Recognizing vs Generating

[A dichotomy involving the difference between checking/verifying/recognizing and generating/coming-up-with-it-yourself. Several examples, with the bulk of the essay applying this dichotomy to advice. Cannibalizes and extends the On Giving Advice essay.]

Here’s a useful dichotomy that I think gets overlooked quite often. It’s the distinction between Recognizing and Generating. The words by themselves don’t explain too much, so here are some examples:

Say you’re writing an essay. You start off with a bullet point outline where you jot down some of your half-formed ideas. Then, you use the outline to write your actual essay, expanding upon the bullet points titled things like “Intro Stuff”, “Ending Hook”, and “Insert stuff about their connection” into actual sentences and paragraphs.

Now, you hand the outline and the essay to your friend. “Can you identify which parts of this outline map onto the essay itself?” you ask them.

It’s not too difficult of a task. With the outline in hand, your friend can probably break up the essay into rough chunks where you expanded upon the bullet points.

But, now, say that you only hand the essay to your friend “Can you use this essay to regenerate the outline that I originally used for this essay?” you ask them.

Now, it’s a lot harder. It’s not just working backward or doing things in a clear cut manner. Your friend could come up with lots of plausible outlines, but there’s only one of them that is the one you used.

In the first scenario, your friend merely needs to recognize which parts of the essay map onto the outline. In the second scenario, your friend has to generate the entire outline.

This, in a nutshell, is the core distinction in the Recognize vs Generate dichotomy.

Another, perhaps simpler example, is checking vs coming up with math proofs. It’s a lot easier to verify that a proof is correct than to have the ingenuity to come up with the proof in the first place.

I also think this is why math ends up being a difficult subject for lots of people. It’s all too easy to fool yourself that you understand a certain subject when you’re just following along with the textbook. Cover up the steps, though, and try to solve the problem yourself, and things get a lot harder.

In a similar vein, when you’re reading any sort of book and the author says, “Now, obviously, we can conclude X,” you should take a step back and actually think. If the author hadn’t presented you with the conclusion, could you have generated it with the pieces you already had?

Lastly, a more general example of this Recognize vs Generate distinction has to do with coming up with ideas. If you’ve ever been in a group discussion, you’ve probably never had a shortage of people willing to critique ideas that are proposed. But having a steady stream of ideas proves to be harder.

Once again, I think the distinction is at work here. It’s far more easier to point out why an idea is bad (after all, basically every plan will have drawbacks!) than to come up with ideas in the first place. This isn’t to say that criticism isn’t important, but I do think it’s important to acknowledge this asymmetry in effort that’s being put in by both sides.

I think that one of the most useful ways to use this distinction is when people give advice.

I claim that basically everyone who gives advice is doing it wrong.

The tldr; is that I think that pretty much all advice people give falls into the Recognizable category, when they should really be thinking in a more Generating style.

Here’s what I think roughly goes into people’s heads when they try to give advice into another human:

  1. Think of several experiences where things turned out well for them.
    EX: “That time I won the free laptop, that time I scored the awesome job, and that time that I asked Andy to the dance.”
  1. Try to find some common threads between all of the experiences.
    EX: “Well, I did always start with a good impression, I always introduced myself well in all of those times…”
  1. Distill it into something general. Give it out as advice.
    EX: “Remember to always start with a strong handshake!”

So your experienced friend, hoping to impart some worldly advice, tells you to always start with a strong handshake. And that advice promptly gets filed under your “Boring Advice” drawer in your head, never to be seen again.

I think that the core problem here is that your friend hasn’t done a good job of accurately conveying the reasons for the advice given.

From inside their heads, they have access to all of their experiences, so once they’ve got the generalized piece of advice—”Remember to always start with a strong handshake!”—they can check it against their memories to verify if it’s a good match.

They’re able to quickly recognize that the advice checks out.

But, for you, the advice-getter, you don’t have access to any of those experiences.

So the best you, the advice-getter, can do is imagine a few situations where the advice kinda works out, and it’s not at all as convincing and obviously “good” as the advice-giver thinks it is.

I think that advice becomes far better when you start taking a Generative approach towards things.

Now, instead of looking for common threads, you ask yourself, “What information would I have needed in those situations in order to have come up with those positive actions in the first place?”

Rather than checking to see if your outcomes match the advice you give, you’re now actually checking to see if the advice you give even leads to those outcomes in the first place.

Which is great because it better aligns your thinking with that of the advice-getter’s. It means you’re thinking about what they do and don’t know. Which means you’re less likely to fall prey to the illusion of transparency, the cognitive bias where we assume that everything we know is also known by the other party.

There’s a certain sense here where you’re modeling humans as complicated input-output machines, and you’re trying to see what sort of input would generate the best outputs.

If we turn this back to the handshake example, an attempt to be more Generative might look like this:

Your helpful friend tells you, “Hey, so whenever you meet someone, you should definitely check out the hills and valley of their hand. It’s totally groovy!”

Now, your curiosity is piqued. The next time you meet another human, you squeeze their hand, trying to figure out where the hills and valleys are. To your surprise, they respond, “Hey, nice handshake!” and the rest of the meeting goes swimmingly well.

As a result, you think to yourself, “Gee, it turns out that handshakes are pretty great when they’re firm!”

This is undoubtedly a silly toy example.

But I hope it illustrates how a large part of advice is letting the other person figure out the actual advice for themselves.

Of course, note that the dichotomy goes both ways: If you’re receiving advice from someone that your brain is quickly labeling as “obvious”, it’s a good idea to do a check and see if you could have come up with it yourself if they hadn’t told you.

And I don’t mean to imply that this model is the only or best way of giving advice.

There’s other ways , from going meta (EX: acknowledging that the advice sounds boring and/or obvious), finding ways to quickly shift intuitions (EX: telling a memorable anecdote), to bridging the differences in experience (EX: give more background information).

But I think that the Recognize vs Generate distinction is still a highly useful one that can find itself into many more discussions. It’s got a lot of use, and I hope that this essay’s illustrated several uses.

In general, taking note of this dichotomy means we want to avoid the surface-level similarities of what, at their very core, are quite different activities (i.e. recognizing vs generating).

The obvious takeaway is that the effort required for the two tasks is very different, despite their apparent similarities. But I think another useful takeaway is that it makes it more explicit that an input-output model can yield very good results, like how we want to judge people based on the information available to them rather than the direct results.

Interesting note: The concepts in this essay roughly parallel a dichotomy in machine learning between discriminative and generative algorithms.



Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s