Alice: Hey honey, I made pasta with tomato sauce!

Bob: Great, let's eat!

Bob: Mmmmm, that's fantastic! It's even better than last time. It's got a sweeter, deeper flavor, which I like.

Alice: Thanks. Last time I only sautéed onions and garlic before adding the tomato puree, but this time I added carrots to the mix for some extra sweetness.

Bob: You know I love your cooking, but I always feel a twinge of skepticism whenever you try to explain why things taste the way they do. Yes, you added carrots to the sauté. But is that the only thing you did differently?

Alice: No. I added some butter along with the olive oil to give it some smoothness. Ummm. I threw some red wine in to the sauce. We didn't have any basil leaves today so I just left that out of the recipe.

Bob: So how do you know that the carrots are what made it taste sweeter? How do you know that it wasn't the butter, the red wine, or the absence of basil leaves?

Alice: Well, the red wine is acidic, so that wouldn't make it taste sweet. I think the butter helped make it taste richer and smoother. The basil leaves add flavor, so leaving them out probably subtracted from the deepness, rather than adding to it.

Bob: Sure, that all sounds plausible. But how can you know any of this?

Alice: It's just common sense. I know that butter is rich. I know that red wine is acidic. I know that basil is flavorful. So I know that adding them will have the corresponding effects on the sauce.

Bob: But isn't it possible that things work differently when you combine them? Sure, red wine is acidic by itself, but isn't it possible that when combined with sautéed onions, garlic and carrots, and olive oil and butter, and fried tomato paste, and whatever spices you used, and cooked. Isn't it possible that after all of that, some chemical reactions occur that cause the red wine to be sweet instead of acidic?

Alice: I don't think so. I think the sweetness came from the carrots. But yeah, I suppose it is possible. So what do you propose I do?

Bob: Cook like a scientist. Only change one variable at a time. Take note of the effect. So in this case, if you hypothesize that adding carrots to the sauté will add sweetness to the sauce, just add carrots to the sauté. Don't add red wine, or anything else.

Alice: Ok, let's suppose I did do that, and I found that carrots did in fact add sweetness. What if I want to add red wine next time. Isn't it possible that carrots add sweetness to the basic sauce, but when combined with red wine, carrots react with the red wine and actually add, I don't know, bitterness to the sauce?

Bob: Yes! You're cooking like a scientist now!

Alice: So I'm supposed to then do another experiment? Make the sauce with red wine, make the sauce with red wine and carrots, and see what the effect of carrots are in that particular scenario?

Bob: Absolutely.

Alice: And then what if I want to use ghee for my fat instead of olive oil?

Bob: More experiments.

Alice: Don't you think that so much experimenting is impractical?

Bob: Maybe. But it's the only way to know the truth. The laws of scientific inference don't care how easy they are to follow. They're laws.

Alice: Maybe. Have you ever heard of bayesian inference?

Bob: Uh yeah, I think so. I think they had a brief section on it in one of my undergrad stats classes. I remember something about it helping solve a game show problem. But we don't use it much in our work at the lab.

Alice: Well the reason I bring it up is because you seem to view inference as a binary thing. Either you conclude that carrots added sweetness to the sauce, or you don't. That's not how I, or other bayesians see things. To me, everything is a spectrum. Of course I can't say that I'm 100% sure that the carrots were the cause of the resulting sweeter flavor, but I will say that I'm about 90% sure. Yes, the fact that I added red wine, butter and removed basil introduce the possibility of confounding, but I feel like I have a pretty good handle on these ingredients given my experience cooking, and my judgement is that it's unlikely that they were the cause of the extra sweetness.

Bob: Sigh. I mean, I guess you're right. But that's just a very unscientific way of doing things.

Alice: And why is that a bad thing?

Bob: Because with that approach, you can never know for certain what the truth is. Like you said, you're only 90% sure that the carrots caused the sweetness, not 100% sure.

Alice: Well I'm ok with that. In a perfect world, I'd do things your way and only manipulate one variable at a time, but that's just not practical in the kitchen. Like you said, every time I change one thing, I'd have to redo the experiment with the carrots. Given how many variables there are, I'd have to do hundreds, maybe even thousands of experiments.

Bob: Again, proper experimentation is the only way to actually know things with certainty.

Alice: Yes, proper experimentation would allow me to boost my confidence from 90% to 100%*. But the cost of this boost would be hundreds of experiments. Thousands of hours of my time. I don't care enough to do that. The return on investment is totally not worth it. When the return on investment isn't worth it, sometimes it makes sense to do science like a chef.

Bob: Well when I cook, I like to do science like a scientist.

Alice: And that's why your tomato sauce stinks.

76

17 comments, sorted by Highlighting new comments since Today at 5:37 PM
New Comment

This reminded me of Perfection Salad.

If the housekeeper could be made to think of herself as a scientist, calmly at work over the beakers and burners in her laboratory, then every meal would emerge as she planned, pristine and invariable. (p. 86)
…an enthusiastic social reformer in the domestic-science movement complained … that “even the intelligent housekeeper still talks about ‘luck with her sponge cake.’ Luck! There is no such word in science, and to make sponge cake is a scientific process!” (p. 86)

I haven’t the time to post at length now, but cooking (and baking) is one of my longtime hobbies, so here are some very quick thoughts—any of which I’ll elaborate on later, if anyone asks.

  • This seems to strawman cooks (in that it leaves out one important epistemic pathway—cf. Christopher Kimball, Serious Eats; leaving it out makes Alice’s position weaker and probabilistically undermines the quality of her tomato sauce)
  • This applies much more to savory cooking than to dessert cooking/baking, and the difference there is analogous to a certain broader class of difference in epistemic approaches
  • What would Alice say to Bob if Bob’s tomato sauce didn’t stink but was great? (Or must Bob’s sauce be bad, with that attitude? A very serious claim…)
  • Implications of differences in flavor perception on points in OP; also, correlation of same with predisposition to certain approaches (does it exist? what does that imply, if it does?)

But good post, overall! We need more of such—upvoted.

What would Alice say to Bob if Bob’s tomato sauce didn’t stink but was great? (Or must Bob’s sauce be bad, with that attitude? A very serious claim…)

I think that Bob's tomato sauce is likely to stink, because it would take too long to iterate if he does things his way. However, it is certainly possible for his sauce to be great, ie. if he stumbles across a good recipe.

So then, even if Bob happened to have great tomato sauce, Alice would demonstrate to him that he was just lucky, and that his way is going to take way too long, such that in practice he will usually fail. Perhaps she could demonstrate this by having a cook-off with a new dish. Although Bob would have to acknowledge that she's right, which is the point that I'm making in this article.

I think that Bob’s tomato sauce is likely to stink, because it would take too long to iterate if he does things his way. However, it is certainly possible for his sauce to be great, ie. if he stumbles across a good recipe.

This is critical: you speak of Bob stumbling—but to the contrary, Bob’s approach allows him to make directed, purposeful movements in recipe-space. Bob’s approach starts out more slowly than Alice’s, but then it accelerates in the rate at which it allows Bob to gain knowledge of the possibilities, because Bob is able to gain (a much deeper) understanding of why recipes work the way that they do.

Here we come to my first point: aggregation of knowledge, being the foundation upon which civilization is built, comes into play here too: Bob need not do all his systematic exploration from scratch, as others have done much of it before him; and what of such Bob does, he can document, and then others who come after him benefit.

So then, even if Bob happened to have great tomato sauce, Alice would demonstrate to him that he was just lucky

And Bob would say: No, Alice, you do me an injustice by attributing my success to luck; I made my great tomato sauce the way that I did because I knew exactly what I was doing.

(And the rest of your comment, I addressed above.)

Good points about directed movements and standing on others' shoulders. After hearing you articulate your thoughts futher, I'm definitely questioning whether cooking was a good example to use.

cough quality of ingredients cough :)

Would you care to be more explicit about the "important epistemic pathway" that's being left out and thus strawmanning cooks? [EDITED to add: oops, didn't see that actually that discussion is elsewhere in the comments. Though I'd say that epistemic pathway isn't left out so much as it's perhaps unrealistically rejected.]

[This comment is no longer endorsed by its author]Reply
This seems to strawman cooks (in that it leaves out one important epistemic pathway—cf. Christopher Kimball, Serious Eats; leaving it out makes Alice’s position weaker and probabilistically undermines the quality of her tomato sauce)

I agree that some cooks can be somewhat "scientific"/Bob-like, but even the more experimental chefs don't really seem to get too close to Bob on the Alice-Bob spectrum. I'm not sure though. Serious Eat's Food Lab can be good. In Search of Perfection (which I've been binge watching recently 😀) is probably as scientific as it gets in the domain of cooking, but even that probably wouldn't fly in academic journals.

Regardless, although some chefs can be Bob-like, I suspect that most chefs are Alice-like, and I think that in using a phrase like "science like a chef", what matters is what most chefs are like (or, arguably, just what the popular perception of what a chef is like). It sounds like you are more familiar with the domain of cooking than I am, so do you think that most chefs are very Bob-like, to the point where "science like a chef" is a strawman? Or just that some minority of chefs are Bob-like, and the phrase misrepresents what they happen to do?

America’s Test Kitchen / Cook’s Illustrated / etc. is what I had in mind, and Serious Eats, yes, as you say. (I invite you to read through The Dessert Bible—my single favorite cookbook of all time—and then tell me that Kimball’s approach is not scientific!)

but even that probably wouldn’t fly in academic journals

You might be surprised—or, perhaps, dismayed—at what would, and wouldn’t, fly in academic journals.

Regardless, although some chefs can be Bob-like, I suspect that most chefs are Alice-like, and I think that in using a phrase like “science like a chef”, what matters is what most chefs are like (or, arguably, just what the popular perception of what a chef is like). It sounds like you are more familiar with the domain of cooking than I am, so do you think that most chefs are very Bob-like, to the point where “science like a chef” is a strawman? Or just that some minority of chefs are Bob-like, and the phrase misrepresents what they happen to do?

A little of the latter, and a lot of something else entirely: most chefs (or at least, most cooks—or did you mean to imply that Alice and Bob are trained professionals? that would cast the claims of the OP in a different light) are not like either Alice or Bob. Most cooks—even those who are very experienced—use a more intuitive and/or more standardized approach than even the one you describe Alice as using.

For example, one of the best cooks I know is my grandmother, who has been cooking for over twice as long as I’ve been alive, and who is a trained pharmacist besides; and yet you would struggle in vain to get from her an explanation of how she prepares any given recipe that is even half as coherent as the one Alice gives in your example. Often it will boil down to “because that’s how it’s done”. (Which is to say, the evolution of the recipe has taken place across generations, not across iterations over one cook’s career.) I would not recommend that you “science like my grandmother” (although you would obviously be quite foolish to dismiss the knowledge gained by that approach—a fact of which I am reminded every time I have a taste of her cooking).

Of those cooks who do explore recipe-space, I think the Bob approach is much more common and more fruitful than you give it credit for—precisely because no one with any sense would think that they must recapitulate the entirety of food science, from scratch, all by their lonesome.

Upvoted. That all makes sense. Thanks for the input! I really question whether this is a good enough analogy to have published the article now.

If Bob's tomato sauce were great, most likely Alice would be saying "... And that's why the only thing you can cook well is tomato sauce, because you've spent every minute you're in the kitchen trying minor variations on your tomato sauce recipe". Of course, Bob might instead have taken a decent recipe someone else made up and not bothered experimenting with it; in that case the right response might be more like "So how come you never tried that with your own tomato sauce recipe? I think it's because deep down you know it's impractical".

Empirically, this is wrong.

It's worth noting that Bob's way of cooking doesn't provide 100% certainty either. Plenty of the science that comes out of that mindset doesn't replicate.

I agree, but I'm not sure that it would be worth talking about that in this particular post, because I feel like it distracts from the core point.

Note that Bob's idea of scientific experiments is pretty limited. Real science does combined-variable tests all the time, and uses tools like regression analysis to actually answer the questions about relationships between the variables. Alice is the better cook AND the better scientist.

So, for optimizing a process with many variables (like tomato sauce), estimate the direction you might improve each variable and move a small amount in that direction, instead of exhaustively testing each variable independently? Because we know that actually works pretty well.

There are some things that Alice does that a gradient descent optimizer doesn't, though, which might also be important. Particularly: she recognizes which variables are likely to affect which features, and she adds a new variable (carrot) from a rather large search space.

I wonder if Alice is vulnerable to a local minimum trap -- she might converge upon pretty good tomato sauce that she can't improve upon, while Bob exhaustively searches for (and might eventually find) a perfect tomato sauce. I agree with the point, though -- if you try Bob's strategy, you'll be eating a lot of bad sauce in the process of exploring all possible ingredient combinations.

Bob strategy doesn't lead to exploring all possible ingredient combinations. The scientific approach of standardizing things generally reduces the search space.

If you look at pharmaceutical interentions you find find often one active ingredient per formula.

On the other hand if you look at the food supplement community you have a lot of formulations that mix a lot more different active ingredients together.