Instrumental Rationality 3: Interlude I

post by lifelonglearner · 2017-10-07T05:22:09.663Z · score: 18 (8 votes) · LW · GW · 5 comments

Contents

  Acting into Uncertainty:

  Fading Novelty:

       A “use it or lose it” phenomenon seems to happen, where either you actually form some new habits as a result of the short-lived excitement, or it falls, forgotten by the wayside.

    1. Going Meta:

    2. Quick Feedback/Incentives/Rewards:

    3. Habituate It:

     4. Contrasting

  Next essay
None
5 comments

[Instrumental Rationality Sequence 3/7].

[Here, we’ll cover two concepts: Acting into Uncertainty and Fading Novelty. They’re both sort of about two (generalizable, I hope) mental feelings (i.e. internal experiences) that can occur when you start trying to act on any of the instrumental rationality techniques.]

++++

Acting into Uncertainty:

[Acting into Uncertainty is about how it can feel scary to get started in environments with incomplete information. It looks at why the feeling of vagueness might be deceptively comforting, and then it dives into explication as a potential solution.]

For many areas of life, I think we shy away from confronting uncertainty and instead flee into the comforting non-falsifiability of vagueness.

Consider these examples:

  1. You want to get things done today. You know that writing things down can help you finish more things. However, it feels aversive to write down what you specifically want to do. So instead, you don’t write things down and instead just keep a hazy notion of “I will do things today”.

  2. You try to make a confidence interval for a prediction where money is on the line. For example: “I am 90% sure Norway has between 100 and 1 billion people”. You notice yourself feeling uncomfortable, no matter what your bounds are. Somehow, it feels bad to set down any number at all, which is accompanied by a dread feeling of finality.

  3. You’re trying to find solutions to a complex, entangled problem. Coming up with specific solutions feels bad because none of them seem to completely solve the problem. So instead you decide to go one level up. You end up just thinking about what properties a good solution should have in the first place and find yourself throwing around buzzwords like "democracy" and "holistic workaround".

In each of the above examples, it feels like we move away from making specific claims because that opens us up to specific criticism. But instead of trying to acknowledge that, we retreat to fuzzily-defined notions that allow us to incorporate any criticism without having to really update.

In other words, there’s a sense in which, in some areas of life, we’re embracing shoddy epistemology (EX: not wanting to validate or falsify our beliefs) because of a fear of being shown wrong.

I think this potential failure is what fuels the badness when we confront uncertainty.

It seems useful to face this feeling of badness or aversion with the understanding that this is what confronting uncertainty feels like. The best action doesn’t always feel comfortable and easy. It can just as easily feel aversive and final.

Look for situations where you might be flinching away from specificity by making vacuous claims that don’t say much at all.

When possible, try to explicate. Be specific.

Explicating and being specific opens up our plans and hypotheses to falsification; it leaves them vulnerable to being affected by evidence. Remaining uncertain means we can’t be shifted either way because we never made a strong statement in the first place.

We want our plans to fall in the face of contrary evidence. We want goals that are actually realistic. A vague goal means that we don’t aren’t required to specify what we actually want to get done, which clearly makes it harder to make progress on them.

Plus, vague goals give you more excuses to wiggle out of your own promises:

In my own case, there’s a secret part of me that is aversive to explicating; it wants to stay in the vagueness. On some level, I think that if I just under-specify what I’ll get done for today, then that leaves open the possibility that I’ll be able to somehow get all my work done. But things don’t actually work like that. It’s important, then, to try and decouple wishes from predictions.

So there's two things here:

One is about how, sometimes, the best action is actually still the one that feels uncertain and scary because confronting uncertainty is just a feature of our existence.

And the other thing is about how being specific can be a good way to solve part of the problem.

I claim that it’s the atomic, mechanistic actions which lead to things getting acted on. The real world runs on specifics—reality always has a Next Action. Thus, when we explicate the felt meaning in our heads, we’re also converting the nebulous feeling into a format that’s more workable in reality.

++++

Fading Novelty:

[Fading Novelty is about how the excitement of stuff can wear off after a while. This can make it harder to learn new stuff after you’ve been exposed to it for a while, and it seems like part of why Obvious stuff gets discarded by our brains.]

Male mammals tend to exhibit a frenzy of mating when first introduced to a female. After some time, they lose interest. Until a new female is introduced, that is, whereupon we tend to see renewed interest from the male.

This phenomenon is dubbed the Coolidge effect.

I find that the Coolidge effect seems analogous to the idea of fading novelty—the biological definition, that is—which is where something new eventually loses its special sheen.

For example, say Carrie gets a new plush cat. She looks at it on her bedside, and it has this sort of "glow" that makes it stand out compared to all her other things. Over time, though, her cat plush fades into the background and it no longer feels special.

I think this is a fairly universal feeling, despite there appearing to be very little about the high-level phenomena in the literature. (Which is why it’s here in an Interlude essay, rather than in a more well-researched essay. )

Other related ideas in this space include conditioning, tolerance, and acclimation; basically, situations where what was once a stressor no longer really elicits much of a response.

I’m interested in looking into fading novelty because it seems like part of the pedagogical problem with learning rationality goes something like this:

Alice learns about back-planning as a new planning skill. Empowered, she starts seeing ways to apply this idea everywhere. Armed with her new hammer, she makes some headway; progress is happening!

Soon, though, realizes that the back-planning idea now feels merely commonplace in her mind. The original feeling of “wow” has faded, and it feels less yummy to keep using the strategy because it's no longer exciting. She stops using it, and there's a general loss of excitement that used to be there.

As a result, I think that when we learn new insights, there is only a small window of time to capitalize on the novelty factor before it starts to feel boring.

A “use it or lose it” phenomenon seems to happen, where either you actually form some new habits as a result of the short-lived excitement, or it falls, forgotten by the wayside.

One reason could be because the novelty of the insight has faded, making it seem less exciting to use.

Now, to be clear, there are obvious reasons for wanting to keep fading novelty in humans:

Fading novelty is our first line of defense against getting stuck in loops. If repeated exposure to the same stimuli in normal contexts always triggered the same response, we’d likely get caught in repetitive actions where we wouldn’t feel incentivized to go off in the world and explore.

Secondly, there’d likely be sensory overload. We’d likely be overwhelmed with the novelty of everything all the time, which would undoubtedly make it far harder to focus on the important things.

However, I think it’s important to at least acknowledge that whenever learning rationality, which is insight-based, fading novelty can reduce the “yumminess” we initially feel towards practicing rationality techniques.

I also think it might be useful to have a few ways to, if not disable, but at least somewhat counter the fading novelty for things that we want to keep feeling new and exciting.

Here are some ideas I’ve brainstormed, along with some examples: (Note that the ideas below all sort of skirt around creating new novelty and don’t exactly give a good solution.)

1. Going Meta:

I touched on this in the end of the In Defense of the Obvious essay, but this basically consists of noticing your lack of enthusiasm after the novelty fades and knowing that it was going to happen like this. I don’t think this brings back the sheen of novelty, but it feels related, so I included it here.

EX: Steve signed up for a difficult calculus course. After the initial excite of manipulating arcane symbols fades, Steve realizes the class is a lot of work. Still, he knew this when he signed up, so he's able to gather back some of his initial fire by knowing everything is All Part Of The Plan.

2. Quick Feedback/Incentives/Rewards:

I also think there’s a sense in where, if the action you’re doing produces some sort of reward/incentive, you’ll probably also feel compelled to do it, in a sense independent of novelty. Think checking Facebook, which keeps you craving that delicious red number hanging on the right edge of the globe icon and how satisfying it feels to click it, over and over, time and time again. We'll come back to this a little more in section 4 about habits.

EX: When I was practicing coin magic in front of a mirror, getting instant visual feedback on my sleight of hand was immediately rewarding, which kept me practicing, even when the novelty of the trick itself faded.

3. Habituate It:

Obviously if you’ve managed to turn the task into a habit, then you don’t need to worry about all this “cultivating novelty” stuff. You’ll just end up…doing it. The next section, Habits 101, will go into far more detail.

EX: Turning journaling into a daily habit so I don’t need to rely on the motivation boost from novelty.

4. Contrasting

Humans are pretty relative. We compare things with regards to our immediate past as reference points. The idea here is that you'd try to reset your "zero-points" for different things by alternating between ascetic and normal states to improve appreciation of the basic stuff.

EX: Deliberately not thinking much for a few days to improve appreciation of thinking. Or, deliberately fasting to improve the taste of food.

Next essay

5 comments

Comments sorted by top scores.

comment by Raemon · 2017-10-07T06:51:04.522Z · score: 15 (5 votes) · LW · GW

I'm glad that the series is getting reposted here, but one thing that sticks out, having read your more recent "see what happens if you try channeling Scott Alexander" essay, is that I think if you revisited the concepts here while applying some skills you explored there, the series would stand out more.

A recurring problem (first acknowledged in 'In Defense of the Obvious' and sort of followed up with here with the novelty-fading thing, is that the skills you're talking about here are mostly composed of obvious-ish things that, while good, don't give you that insight-porn quality that a lot of early LW stuff had.

A potential solution to that is to make the examples more exciting. Open with a compelling/cute anecdote that gives the reader an initial "ooh!" kick, even if the underlying concept you want to focus on feels a bit obvious, then segue into the main point, and I think ideally ending on a note that forces the reader to think about how to apply this to themselves.

This may be beyond scope for the current re-posting - I think coming up with the right anecdotes and examples is a big chunk of work, but something to at least bear in mind for whatever the next iteration of this project is.

comment by lifelonglearner · 2017-10-07T14:04:17.453Z · score: 13 (4 votes) · LW · GW

This is a good example of a concrete change I could make which would likely make these essays a lot more interesting to people (in terms of readability). Thank you for pointing it out.

I agree that it would take more work than the minor cleaning up / formatting I've been doing, and I also think it could be very good in terms of generating more interest.

I will commit to spending at least 30 minutes thinking more about how I might be able to channel some of that into the existing posts as well as the last 4 essays.

comment by lifelonglearner · 2017-10-07T05:25:05.474Z · score: 7 (2 votes) · LW · GW

I think this maybe one of the weaker essays in the sequence, even though the two things covered are (I claim) important.

The next post in the sequence, Habits 101, is going to be very long (~9,000 words), so it might not show up on time because of formatting / image embed things.

comment by Gunnar_Zarncke · 2017-10-07T15:18:38.855Z · score: 4 (2 votes) · LW · GW

One other technique for countering Fading Novelty I use is to turn insights into Anki cards. Anki's spaced repetition logic ensures that you will be reminded of this insight exactly around the time when you are about to forget it and thus either you just refresh it or it will feel new. In the latter case the excitement will also come back and you get a chance for a new start.

comment by StefanDeYoung · 2017-11-09T19:25:12.289Z · score: 2 (1 votes) · LW · GW

I have also seen other users post about using Anki cards to remember insights from LW. However, I've had difficulty with formulating good flashcards related to this material.

Right now, I have a card for the Litany of Tarski. On one side is the litany ("If the sky is blue...") on the other side, "Litany of Tarski." When I see the card, I try to recite the litany in that form, but also consider the underlying idea that there is a territory to be mapped, and that the map is supposed to reflect the territory. I might also create a new litany with some other object than the blue sky.

Is this the kind of card that you create? Can you give an example of how you use a card to remind yourself of insights rather than definitions?