TrackBack URL:
The author at voluntaryXchange in a related article titled The Shangri-La Diet writes:
    I started on the Shangri-La Diet about 6 weeks ago. Yes, it's crazy. Sip sugar water slowly all day, and lose weight. I wasn't going to post about this. I just thought I'd try it for quite a while and [Tracked on May 22, 2006 12:03 PM]
COMMENTS (16 to date)
mikeg writes:

Ford makes some good points, but really, how many diet books out there have been subject to clinical study and peer review?

Here Kling says he tried it and failed. But if we are to take Ford's argument about science to heart, we should disregard such anecdotal claims (along with the many positive claims posted in the Amazon reviews).

The confirmatory bias at work here may be Kling's favorable clipping from TCSDaily of a story that supports his personal views about the diet. I admit the logic connecting "I failed to lose weight using this diet" and "Spread the word about an article claiming that Roberts should have had his research results peer reviewed" is a bit queer, but that's the confirmatory bias at work.

asg writes:

Aren't you committing almost as serious a sin as Roberts', by passing judgment on the diet before it has been scientifically tested? His sample size was 1; yours is 2. That doesn't suggest a much higher reliability for your conclusions than for his.

bartman writes:

It worked for me: 20 lbs in two months.

Tim Lundeen writes:

I read Robert's paper when it was linked to by Marginal Revolution and tried it, and it worked as advertised. But then I moved and dropped out of the diet for a while in the course of all the perturbations involved. When I went back to try it again, it wasn't working. I realized that my diet had changed in the the interim -- I had been eating a relatively low glycemic index diet when Robert's diet worked, and had gradually added more and more high-glycemic index foods to my diet (candy, cake, ice cream, chocolate, breads, rice, potatoes). When I went back to a combination of a low-glycemic-index diet and Robert's set-point lowering strategy, it again worked fine.

Roberts, if you read his paper, was already on a low-glycemic-index diet when he had his revelation.

The seductive part of his diet strategy is that you can "eat anything". I don't think this is true: I think you can overwhelm the negative bias from the unflavored calories by having too much high-GI food. I don't know if your weight set-point is affected by the glucose/insulin surges that high-gi diets provide, or if the flavor theory is correct and the flavor-calory combination of the high-gi foods strongly offsets the negative push from the unflavored calories.

So, I'd suggest combining a low-GI diet and Robert's unflavored calories.

I've lost about 25 effortless pounds since I found this combo, without ever being hungry. I still have some craving for carbohydrates, but not as strong as before :-)

Roberts is a scientist, and his paper was quite sound. The book makes somewhat larger claims than the paper, and I don't think some of them are warranted given my and other's experience. But Roberts does talk about low-GI diets in his book as a way to lose extra weight. Low-GI diets are a good thing anyway, regular strong glucose-insulin surges are not healthy on a long-term basis.

balibari writes:

I think you've explained my cigarette diet. I know that smoking increases your metabolism, but I was beginning to wander if it was working too well; and wondered why more people don't take up smoking. (I know, smoking is bad, and will lead to cancer, but obesity is unhealthier, and some diets are pretty silly as well) I've been reducing my calorie intake to somewhat ridiculous levels these past 5 years. 1 meal every other day this year. I don't think it's possible to eat less. And I'm ok, I think.

Robert Speirs writes:

One meal every other day?? Why not try the Eating Man's Diet? You eat 900 calories one day and a regular "diet" - 2700 calories if you're shooting for 160 pounds, for instance - on the next day. Repeat until thin. Sharkey's book was published in 1969. It's been out of print for decades. I've read hundreds of diet books and this is the only one that makes sense. And so far, in two weeks, I've lost four pounds. We shall see. But, darn, it lets you eat.

Michael Stack writes:

I forgot where I read this, but I think it was on MR. Anyway, an economist thinks that the way most of these diets work is that they place some strange restriction on what it is that you're allowed to eat. So even though it seems as though you can eat many different foods, in reality, you're quite limited to what you can eat until you learn to "game" the diet. Until that point, you lose weight through good 'ole calorie restriction, though the dieter often isn't aware that his caloric intake has declined.

imsovain writes:

I find Tim Lundeen's comment about the diet only being effective on low-GI diets very interesting. I've been on the diet for 2 weeks, and I found the appetite suppressing effect of the oil to be immediate and obvious. I eat a fairly low GI diet.

david writes:

Cutting out 150-200 calories per day will help you lose weight slowly and safely, but it won't bring commercial success. I lost forty pounds over two years and have kept it off for a couple years since. Everyone says I look great, but no one wants to know how I did it. They want to know how to lose ten pounds in a week.

Seth Roberts writes:

Confirmatory bias? No. As I describe in a paper about my self-experimentation (, I tried six different ways of losing weight. The final one -- on which my book is based -- worked far better than the previous five. I was stunned how well it worked -- there was no precedent anywhere for such an effect. My theory suggested it might work, sure, but how well it worked is essentially a matter of parameters and my theory was not quantitative. This does not sound like confirmatory bias to me. Confirmatory bias was equal for the six methods I tried, so it cannot explain differences between them. One of the six methods I tried worked only for a few weeks, so apparently confirmatory bias cannot produce long-term change. I have kept off the weight I lost with the final method for almost six years.

Mike Wells writes:

"I was foolish enough to buy the book and try the diet."

Apprently also foolish enough to buy it and try it without reading it. The diet has nothing to do with consciously reducing caloric intake and somehow trying to motivate one's self to stick with it. It involves adding calories in a specific way that is supposed to reduce appetite so that your net calories decrease without conscious effort.

There are many possible explanations as to why the diet didn't work for you. What's harder to explain is why it's working so well for so many people who approached it with similar skepticism.

alcibiades writes:

So how do we raise the psychological cost of (over)eating? Personally, I don't keep any solid food in my fridge. A richer person might hire a personal chef who cooks three square meals. (And beats him mercilessly if he tries to run off to McDonalds?) A better electrician than I might create a programmable refrigerator.

Jim McCloskey writes:

Sigh. So the diet is working for me. Weight down (about 2 lbs. Per week), blood glucose down, blood pressure down. What am I supposed to think about someone that says "It doesn't work for me so it must be a fraud?" This is the single most depressing post I have seen from you.
I have no problem with you saying it doesn't work for you. You look trim in your photos. For a fatty like me losing 2" on the waist line is important. Having a regimen that is easy to keep up with is priceless.
Regarding the TCS article. If this was a weird diet I might agree for testing Maybe the big sugar-water or big lite-tasting-olive-oil consortium would fund a study. Since they don't seem to be awake what should we do? Wait? Taking two tablespoons of oil or sugar water twice a day is pretty painless. If it worked for 10% of the people that tried it would be worthwhile. Depressing....

Eric H writes:

Michael Stack wrote, "I forgot where I read this, but I think it was on MR."

Yes, it was. It's called the Alphabet Diet. The main trick is that as soon as you learn to game it, you should switch to another diet. It's too simple to justify a book written about it (and lots of co-merchandising), so it's not well known.

If you're going to call me an economist, however, you should probably prepend "amateur" to that title ;~)

Well, any diet works for about 2-3 weeks, and holds on for another week. It does that by inducing change in what you are eating. It is why the general fraud of the diet book industry has been so successful.

Having followed the diet and the developing ancedotal evidence (and yes, I tutored advanced statistics and know that ancedotal ~ wrong isn't far off most of the time), for people who are morbidly obese, the diet seems to kick in immediately. For people close to their optimum weight, it seems to take 2-3 weeks before there is any effect at all.

I'd be curious how long you tried the diet, what your ht/wt factors are, what method you used and the doses, and how much tweaking you did.

I tried it in November of 2005. The mental changes in how I related to food were immediate and powerful. I had no expectation or experience of anything like that. I'd also tried a number of diets and never had my metabolism let me lose more than twenty pounds without shutting down.

I'm currently down 57 pounds. I find it frankly amazing. I've had a number of people I interact with try the diet with immediate success. I've given away two bulk orders of the book to friends (I e-mailed Seth before the book came out, trying to send him some money via pay-pal to say thanks. He turned me down, so I've bought books and given them away instead).

What is fun is just how fast the effect (I'm not sure diet is the right term for the SLD method) kicks in. Most of the people I've suggested the method to have felt the effects within a few hours.

It is also interesting that the book is nice, but really surplusage to many.

Anyway, I'd be interested in more data on your experience, not to mention an analysis of how this differs from the typical diet which relies on the change in diet effect to generate a temporary effect. is an excellent post to read just to get a flavor of the way the diet industry should be seen.

As for the weight loss surgery industry, well, the reason I skipped surgery is that the operation to complete return of lost weight time frame is seven years.

Should be interesting to see how it all goes.

Drop me an e-mail with the missing data, I'm really curious.

I was hoping to hear back on my questions, I was actually curious about the details. Failures often teach more than successes.

Comments for this entry have been closed
Return to top