Responding To Overconfidence

Without Googling, what's your 95% confidence interval for the longest time spent in labor? Feel free to post it in the comments before reading on.

I just encountered this question on William MacAskill's Facebook wall. When I looked it up, after posting my answer, I discovered that my upper bound was off by a factor of 10.

I was reluctant to answer the question in the first place, but I didn't stop to examine why. It is now clear to me: When it's revealed that I'm extremely overconfident about something, my default response is shame and regret.

I used to respond much more strongly with shame and regret than I do now. I recall an occurrence of this reaction from about three years ago. The reaction was so strong that many vivid details of the context are readily available in memory. (Whether they're accurate is a separate question.) Robby and I were in the the pizzeria on Kirkwood sitting at a table by the door. It was raining. We were having bread sticks with cheese sauce, and he asked me, "What's your 90% confidence interval for when Reverend Bayes was born?"

Immediately I felt attacked and defensive. I did not know what a confidence interval was at the time, so he spent a few minutes explaining it to me. After that, he wanted me to answer, and I was so scared. I don't remember why, but I remember the feeling very well. I was terrified that I'd be way off, and that this test would reveal my embarrassing mistake. I answered anyway because I'd recently discovered Lesswrong, so I felt that this was a kind of question I was Supposed To answer if I wanted epistemic improvement. (The question was actually taken from a Lesswrong survey.) And I was definitely Supposed To know when Bayes was born!

Sure enough, I was way off. And sure enough, the shame flooded over me like a bucket of ice water. I felt terrible, and I regretted answering the question.

A lot of things have happened in the intervening time to lesson the shame response. I don't know what most of them are, but watching people I respect readily and casually test their predictions and reveal their mistakes has surely been part of it. Curing social anxiety also contributed, obviously.

It's not been enough, though. When I answered the labor question, I still felt enough of the shame to overshadow the recalibration going on beneath it. I did feel the recalibration as well, but it was subtle enough by comparison that if I hadn't been training mindfulness of this sort of mental motion in the recent past, I'd have missed it. And I certainly didn't catch the details.

That's a problem. I can't optimize a process I'm never aware of in real time. It doesn't matter how well I understand Bayesian updating when some other sensation is drowning out all my opportunities to apply my understanding.

You might think, "Ah, but those negative feelings are useful, because if you're punished for being overconfident then you might be less confident in the future!" What actually happens is that I'm less likely to put myself in situations that would reveal my overconfidence if it existed. Which shouldn't be surprising from a behavioral psychology perspective: The immediately preceding action was "answer calibration question", not "form a belief and establish a level of confidence in it".

If I were going to train a better reaction to calibration opportunities properly, I'd spend a few days studying my default reaction and becoming as mindful of it as possible. I'd also examine whether my default reaction suggested an emotional need of some sort that the optimal response ought to address, especially if my reaction were as strong now as it was three years ago. Only then would I begin considering possible interventions.

But this particular reaction seems to be in a class of habits that are very important but whose triggers are much too rare for the current version of the Tortoise Skills installation procedure. So instead, I'm going to try doing an abbreviated version of the procedure in case it turns out that I can get marginal improvements quickly from isolated cases like this.

My best guess at how I'd rather respond to discovering I'm extremely overconfident is about the same as the response I learned to have to failures. I'd like to feel nonchalant interest in my overconfidence. Further, I'd like that interest to inspire targeted curiosity about the cause of the overconfidence, and increased sensitivity to similar contexts or patterns of thought that might signal severe overconfidence if I encounter them when forming or considering other beliefs.

But the most important part is just letting go of the thing that drags me down into counterproductive emotions. A flavor of wu wei, maybe, of fluidity. My brain's pretty good, really, when I can keep from getting in its way. If I can just stop doing the stupid thing, I often don't need a brilliant solution on top of it.

I don't know how I do that "letting go of the dragging-downward" thing, but I do know that I've learned to do it at least once before. I'll plan to imagine Eliezer discovering overconfidence and his usual response, as a reminder that other responses are possible, in case I need some extra help.

So, here's the new trigger-action plan, which I will not train but will instead simply intend and await: If I notice that my overconfidence has been revealed, then I will loosen my grip on the downward-dragging sensations and direct my attention instead to even the tiniest sensation of reflective interest. If I have trouble with that, I will imagine how Eliezer would react in my place.

Brienne's Workflow

I've dramatically improved my workflow over the past month or so. I don't expect this exact formula to work for you, but working is *so* much more fun now that I feel like I've got to share, just in case somebody gets a small part of the benefit from one of these ideas.

Here is the formula I use today. I added each thing in succession, and each one improved my experience of working immediately and obviously.

1. Pomodoros. The pomodoro technique is a work schedule with 25 minute blocks separated by short breaks. I've worked in pomodoros off and on for a couple years, but until now I've used them on an as-needed basis. They've always been good for getting me through highly aversive work, but they've never felt like a boost to projects I don't mind working on. I seem to be pretty good at focusing and not procrastinating by default, so I think I had less to gain from pomodoros than a lot of people. For example, my instinctive response when I first heard about browser extensions that block distracting websites during work periods was "...why don't you just close those tabs?". (I understand why, it's just not how my brain works.) Pomos in a social context, though, have proved powerful for me.

2. The Less Wrong Study Hall is a Tinychat room where people work together silently, webcam broadcasting optional, in extended pomodoros of 32 minutes with 8 minute breaks. Chatting and being social during break times is encouraged, as are bragging about what you accomplished in the past pomo, seeking moral support during difficult projects, encouraging others, and announcing your intentions for the next work block. The password is lw. Social conventions for the room can be found here.

Breaks sometimes get a little silly:

I tried this expecting it not to work. I was really just curious, because I'd never tried anything like it. I'm a very introverted person, and when I had social anxiety I found video chat completely terrifying. I also don't tend to respond well to punishment-based motivation methods, and the original idea behind the LWSH was to create a sense of accountability to others who are watching you work (or something like that).

Instead, it seems to have restored a positive social element that I've been craving since I left college. Maybe I like people after all, but only when they're being quiet, productive, highly predictable, and completely independent of me. Those things happen all the time in libraries, coffee shops, and dorm lounges--at least on college campuses--but I've not encountered the combination in many other contexts.

Hanging out in a video chat room also seems to be breaking down my negative associations with video chat that accumulated during the years of social anxiety. I scheduled a Skype meeting the other day with zero discomfort, and I'm still not worried about it even though it's happening tomorrow. I also participated in a CFAR alum Google hangout a while back, and didn't feel even the tiniest twinge of anxiety.

3. Ambient sounds of rain, thunder, and a coffeeshop (with the voices as indistinct murmurs). I went through a a list of ambient websites trying out each, just because I happened upon the list and was curious, and found Rainy Cafe to be the best for me. It turned out to be on too short a loop, so I've switched to A Soft Murmur, which offers all three types of sounds as options to combine. (I think the loop here might also be too short, and I need to look for more alternatives, or maybe make my own track.) I've tried each of these sounds individually, and the combination works best. Like with the LWSH, I was surprised that the coffee shop sounds were a plus. They seem to increase my sense of comradery in the LWSH. When I shut off the cafe sounds, I suddenly feel slightly lonely, and work becomes less fun.

4. Complice is Malcolm Ocean's productivity startup. It targets a type of person I am not, so it didn't really work for me when I first tried it. I imagine it's great if you have trouble staying focused on your goals, acting in line with your priorities, and not procrastinating. I guess you could say I stand to gain from productivity vitamins, not productivity medicine.

But when he added Less Wrong Study Hall integration, that feature turned out to be a big boost for me. (He plans to make other similar rooms eventually.) It added to my experience of the LWSH more precise and automatic timing, the ability to see what other people are working on and to show others what I'm working on, a simple to-do list, and a visual reminder of how many pomos I've completed. Here's the LWSH as it appears in Complice at the moment.

5. Chocolate at the beginning of each pomo break. I tried a few different kinds, and of those the best for this purpose proved to be After Eight Mint Chocolate Thins. They have a distinctive flavor and texture that I can tie to pomos exclusively (unlike plain milk chocolate, which I'll encounter in many different situations), but I'm pretty sure their superiority is mostly about the packaging. (LOOK HOW PRETTY!!!)

The result of all these things together is that I look forward to getting to work, even if the project isn't all that interesting. I'm usually sad to stop working. Several times I've sat down to complete a single task, and ended up knocking out a bunch of things it would have been fine to do later, just because of the momentum.

Without any of these elements, I usually do about the minimum without a lot of trouble. Procrastination and distraction aren't big problems, but long work periods take a lot out of me, and I'm not motivated to do more than necessary.

My main workflow problem at this point is lunch. I don't want to stop focusing to make or eat it. I've even stated putting "lunch" on my to-do list and staying signed in to the LWSH while I cook, so it feels more like food is part of work.

Tortoise Report 1: Growing the Roses Of Success

This post is part of a year-long project for learning to install habits of thought. For more about the tortoise skills project itself, see the Tortoise Skills Page.

Summary

Habit: Growing the Roses Of Success
Duration: 7 Days
Success: 7/10
Trigger: The very beginning of a trapped, sinking sensation in my stomach and chest associated with having failed.
Action: ???Magic unconscious hypnosis repair???
Result: Upon encountering the beginning of a slight sinking sensation associated with a failure, I no longer get dragged into counterproductive emotions. Instead, I feel nonchalant interest in what went wrong, an impulse to weigh whether it's worth trying to repair completely, and a motivation to make any cheap repairs that are available immediately.

Strategy Updates

Here's what I've learned over the past week about habits and installing them, and what I plan to do about it.

1. The current version of the installation procedure works best for a narrower class of habits than I recognized at first.
   
   Next actions:
   1. Pin down more precisely what kinds of habits it's good for.
   2. Look for small tweaks to the procedure that might accommodate more kinds of habits.
   3. Consider investing in large changes or multiple procedures.
   
   
2. I need to dig into Rule 1. ("Aim: I will endeavor for every habit I train to be the one I most desperately need at that time.") I meant for it to be an often unattainable ideal to strive for, something to keep my from getting distracted and losing my purpose, and not so much a "rule" that I must adhere to perfectly. My intuitive feel for what I need most isn't turning out to be quite as strong as I expected, and I'm experiencing some analysis paralysis.
   
   Next actions:
   1. Make a list of possible criteria for choosing the next habit.
   2. Write it up as a blog post if it goes well.
   
   
3. Offline training should definitely be more streamlined. How best to use my offline training time will vary a lot by context and mood, but I found myself wishing I had a list of questions posted in front of me to guide me. (Terminology: "Offline training" comes from machine learning. Online learning updates mappings when each new data point comes in. It's good when data become available sequentially. Applied to humans, we call this "learning on the fly". Offline learning techniques are good when a large batch of data is available at once. Cramming for an exam is a human example. What I'm calling "offline training" in this context is whatever I decide to do when I sit down for a few minutes to look at all the relevant facts at once.)
   
   Next actions:
   1. Brainstorm a list of offline training questions
   2. Pick the best ones and make a list to post in the zendo
   3. Write a blog post about offline habit training (pending feedback from at least one more installation)
   
   
4. Offline meta sessions (to reflect on and strategize about the overall procedure) aren't built into the current installation procedure. In retrospect, it's obvious they should be.

Next actions:
1. Decide what the schedule should be for meta strategy sessions
2. Make a list of questions to guide meta strategy sessions

Log

12/31/2014

[This first entry is all prep work. It's probably more detailed than future reports on prep work will be.]

My best guess at the skill I most desperately need right now is resilience: the ability to recover rapidly, especially from failure; to bend without breaking.

1. Be able to generate concrete examples of successes and failures to apply the skill.
   
   An example of successful application: Every time another approach to teaching epistemic rationality failed, CFAR adjusted and tried something else, rather than giving up on teaching epistemic rationality.
   
   An example of failure to apply the skill: I got a C on my very first logic test in college. Rather than correct my mistakes and study for the next test, I was crushed and spent several days agonizing over whether to drop the class. Complete failure would have been dropping the class at that point (which I didn't and went on to excel in highly advanced logic courses), but perfect resilience would have prevented any waste of time or energy.

2. If a skill requires multiple habits, train them serially, and repeat step 1 for each individual habit.
   
   This skill seems to require several habits. It's difficult to pin them all down, but I have at least identified a few. I'll start with "growing the roses of success": feeling emotions in line with knowledge that my failure has been educational.
   
   For every big mistake you make be grateful!
   That mistake you'll never make again!
   Every shiny dream that fades and dies,
   Generates the steam for two more tries!
   So when it gets distressing it's a blessing!
   Onward and upward you must press!
   From the ashes of disaster grow the roses of success!
   
   An example of growing the roses of success is burning a batch of cookies and feeling happy to have learned that my new oven is hotter than my old oven. Failure to grow the roses of success in the same situation would be sulking about having burnt the cookies.

3. Clearly define at least one high-quality trigger for the proposed action before beginning to train that habit.
   
   When I imagine burning the cookies, the deciding moment that splits the success worlds from the failure worlds is the moment when I'm surprised to find smoke and blackened cookies after opening the oven door and I feel a trapped, sinking sensation in my stomach and chest. In the failure worlds, I let that feeling drag me into an inescapable pit of negative emotions. In the success worlds, I respond to it in a way that shifts my focus from the badness of my mistake to the goodness of information. (Figuring out exactly what intervention will cause that shift comes later.)
   
   Imagining other concrete examples produces the same results, so my first guess at the right trigger is "the experience of unpleasant surprise at my mistake accompanied by a trapped sinking sensation in my stomach and chest". Therefore, if I encounter that experience, then I will activate reflective attention to reveal further details and inconsistencies with my prediction.

That's it for the prep work!

1/2/2015

I'm not encountering enough instances of my trigger. It happened once yesterday, and I didn't catch it fast enough. That means it's time for...

4. Seek opportunities to practice.
   
   I will now study the experience of realizing I've made a mistake by playing 2048.
   
   Results: Oh man, awesome side effects.
   
   
   1. I'm using my knitting counter, and since that's already a conditioned reinforcer, I'm automatically coming to associate noticing I've made a mistake with positive feelings. I didn't even notice before how much I direct my attention away from my own mistakes. I wonder if I could break that habit even faster using a primary reinforcer.
   2. This is quickly training me to notice the difference between an error of judgement and a random "shit happens", since I only get to click the counter for errors of judgement.
   
   This is the best game of 2048 ever. I'm rewarded in the natural way by the game when I don't fuck up, and I'm rewarded by the habit training every time I notice I've fucked up. I'm literally laughing out loud at my fuckups. This is so much fun. I love rule 4.
   
   My count for today is 38 so far, so I'm clearly in the middle of...
   
   
5. Train triggers before actions.
   
   I actually updated my trigger partway though without being foveally aware of it. I think my first hypothesis for the trigger was wrong. The surprise at my mistake and the dread/sinking sensation are not simultaneous. In fact, the dread/sinking sensation isn't even my usual response to noticing I've made a mistake. My usual response actually seems to be to try to ignore the mistake. It's only when I fail to ignore it that I experience the dread.
   
   Trying to ignore a mistake feels like trying to avoid eye contact. I even seem to be more likely to make another mistake immediately afterward, because I act hastily. I think maybe I'm trying to distract myself from the first mistake, though it actually feels more like I'm trying to distract the world, like if I move fast enough the world won't notice I messed up and it won't count. Same as the five second rule when I dropped food on the floor as a kid.
   
   Updated trigger: The sensation of surprise directed at something I recognize as my mistake, independent of the sinking sensation or even the sensation of trying not to look at the mistake.

1/4/2015

I feel like I'm doing something wrong, but I'm a bit sleep deprived and I'm having a lot of trouble concentrating enough to work out what it is.

It might be that I'm practicing the wrong thing. My current trigger is "the sensation of surprise directed at something I recognize as my mistake", but I updated to that in an attempt to not ignore my mistakes, which wasn't the original goal. The original goal was to cut back on despair in response to mistakes and promote something like satisfaction and curiosity. It's only the very tiny mistakes that I'm able to ignore anyway, so although not ignoring tiny mistakes is an important skill (one I'm adding to my wishlist), I don't think it's part of resilience, and I don't think it's The Most Important Thing for me to learn right now.

The times when I've made and noticed mistakes on my own so far this week, I've not felt the despair-type feelings that I flagged as problematic before. Like when I accidentally left my knitting counter upstairs this morning. I just felt "oops" and maybe a tiny bit of frustration, then I ran upstairs to retrieve it. That's all there was to it. That kind of feeling doesn't have the potential to get in my way.

The only times in the past few days when I've felt the problematic thing I flagged have been while interacting with other people. And I don't think I clicked the knitting counter for any of those, because they weren't straightforwardly mistakes. In retrospect, some of them actually were things I perceived as evidence of mistakes, but I didn't notice that at the time: for example, when I made a Facebook update and people responded with apparently off-topic comments, indicating I hadn't made my point clearly.

I'm thinking the problem is closely related to inadequacy in the eyes of other people, not so much myself. It definitely feels like every time I've felt big anti-resilience emotions, it has been because other people have not responded the way I hoped for them to. It's a little confusing, because if I perceive a failure myself that I don't believe others perceive as a failure, I still feel the despair thing, but only if other people are somehow involved. If I write a blog post that includes a mistake people criticize, I feel it, and if I write a blog post that people like but don't interpret as I intended, I also feel the thing. I mostly don't feel the thing if I make a private mistake that nobody else finds out about.

Updated trigger: I think I'll go back to noticing the trapped, sinking sensation in my stomach and chest, and I'll seek opportunities to practice by reading critiques of things I've written.

1/6/2015

6. Test a variety of actions if required.
   
   This sometimes happens. It's a little inconvenient given that I wanted to use this first habit to demonstrate in quite a bit of detail how the habit installation process works. But for me, at least, it happens at least half the time.
   
   Sometimes, without my conscious direction, my brain skips the "test a variety of actions" part. I jump from "ok, I mostly have a handle on my default response to the trigger, and I can notice it reliably" to "have the preferred response to the trigger instead", with no purposeful intervention at all beyond simply learning to notice the trigger. In this case, it's happening even without me having become consciously aware of what exactly my preferred response to the trigger is.
   
   The new response isn't exactly like I predicted. What I imagined originally was more of a focused curiosity and maybe a triumphant feeling similar in intensity to the sinking sensation from before. Instead, I've replaced the trapped feeling and sinking sensation with a nonchalant interest in what went wrong, an impulse to weigh whether it's worth trying to repair completely, and a motivation to make any cheap repairs that are available immediately. In retrospect, that does seem like the best emotional response for producing the most desirable behavioral responses. I suppose I was imagining overpowering the negative reaction with a positive one. This seems better.

I still need to stick with it for a few days before starting on another habit to make sure I don't lose the ability to notice the trigger, but at the moment it looks like the problem has mostly been fixed, and the new habit mostly installed.

The main problem when I perform an unconscious intervention like this is that if in the future it fails to work, I won't know what levers to manipulate to get it working again. Since I don't know that that issue will actually arise and I can just take a few days to implement step six if it does, I declare this habit 80/20d. I'll move on to my next habit on Thursday (a week from the start date) if I don't encounter more problems.

1/8/2015

I'm not entirely satisfied with the installation of this habit because the intervention (whatever it is) hasn't been tested harshly enough for me to feel confident that the problem's mostly fixed. But I also have a feeling it's not quite the right kind of habit for this process. Instances of the trigger that are high enough intensity to thoroughly test my progress are quite context dependent, and aren't happening frequently enough for training on the scale of one week to a month. I suspect I either need habits with more frequent triggers, I need to be more opportunistic by picking habits with triggers that will be frequent in contexts I predict will occur in the near future, or I need to change the procedure to accommodate less frequent triggers, perhaps by training more than once habit at a time. Or perhaps I should have a tiered system, where at any given time I'm training one high-frequency habit, one mid-frequency habit, and one low-frequency habit. I'll think on it.

The meta stuff is really important, especially this early on, so I'm going to hold off on choosing a new habit for a few days while I work out how to respond to problems that have arisen so far.

2/9/2015

One month since this post, and things seem to be holding steady with Growing the Roses. I fairly rarely notice the trigger consciously (maybe once a week), but my experience of small failures has been awfully smooth sailing. (Performing the desired action without noticing the trigger consciously is part of the goal. Noticing is essential for training, but mastery of a habit means completely effortless, automatic performance.) My failures are notable for their lack of salience, so the change isn't obvious when I'm not reflecting on it, but my memory of the past month is not punctuated by failures, and that's definitely new. I still haven't encountered anything I consider a really big failure. I'll update again with a full report on my experience of it as soon as one happens.

Reflective Recording

Related Posts: Mindfulness, How To Train Noticing, Feeling Clearly, Tathatā: Why Be Here Now?, Simulating Confusion, What It's Like To Notice Things

What is a reflective record?

A reflective record is anything you write down while in reflective attention.

What does a reflective record look like?

Here's a typical example of one of my reflective records from a couple months ago.

The sound of cars on the road, and a fly flitting through a beam of sunlight. I’m sleepy and my head feels fuzzy. The laptop is uncomfortably warm on my legs, and I think I should move it. The room smells like empanadas from lunch. A moment of blankness, which gives way as I realize I’ve simply lost direction for a moment. I gently nudge it back to the flow of my stream of consciousness. I notice that I have a Facebook notification, wonder what it is, and now I’m deciding to close all my tabs but this one. A feeling of familiarity like an openness in my chest, and as I watch that, memories of having performed this exercise many times. Through my inner monologue, the words “What will my readers think of the chapter I'm working on?”, accompanied by a very dull and mild pang of anxiety. I take a deep breath, and I’m enjoying the sensation of the air rushing out of my nostrils as I exhale. The words “categories of experience”, and I imagine circling phrases with colored pencils. A feeling of sufficiency and completion, part of my experience of the belief that I should stop writing now.

And here's what happens if I make a reflective record right now.

Cars passing by on the road making a swooshing noise. I'm imagining the scene out the window, though I'm looking at my computer screen. I'm imagining the visuals of a sunset over the ocean framed by hills and buildings, though it's actually morning and the sky is cloudy and gray. A pang of hunger rising in my stomach, feeling sharp and insistent. Words in my head: "what should I eat?", and a little frustration. Stopping work to eat feels at once aversive and enticing. The thought of eating causes relief and happiness, but the thought of cooking causes gumbly dark denial and I want to ignore the thought. My socks are gray-blue and fuzzy, and they make me content and comfortable. A memory of the way my attention suddenly retreated from the thought of food to grasp the nearest non-food-related sensation. Sleepiness, a constant temptation for my attention to wander away and forget itself, and apparently I'm more willing to describe my experience in imprecise metaphor than I feel I remember having been in past reflective records. My mind wants to focus on the difference between reflective recording and free writing. The locking-in-place-resolution of a decision not to bother writing about free writing in this post, but to reconsider after I eat.

What is reflective recording good for?

I use reflective recording for three things.

1. Habit training. Suppose you're trying to learn a more productive psychological response to confusion than the one you usually have. If you want to respond with curiosity, you'll need an intervention that inputs the beginning of your default response and outputs curiosity. To figure out what that intervention should be, it helps to have a detailed model of the input. Human memory isn't designed to store most of the sorts of things that go on in moment-to-moment awareness, so if you don't capture the details right away, you'll probably forget something important. If gather several reflective records after the same trigger over time, you'll get a better idea of how widely your default responses to the trigger vary.
   
   The same goes for testing the output: To know quickly if the intervention reliably causes the desired mental state, you need to know what mental state it causes, and you need to keep track of the results over time.

2. Responsible introspection. Responsible introspection is a way to gain self-knowledge while bypassing the introspection illusion. It means paying attention to immediate experience first, and reasoning abstractly about that data later.
   
   We do not have direct access to the origins of our mental states, but we do have mental states, and the contents of those mental states aren't arbitrary. Our experiences provide data about our patterns of thought. To introspect responsibly, collect that data by activating reflective attention in the presence of whatever stimulus interests you (a thought about a new job offer, for instance), and then writing down what you experience. You can repeat that a few times to find out how your reactions to the thought vary over time.
   
   Once you have detailed first-person data that isn't contaminated by inference and belief about belief, you can add it to third-person observations about your past behaviors. From there, it's relatively safe to reason abstractly about problems that depend on predictions about how you'll think and feel.

3. Predicting experience. Most of immediate experience is forgotten. Most of it doesn't matter, isn't vivid, isn't unusual, and doesn't make a lasting impression. It takes an extra reflective effort of become aware that your mind's doing whatever it's doing. A lot of the truth of what it is to be a mind slips through the cracks, so our default models of immediate experience lack crucial information. For example, we tend to hold onto beliefs we form and dispense with memories of what observations led us to form those beliefs, and what emotions colored our perceptions as we integrated those observations.
   
   When you have a better model of immediate experience, you can make better predictions about how you'll think and feel on a moment-to-moment basis in the future. Practicing reflective attention regularly can't bring back information you've already lost, but it can reduce illusions about experience that result from biases of memory.
   
   Making a reflective record now and then is even better than reflective attention alone, since it lets you review data taken from many time slices all at once.

Reflective recording is inspired by free writing and (my problems with) Gendlin's focusing, but it's a practice I developed myself. To my knowledge, nobody else has tried it yet, so I'll be very interested to hear about how it works, or doesn't work, for you.

---

Here's a conversation in response to this post from Facebook. I'll incorporate what I learned from it into the post soon (probably), but for now, I'm putting it here because it might clear some things up.

Malcolm: I would expect the act of writing stuff down to be way too slow and I wouldn't be able to think things in time. Might try this with speaking aloud and recording it as audio (which is actually what I expected it would be, based on the name).

Jamie:I found that attempting this slowed me right down. I can't write, type or speak even close to the speed I notice thoughts. Converting impressions and awareness into words and then into movement instructions for recording them is almost uselessly slow for stream-of-consciousness stuff. I can sort of 'buffer' because I can sustain two mental streams at once, but even so it's the mental equivalent of trying to run in knee-deep water. By the time I've finished writing something I lost awareness of 90% of the other things I was experiencing at the same moment as whatever it was I was writing down, and almost have to pick the next thing to write at random.

On the plus side, it did make me aware of just how MUCH I notice and immediately throw away without acting on, including stuff I probably ought to record or remember.

Me: You're both probably trying to catch a whole lot more than I am. I wait for a particular kind of thought when I'm actually using this for specific things. When I'm not I pick sort of at random with a huge bias toward stuff it's easy to put into words.

Malcolm: Hmmm... oh! Okay, yeah, I think I have a better sense of the structure. I think the examples you give are kind of misleading about this, as they imply just the random version.

Jamie: Right, so it's not a logfile, it's either a listener or a random activity sample. That feels a lot less close to 'awesome superpower', but a great deal closer to 'physically possible for unaugmented humans'. Your examples felt pretty much stream-of-consciousness, so I had interpreted it as 'log everything that seems important about a given moment'.

Me: Yeah, I was erring on the side of not including enough because I'm trying to learn to only say precisely what is needed. But when I designed a series of exercises on reflective attention, most of the point was to get to "partial reflection", which means keeping your attention fixed on a single category of thought (like physical sensations, emotional sensations, reactions to another person).

But I expect different people to parse their experiences differently, and I found that even for myself sans communication with others, it helped to have identified the most natural system of categorization for my moment-to-moment experiences. The random-ish sampling was originally just for finding those categories. It turns out that it's also great for moving into partial reflective recording if you're having a hard time getting a particular category in focus at first. Also, beginning with partial reflective recording and moving out of it only when it feels right tends to make my free writing sessions a lot more productive a lot more quickly.

Here's my categorization of the first example in the post. (It was edited a little to make more sense in context.)

How To Train Noticing

Alice wants to stop treating her beliefs as binary and start treating them probabilistically—that is, she wants to update herself incrementally. So she's hoping to work on the skill of raising her credence a little bit when she encounters weak evidence against her beliefs, instead of entirely disregarding anything that doesn't completely "change her mind". What should she do?

Obvious plan is obvious: If she encounters weak contrary evidence, then she should update slightly away from the hypothesis.

But obvious plan is not best plan. Why not?

Let's assume that Alice already knows exactly what she means by "update slightly away from the hypothesis" and knows exactly how to do it. (So the first problem is that in real life, she might not know either of those things.) The problem I want to focus on in this post is that "encounter weak contrary evidence" is a shitty trigger no matter how good the action you plan to take when the trigger happens.

Imagine one of those fake duck ponds you see at carnivals, the ones with the kiddie pool and the yellow rubber ducks. A current is pushing the floating ducks in circles around the edge of the pool. There are nine ducks with their bellies painted red, and one duck with its belly painted purple. To win the prize, you have to grab the purple-bellied duck when it floats by.

Now imagine the same duck pond, but instead of their bellies being painted, it's their backs. There are nine red-backed ducks and one purple-backed duck, and to win the prize, you have to grab the purple-backed duck when it floats by.

The second game's a lot easier, right? Why is that?

The mere fact that the purple duck is in front of you is an insufficient trigger. When you play the second game and win, you're not just grabbing the duck in front of you when it's purple. You're grabbing the duck in front of you when you see that it is purple. You notice a purple experience happening in your mind, and that's how you know to grab the duck. In the first game, you lose, because there's nothing to notice. Even though the ducks are in fact different, they all look the same from your vantage point.

Back to Alice.

The game she's playing is "update slightly when I encounter weak contrary evidence". The duck pond is the world, the current is time, and the ducks are events. Most of the ducks are red, and the purple ducks are "weak contrary evidence". "When I encounter weak contrary evidence" is a bad trigger in exactly the same way that "when the purple duck is in front of me" is a bad trigger. It doesn't pick out a subjective experience that distinguishes the attempted trigger from everything else. There's nothing to notice.

To make a good training plan, Alice needs an analogue to an experience of purpleness. She needs to know exactly what it feels like to encounter weak contrary evidence. Once she has that, then she has a reliable trigger.

So how can Alice find out what subjective experience is a function of weak contrary evidence? First of all, she's got to know what weak contrary evidence is. Not just what it feels like, but what it means. Let's assume she knows that already. So what's left is to identify the corresponding subjective experience.

Here's how I do it.

1. I guess. I remember or imagine a few specific instances of encountering weak contrary evidence (such as when I thought my friend wasn't attracted to me, but when I made eye contact with him across the room at a party he smiled widely). On the basis of those simulations, I make a prediction about what it will feel like, in terms of immediate subjective experience, to encounter weak contrary evidence in the future. The prediction is a tentative trigger. For me, this would be "I feel a sort of matching up with one of my beliefs, there's a bit of dissonance, a tiny bit of fear, and maybe a small impulse to direct my attention away from these sensations and away from thoughts about the observation causing all of this".
2. I test my guess. I keep a search going on in the background for anything in the neighborhood of the experience I predicted. Odds are good I'll miss several instances of weak contrary evidence, but as soon as I realize I've encountered one, I go into reflective attention so I'm aware of as many details of my immediate subjective experience as possible. I pay attention to what's going on in my mind right now, and also what's still looping in my very short-term memory of a few moments before I noticed. Then I compare those results to my prediction, noting anything I got wrong, and I feed that information into a new prediction for next time. (I might have gotten something wrong that caused the trigger to go off at the wrong time, which probably means I need to narrow my prediction.) The new prediction is the new trigger.
3. I repeat the test until my trigger seems to be accurate and precise. Now I've got a good trigger to match a good action.

If I were Alice, I'd take one more step toward noticing every instance of weak contrary evidence. A precise and accurate trigger is necessary, but it's not always sufficient. This kind of skill takes practice.

I have a knitting counter, which I bought for $7.13 on Amazon. Knitting counters are very simple: You press a button, and it advances the count by one. When I'm training myself to notice a trigger, I carry the knitting counter in my pocket. Every time I notice the trigger, I push the button. I reset the counter to zero at the end of the day, and the next day I try to beat my highest score.

(There are plenty of substitutes for the knitting counter, of course, such as keeping track in your head. But it does make a highly satisfying cliking sound.)

I keep doing this until my score levels out. Then, I swap out the action of pressing the button for whatever other action I think is useful. In this case, it would be "update slightly away from the hypothesis".

Usually, the leveling out process runs into the action-swapping process, so for a while I'm responding with the action while I'm still getting better at noticing the trigger. But if the action is any more complicated than pressing a button, I hold off on taking it and train noticing specifically until I'm feeling pretty comfortable with the noticing itself.

So in short, here's how to train noticing: Identify a subjective experience you want to notice, predict what the experience will be like, test your prediction, repeat 'til you've got it right, and gamify your practice.

Reflective Attention

And somewhere in the back of his mind was a small, small note of confusion, a sense of something wrong about that story; and it should have been a part of Harry's art to notice that tiny note, but he was distracted. For it is a sad rule that whenever you are most in need of your art as a rationalist, that is when you are most likely to forget it. —HPMOR, Ch. 3

A rationalist’s art is most distant when it is most needed. Why is that?

When I am very angry with my romantic partner, what I feel is anger. I don’t feel the futility of throwing a tantrum, or the availability of other options like honest communication, or freewriting, or taking a deep breath. My attention is so narrowly focused on the object of my anger that I’m likely not even aware that I’m angry, let alone that my anger might be blinding me to my art.

When her skills are most needed, a rationalist is lost in an unskillful state of mind. She doesn’t recognize that it’s happening, and she doesn’t remember that she has prepared for it by learning and practicing appropriate techniques.

The following exercise trains a skill I call reflective attention, and some call mindfulness. For me, it serves as an anchor in a stormy mind, or as a compass pointing always toward a mental state where my art is close at hand.

Noticing that I am lost in an unskillful state of mind is a separate skill. But when I do happen to notice—when I feel that small, small note of confusion—reflective attention helps me find my way back. Instead of churning out even more pointless things to yell at my partner, it allows me to say, “I am angry. I feel an impulse to yell. I notice my mind returning over and over to the memory that makes me more angry. I’m finding it hard to concentrate. I am distracted. I have a vague impression that I have prepared for this.” And awareness of that final thought allows me to ask, “What have I trained myself to do when I feel this way?”

The goal of the following exercise is to practice entering reflective attention.

It begins with an instruction to think of nothing, because when you monitor yourself to make sure you’re not having any thoughts, your attention ends up directed toward the beginnings of thoughts. Since the contents of consciousness are always changing, maintaining focus on the beginnings of thoughts prevents you from engaging for an extended period with any particular thought. It prevents you from getting “lost in thought”, or keeping attention focused on a thought without awareness of doing so. The point is not actually to be successful at thinking nothing, but to notice what happens when you try.

Keeping your focus on the constant changes in your stream of consciousness brings attention to your experience of awareness itself. Awareness of awareness is the anchor for attention. It lets you keep your bearings when you’d otherwise be carried away by a current of thought or emotion.

Once you’re so familiar with that feeling of mindfulness that creating it is a primitive action, you can forget the introductory part, and jump straight to reflective attention whenever it occurs to you to do so.

---

This will probably take around five minutes, but you can do it for much longer if you want to.

Notice what your mind is doing right now. One thing it’s doing is experiencing sensations of black and white as you read. What else are you experiencing? Are there words in your inner monologue? Are there emotions of any kind?

Spend about thirty seconds trying not to think anything. When thirty seconds is up, stop trying not to think, and read on.

What’s happening in your mind is constantly changing. Even when you were trying not to think, you probably noticed many times when the stillness would shift and some new thought would begin to emerge in conscious awareness.

Turn your attention to those changes. When a new thought emerges in consciousness, see if you can notice the exact moment when it happens, becoming aware of what it feels like for that particular change to take place.

If it helps at first, you can narrate your stream of consciousness in words: “Now I’m seeing the blue of the wall, now I’m hearing the sound of a car, now I’m feeling cold, now I’m curious what time it is…” You’ll probably find that you can’t narrate anywhere near quickly enough, in part because thoughts can happen in parallel, while speech is serial. Once narrating starts to become frustrating for that reason, stop slowing yourself down with words, and just silently observe your thoughts as they occur.

If you’re finding this overwhelming because there are too many thoughts, narrow your focus down to just your breathing, and try to precisely identify the experience of an exhale ending and an inhale beginning, of an inhale ending and an exhale beginning. Keep doing that until you feel comfortable with it, and then slowly expand your attention a little at a time: to other experiences associated with breathing, to non-breath-related bodily sensations, to non-tactile sensations from your environment, and finally to internal mental sensations like emotions.

If you notice an impulse to focus your attention on a particular thought, following it and engaging with it—perhaps you notice you feel hungry, and in response you begin to focus your attention on planning lunch—instead of letting that impulse take over your attention, recognize it as yet another change in the activity of your mind. If you’re narrating, say, “now I’m feeling an impulse to plan my lunch”, and keep your focus broad enough to catch the next thought when it arises. If you realize that you’ve already become lost in a particular thought, notice that realization itself as a new thought, and return to observing your stream of consciousness by noticing the next new thought that happens as well.

You might need to practice this many times before you get the hang of it. I suggest trying it for ten minutes to half an hour a day until you do.

Once you feel like you can recognize the sensation of reflective attention and enter that state of mind reliably given time, begin to train for speed. Instead of setting a timer for fifteen minutes or however long you want to practice, set it to go off every minute for the first half of your practice, spending one minute in reflective attention, and one minute out. (Don’t do this for all of your practice. You still need to practice maintenance.) When you can consistently arrive in reflective attention by the end of the minute, cut the intervals down to 45 seconds, then thirty, fifteen, and five.

---

In real life, the suspicion that you may be lost in an unskillful state of mind will be quiet and fleeting. “Quiet” means you’ll need to learn to snap your attention to the slightest hint of that feeling. For that, you’ll need to train “noticing”. “Fleeting” means you’ll need to be able to respond in less than five seconds. You’ll need to begin the process in less than one second, even if it takes a little longer to fully arrive in reflective attention. For that, training for speed is crucial.

Feeling Clearly

Every stripper knows to name a higher price than he expects to get when selling a lap dance. He'll start out by telling you it'll cost $75, and you'll say that's too much. Then you'll counter with $40, and he'll say, "How about $50, and I show you a new trick I learned yesterday." If I ask you later why you agreed to $50, then unless you already know about anchoring effects (or perhaps even if you do), you'll say the guy was hot and you thought $50 was a fair price.

But if I asked you the moment you walked into the club, "What's the highest price you'd consider fair for a lap dance with the guy on stage right now?" you'd say, "$25" (or something lower than $50, anyway). You wouldn't be lying to me. It would feel true to you.

Anchoring is just one among the many guises of the introspection illusion.

People tend to think they have direct access to the origins of their mental states. They think they're infallible when it comes to certain kinds of self-knowledge, like why they chose to be a teacher, whether they like broccoli, or why they agreed to pay $50 for a lap dance. But they're wrong.

This is a big deal.

Suppose you want to be more productive by making your work periods more enjoyable, so you've decided to start playing your favorite kind of music (French house) as you work. Here are some judgements that might have influenced that plan:

• your favorite kind of music is French house
• you're more productive when you're happy
• you don't find music distracting

Maybe all of these things are true, and maybe not. If they're not, your plan isn't going to work so well. Will you notice when it fails, or will you go on believing all of these things whatever happens, as long as they keep feeling true? The introspection illusion means that how true they feel to you is not an excellent indicator of how true they actually are, even though they're mostly about your own thoughts and beliefs. Empirical observations about productivity under various circumstances must be part of the story.

But not all kinds of introspection are equally subject to this problem.

The introspection illusion happens when we try to access the processes underlying our conscious mental states. Processes underlying our conscious mental states are not themselves part of our conscious mental states. So this is the illusion of feeling as though we are conscious of unconscious processes.

But we really are conscious of some things. You're conscious of the temperature of the room, now that I've brought your attention to it. You're conscious of the color of your shirt. You're conscious of the emotional sensations that occur upon reading the phrase "your grandfather's voice".

The surface level introspection you employed to become aware of each of those mental contents is far more reliable than the deep soul-searching people often associate with the word "introspection". And you can get a lot of mileage out of that if you know how to use it. This is what all the "mindfulness" and "being in the moment" stuff is really about.

If just asking yourself the question "How do I feel about my boyfriend's new girlfriend?" and going with the first judgement that occurs to you won't do the trick for predicting the emotions that will influence your interactions with her, what will work?

There's a technique I use often for making more accurate predictions about my future mental states. I call it "Feeling Clearly". It's not a method for revealing the true feelings hidden at the core of your being. It's just careful observation of what does in fact happen to your mind when it encounters whatever you're wondering about. If you're right about that, what you really truly feel deep down (if there is such a thing) isn't so important, is it? Predicting and influencing the contents of your consciousness is all that matters.

This is applied experimental phenomenology. It lacks many virtues of a randomized, double-blind, controlled, peer-reviewed study. But your feedback loops can be way fast.

How fast?

Quick, make a prediction about how much you will enjoy imagining smelling a rose. Ok, now imagine smelling a rose. How much did you enjoy it? Did you overestimate, or underestimate? Taking that into account, make another prediction about how much you'll enjoy imagining smelling a rose. Ok, now imagine smelling a rose again. Were you closer this time?

That fast.

Here's how it works.

---

Choose a simple idea or topic that makes you a little uncomfortable. Nothing really important or painful, just something small that's been worrying you, or that feels unresolved. It might be something a friend said to you yesterday. It might be an upcoming responsibility, or a recent event that didn't go as well as you'd hoped. Whatever it is, be specific, and then set it aside for later.

Notice what your mind is doing right now. One thing it’s doing is experiencing sensations of black and white as you read. What else are you experiencing? Are there words in your inner monologue? Are there emotions of any kind?

What’s happening in your mind is constantly changing. Turn your attention to the changes. When a new thought emerges in consciousness, see if you can notice the exact moment when it happens, becoming aware of what it feels like for that particular change to take place.

If it helps at first, you can narrate your stream of consciousness in words: “Now I’m seeing the blue of the wall, now I’m hearing the sound of a car, now I’m feeling cold, now I’m curious what time it is…” You’ll probably find that you can’t narrate anywhere near quickly enough. Once narrating starts to become frustrating for that reason, stop slowing yourself down with words, and just silently observe your thoughts as they occur.

If you’re finding this overwhelming because there are too many thoughts, narrow your focus down to just your breathing, and try to precisely identify the experience of an exhale ending and an inhale beginning, of an inhale ending and an exhale beginning. Keep doing that until you feel comfortable with it, and then slowly expand your attention a little at a time: to other experiences associated with breathing, to non-breath-related bodily sensations, to non-tactile sensations from your environment, and finally to internal mental sensations like emotions.

If you notice an impulse to engage with a particular thought—perhaps you notice you feel hungry, and in response you begin to focus your attention on planning lunch—instead of letting that impulse take over your attention, recognize it as yet another change in the activity of your mind. If you’re narrating, say “now I’m feeling an impulse to plan my lunch”, and keep your focus broad enough to catch the next thought when it arises.

Do that for about five minutes, or until you’re ready to move on.

When you’re ready, think the thought you chose at the beginning. Drop it into your stream of consciousness. Then immediately go right back to noticing thoughts, emotions, and sensations as they arise.

You’ll probably notice some activity occurring in response to the thought you just dropped in. Observe those responses, one after another, not being drawn into any one of them but remaining aware of each.

Do this for as long as needed. Think the thought again whenever you feel the responses to it have died down.

When you’re done, write down each of the reactions you recall, before they fade from memory.

As a variation, you can write down the reactions in the middle of the exercise, as they’re happening. I don’t suggest starting off with this variation, because it introduces a focus on words that might be disruptive. Find out what it’s like without writing the first time you try it.

Try several sessions of this spread out over the course of a day, or over a few days, and keep notes each time.

---

So what does this get you? It gets you reliable data on what happens when you encounter whatever thought you're interested in. It circumvents the introspection illusion to help you make more accurate predictions about your mental states, and therefore about whatever behaviors are influenced by them.

Now, that's not going to perfectly map onto real-world situations.

For one thing, it takes practice to get really rich, precise data; to distinguish "fear" from "a cold tightness in my chest I associate with anxiety, plus a feeling of directedness at an image of being abandoned".

Secondly, a thought about something isn't the thing itself. Your simulation of what it will be like to meet Tiffany will have some correlation to what it will actually be like to meet Tiffany, but it won't be perfect.

You can get better at that too, though. You can calibrate.

Get really comfortable with reflectivity, the central skill of this exercise. Then, when you actually meet Tiffany in real life, activate that reflective mode, and take note of how exactly your predictions fail. Form hypotheses about why, and feed those back into your next round of simulation.