The Ben Franklin Method: A Deliberate Practice case study
The Ben Franklin Method
We all know that the way to increase skill is “deliberate practice”. However, what exactly is “deliberate practice” and how do you do it? Practice seems straightforward enough in the area of music or chess or sports and, in fact, is what most books on practice focus on. But how can I increase my skill at programming? Or at thinking scientifically? Let’s taboo “deliberate practice”.
Specifically, let’s extract some lessons from Ben Franklin’s famous practice of sentence structure, as described in Talent is Overrated.
First, he found examples of prose clearly superior to anything he could produce, a bound volume of the Spectator, the great English periodical written by Joseph Addison and Richard Steele. Any of us might have done something similar. But Franklin then embarked on a remarkable program that few of us would ever have thought of. It began with his reading a Spectator article and making brief notes on the meaning of each sentence; a few days later he would take up the notes and try to express the meaning of each sentence in his own words. When done, he compared his essay with the original, “discovered some of my faults, and corrected them.”
– Chapter Seven, Talent is Overrated
What was he doing? Let’s follow along by replicating his experiment. The role of the Spectator article will be played by Paul Graham’s essay Schlep Blindness and the role of the young Ben Franklin will be played by yours truly.
Here are the opening three paragraphs from PG’s essay:
There are great startup ideas lying around unexploited right under our noses. One reason we don’t see them is a phenomenon I call schlep blindness. Schlep was originally a Yiddish word but has passed into general use in the US. It means a tedious, unpleasant task.
No one likes schleps, but hackers especially dislike them. Most hackers who start startups wish they could do it by just writing some clever software, putting it on a server somewhere, and watching the money roll in - without ever having to talk to users, or negotiate with other companies, or deal with other people’s broken code. Maybe that’s possible, but I haven’t seen it.
One of the many things we do at Y Combinator is teach hackers about the inevitability of schleps. No, you can’t start a startup by just writing code. I remember going through this realization myself. There was a point in 1995 when I was still trying to convince myself I could start a company by just writing code. But I soon learned from experience that schleps are not merely inevitable, but pretty much what business consists of. A company is defined by the schleps it will undertake. And schleps should be dealt with the same way you’d deal with a cold swimming pool: just jump in. Which is not to say you should seek out unpleasant work per se, but that you should never shrink from it if it’s on the path to something great.
– Schlep Blindness, Paul Graham
My “brief notes” on each sentence in the third paragraph:
schleps inevitable
just writing code - not enough
PG made that mistake
business mostly defined by its schleps
deal with it as soon as possible
don’t seek it
don’t shrink
And, based on those notes, here is my execrable rendition:
Schleps are inevitable in business. Most programmers think that just writing code is enough. I myself made that mistake in 1995. It took me a while to realize that a business is mostly defined by the schleps it will undertake. So, dive into them without waiting for the right time. However, while you shouldn’t seek out schleps, you shouldn’t shrink from them either.
I didn’t spend too much time writing that, but yeah, I literally turned my eyes away from the screen when I compared my paragraph to his. Not a fair fight!
Anyway, we need to compare each sentence to the original like Ben Franklin, but what does it mean to “compare”? If it were just that PG used one strong word in place of the many weak words I used, like “decide” instead of “make a decision” or “conclude” instead of “come to a conclusion”, it would be easy. But the sentences differ in all kinds of ways.
In sports, it can be pretty easy to compare two performances because we have dimensions like time taken or height jumped or run rate. For example, Usain Bolt ran faster than whoever else ran with him. Dhoni had a higher run rate than Dravid. That tells you exactly where you need to improve.
Given that writing doesn’t seem to have such objectively-measurable dimensions, I’ll use the first 10 tips from this writing advice book called 50 Writing Tools by Roy Peter Clark. To a first approximation, the more of them you use, the better your writing. (You don’t have to understand these rules right away, just know that the book describes them at length.)
Writing Tool #1: Branch to the Right
Begin sentences with subjects and verbs, letting subordinate elements branch to the right. Even a long, long sentence can be clear and powerful when the subject and verb make meaning early.
Writing Tool #2: Use Strong Verbs
Use verbs in their strongest form, the simple present or past. Strong verbs create action, save words, and reveal the players.
Writing Tool #3: Beware of Adverbs
Beware of adverbs. They can dilute the meaning of the verb or repeat it.
Writing Tool #4: Period As a Stop Sign
Place strong words at the beginning of sentences and paragraphs, and at the end. The period acts as a stop sign. Any word next to the period says, “Look at me.”
Writing Tool #5: Observe Word Territory
Observe “word territory.” Give key words their space. Do not repeat a distinctive word unless you intend a specific effect.
Writing Tool #6: Play with Words
Play with words, even in serious stories. Choose words the average writer avoids but the average reader understands.
Writing Tool #7: Dig for the Concrete and Specific
Always get the name of the dog.
Writing Tool #8: Seek Original Images
Seek original images. Make word lists, free-associate, be surprised by language. Reject cliches and “first-level creativity.”
Writing Tool #9: Prefer Simple to Technical
Prefer the simple to the technical: shorter words and paragraphs at the points of greatest complexity.
Writing Tool #10: Recognize the Roots of Stories
Recognize the mythic, symbolic, and poetic. Be aware (and beware) that common themes of news writing have deep roots in the culture of storytelling.
I’ll admit it’s hard for me to tell for sure in some cases whether someone is following a particular rule (more on this difficulty later). For now, let me measure these dimensions as best as I can.
Here we go:
PG: One of the many things we do at Y Combinator is teach hackers about the inevitability of schleps.
Me: Schleps are inevitable in business.
Rule #1: I think he branched to the right since the subject was “we (at Y Combinator)” and the verb was “do”. I don’t think “Schleps are” counts as subject and verb. (See how unsure I am about the dimension itself?)
Rule #2: He used the strong verb “teach”. I used “are” (I don’t think that’s even a proper verb). Strike two for me.
Rule #3: I didn’t use adverbs. Phew. No strike.
Rule #4: I used “schleps” at the beginning of the sentence and he used it at the end. I think the effect of using it at the end is stronger. (See what I did there?) So, this isn’t a complete strike for me; that would be if I used “schlep” somewhere in the middle. Make it half a strike.
Rule #5: I find “word territory” hard to measure. Moving on.
Rule #6: PG played with the novel word “schlep”. I just copied him. I don’t think I would have used that word if I were writing the essay by myself. Strike.
Rule #7: He spoke about the concrete things they did at Y Combinator. I made a vague statement about “business”. Strike number whatever.
Rule #8: “Seek original images” doesn’t apply here, as far as I can tell.
Rule #9: “Prefer simple to technical” - Wow. Again, I can’t tell whether this applies. I need to read that writing advice book again.
Rule #10: He seems to be describing a story (“One of the many things we do at Y Combinator”). I’m making an abstract statement (“Schleps are inevitable in business”). Jesus. How many things can you do wrong? And the book has another 40 rules!
People have told me I’m a pretty good writer, but it takes just one sentence for PG to outclass me. The sad tale continues:
PG: No, you can’t start a startup by just writing code.
Me: Most programmers think that just writing code is enough.
Rule #1: We both branched to the right with the subject and verb.
Rule #2: He used the strong verb “start a startup” whereas I talked about “think”. Hell, I didn’t even mention what it is that “just writing code” is not enough for. Strike one.
Rule #3: No adverbs, thankfully.
Rule #4: He ended on “code”. I ended on “enough”. ’nuff said.
Rule #5: No clue how to observe “word territory”. Skip.
Rule #6: I don’t think either of us played with words here.
Rule #7: I think he spoke directly to the reader to evoke images of people who insisted that they could start a startup by just writing code. I wrote about the abstract set of “most programmers”. Man, that is such bad writing. And the contrast makes it look even worse. Not fair!
Rule #8: No images for either of us.
Rule #9: Both of us kept it simple.
Rule #10: He’s kind of telling a story about wrong-headed programmers starting doomed startups. I’m not.
I won’t torture you (or myself) with the full analysis of the other sentences, but here a couple of interesting rules that I violated but he didn’t:
PG: And schleps should be dealt with the same way you’d deal with a cold swimming pool: just jump in.
Me: Dive into them without waiting for the right time.
Rule #8: Seek original images - diving into a cold swimming pool isn’t original, but it’s unique for describing schleps. I didn’t even bother constructing an image.
PG: Which is not to say you should seek out unpleasant work per se, but that you should never shrink from it if it’s on the path to something great.
Me: However, while you shouldn’t seek out schleps, you shouldn’t shrink from them either.
Rule #4: Period as a stop sign - he ended on “great”; I ended on “either”. FML.
Man, that was tiring (and painful). Measuring ten dimensions for just six sentences left me pooped. (Did somebody say “[deliberate practice] is highly demanding mentally”?1)
The Skill Ladder
It’s pretty obvious by now that I suck. But now, thanks to deliberate practice, I know exactly how much.
So, what can I say about “deliberate practice” based on the above study?
Skill refers to the quality of your response to a given situation. So, the “brief notes” I made about each sentence represented the thought I wanted to convey, i.e., the input situation. The sentence I wrote was my response.
The 10 rules above helped me grade the quality of my response along 10 dimensions. We can think of a ladder of quality, with the lowest rung representing a sentence that satisfies none of the rules, the highest rung a sentence that satisfies all 10 rules, and the intermediate ones representing sentences that missed one or more rules. Of course, some rules probably matter more than others, so being concrete and specific (rule #7) may help little if you bury your subject and verb in the middle of your sentence (violating rule #1) whereas things may go better the other way around. I will need such dimensions for whatever skill I want to practice.
In fact, so crucial are these rules that you need to practice the skill of observing those dimensions automatically. You should know immediately when you violate a rule. This is what Daniel Coyle, in his book The Talent Code, called “learning to feel it”. Remember how I completely failed to recognize whether I had obeyed rule #5 and rule #9. That means I can’t tell, along that dimension, if my writing is good or bad (probably bad). Before I can practice the skill of writing, I must learn to tell if a paragraph “observes word territory” or “prefers simple to technical”. This goes for other skills. For example, cricket batsmen need to know when they are hitting the ball in the middle of the bat. Otherwise, they may just nick one to the slips. That’s why you see them shadow batting after playing poorly. They’re making sure that they don’t slip down the ladder because of lack of foot movement or overhitting.
So, these rules help me evaluate my writing and give me feedback. Then, why do I need sentences from PG? His sentences provide concrete examples of the higher (maybe highest) rungs of the ladder. That way, I can see firsthand how to express the same idea with high quality.
However, I suspect having terrific examples can be a curse when trying to practice because you feel like a complete loser there are too many things being changed at the same time. Recall how I kept remarking on how many things he got right that I got wrong. To write like PG, I would have to, at minimum, obey seven or more rules than I currently do. How do I leap from my current level to a rung seven levels above?
Imagine a writer who sucked just as much as I did but did one thing better, such as ending on a strong word. I might write “Place emphatic words at the end of the sentence” whereas he would write “Place emphatic words in a sentence at the end”2. See how easy it is to tell the difference? The only change is “at the end of the sentence” instead of “in a sentence at the end” and I now know that the latter is better because the strong word here is not “sentence” but “end”. If I do that for enough sentences, then somehow, through the magic of repetition, I will notice immediately if I ever write a sentence with a poor ending and I will rewrite it to end better.
So, what we need for deliberate practice is not just better examples, but examples that are just a few levels above ours (ideally, one level above).
Of course, if you have an algorithm - a fixed procedure - for generating sentences that obey all 50 rules no matter what the input idea, then you don’t need any of this work. You can focus on learning how to use that algorithm and, within a short while, reach mastery. That is what I believe we do for long division. We don’t start with a response that gets just one of the digits right (like “3550 / 62 = 873”, where only the 7 is right) and improve till we fix the other digits (to get “3550 / 62 = 572”). We start with a procedure that is guaranteed to get every digit right as long as we correctly apply each step. While you may work more on your speed, your answers are guaranteed from the start to be correct. Real-world skills, however, are more “complex” than long division, which is to say, we haven’t yet figured out their algorithms. Until we do, we will have to use deliberate practice to climb up the skill ladder one rung at a time.
Notice how PG satisfied a ton of rules in the same sentence. That’s mastery. His response to an idea like “just deal with schleps as soon as possible” comes out in the form of a vivid sentence “And schleps should be dealt with the same way you’d deal with a cold swimming pool: just jump in.” I don’t think he’s ticking off the rules one by one. His “instinctive” response is probably one that obeys a lot of rules in one shot. He has gone up the ladder over time and eliminated bad writing habits. Similarly, at my level of skill at programming, I immediately frown when I see a long, messy function. I have gone up enough rungs on that ladder to notice “code smells”, i.e., violations of the rules of good programming.
Finally, notice that I needed input situations, like “schleps - deal with them as soon as possible”, so that I could evaluate my responses. Great skill needs great responses to a lot of different situations. A master batsman needs to play well against bouncers, off-spin, and yorkers. Similarly, when writing, I need to express sentences that introduce a new idea or compare an idea against another one or describe an involved procedure. I need to find examples of each type, get the input situation or idea in the form of brief notes, write my version, evaluate it along various dimensions, and compare against a slightly better version to see how I could have improved. Repetition on such a variety of inputs is what I believe takes the much-vaunted 10,000 hours. The more the types of situations, the more you will have to practice to reach the top rung of the skill ladder in each situation.
Note that this means there’s no such thing as overall skill. It depends on the situations for which you have practiced high-quality responses. Even if you’re a world-class finisher like Dhoni, you will struggle if you come across the swinging balls that openers have to negotiate. And, likewise, great openers will struggle at the death when facing Malinga’s toe-crushers, while Dhoni will pull off a helicopter shot.
Road to 10,000 Hours
The above model of deliberate practice covers all three archetypes that Geoff Colvin puts forth in Talent is Overrated. The music model is when you have a fixed set of input situations, like the parts of a Beethoven song, for which you need high-quality responses along the dimensions of harmony, melody, tempo, etc. (“Were you rushing or were you dragging?”3). The chess model differs in that there is a wide variety of input situations, i.e., chess board positions, to which you need to respond quickly and strategically. Again, you need to compare against people slightly better than you, because the experts might literally be playing 4D chess. The sports model is one where the input situations vary fluidly, like the demands of a T20 chase, and you need to respond well across the dimensions of run rate, chances of getting out, sixes, and so on. In all these cases, the idea is to evaluate your performance across several dimensions (either by yourself or through an external coach), figure out the difference from someone who’s slightly better, and bridge the small gap.
While we’re at it, we can see that this model includes all five elements of deliberate practice:
Deliberate practice is characterized by several elements, each worth examining. It is activity designed specifically to improve performance, often with a teacher’s help; it can be repeated a lot; feedback on results is continuously available; it’s highly demanding mentally, whether the activity is purely intellectual, such as chess or business-related activities, or heavily physical, such as sports; and it isn’t much fun.
– Chapter Seven, Talent is Overrated
The kind of practice I described in this essay is designed to improve performance because it specifically aims to improve along one or two well-defined output dimensions (like ending on something emphatic). Contrast this to the poor practice of just hitting golf balls without measuring key output dimensions like distance covered, timing, and accuracy, or that of writing tons of prose without evaluating each sentence. It can be repeated as many times as the number of input situations you have. For example, I can get a lot of sentences from PG essays and learn where I fall short along the output dimensions. Contrast that to just writing one 3,000-word essay and calling that practice. Feedback on results is continuously available even when it makes your eyes bleed (as we saw above with the ten dimensions). Compare that to how hard it was to evaluate my own writing without the dimensions or without PG’s essay. It’s highly demanding mentally because I started tapping out after one paragraph of intense scrutiny. Sure, you won’t normally go around evaluating ten new dimensions at a time, just the one dimension that you’re currently failing at, but still it was tiring. And you bet your ass it isn’t much fun to hit yourself in the face with how much you suck compared to a brilliant writer.
Great. This seems like a workable blueprint of deliberate practice. So, how do we put it all together to practice programming or scientific thinking? To be continued.
comments powered by Disqus