Outward Change Drives Inward Change
The subsumption architecture for robotics invented by Rodney Brooks is based on the idea of connecting behavior to perception more directly, with fewer layers of processing and ideally no central processing at all. Its success, e.g. the Roomba, stands as proof that something akin to control theory can be used to generate complex agent-like behavior in the real world. In this post I’ll try to give some convincing examples from literature and discuss a possible application to anti-akrasia.
We begin with Braitenberg vehicles. Imagine a dark flat surface with lamps here and there. Further imagine a four-wheeled kart with two light sensors at the front (left and right) and two independent motors connected to the rear wheels. Now connect the left light sensor directly to the right motor and vice versa. The resulting vehicle will seek out lamps and ram them at high speed. If you connect each sensor to the motor on its own side instead, the vehicle will run away from lamps, find a dark spot and rest there. If you use inverted (inhibitory) connectors from light sensors to motors, you get a car that finds lamps, approaches them and stops as if praying to the light.
Fast forward to a real world robot [PDF] built by Brooks and his team. The robot’s goal is to navigate office space and gather soda cans. A wheeled base and a jointed hand with two fingers for grabbing. Let’s focus on the grabbing task. You’d think the robot’s computer should navigate the hand to what’s recognized as a soda can and send out a grab instruction to fingers? Wrong. Hand navigation is implemented as totally separate from grabbing. In fact, grabbing is a dumb reflex triggered whenever something crosses an infrared beam between the fingers. The design constraint of separated control paths for different behaviors has given us an unexpected bonus: a human can hand a soda can to the robot which will grab it just fine. If you’ve ever interacted with toddlers, you know they work much the same way.
A recurrent theme in those designs is coordinating an agent’s actions through the state of the world rather than an internal representation—in the words of Brooks, “using the world as its own model”. This approach doesn’t solve all problems—sometimes you do need to shut up and compute—but it goes surprisingly far, and biological evolution seems to have used it quite a lot: for example a moth spirals into the flame because it’s trying to maintain a constant angle to the light direction, which works well for navigation when the light source is the moon.
Surprising insights arise when you start applying those ideas to yourself. I often take the metro from home to work and back. As a result I have two distinct visual recollections of each station along the way, corresponding to two directions of travel. (People who commute by car could relate to the same experience with visual images of the road.) Those visual recollections have formed associations to behavior that bypass the rational brain: if I’m feeling absent-minded, just facing the wrong direction can take me across the city in no time.
Now the takeaway related to akrasia that I’ve been testing for the last few days with encouraging results. Viewing your brain as a complete computer that you ought to modify from inside is an unnecessarily hard approach. Your brain plus your surroundings is the computer. A one-time act of changing your surroundings, physically going somewhere or rearranging stuff, does influence your behavior a lot—even if it shouldn’t. Turn your head, change what you see, and you’ll change yourself.
[edited to fix formatting]
I agree.
I noted in a recent comment here To do my PhD writing I take my laptop down to a local coffeeshop and sit there for one to two hours of solid writing. It’s become habit. I go to the coffeeshop, I write. My body/brain knows that’s what I do there, so there’s nothing to fight.
I think this notion that external things can shape how we perceive, think and do things is quite general and applies to all aspects of our psychology. On the one hand that’s fairly obvious, but I don’t think it’s properly acknowledged in most people’s outlook—it’s still far too common to think that our behavior is simply a matter of our choices, and that when the environment influences you, it’s only in a superficial manner.
Here’s some examples from other areas. Donald Norman, in his The Design of Everyday Things, about affordances—how the design of things like door handles and chairs—can shape how you use them.
Jane Jacobs (The Death and Life of Great American Cities) and Richard Florida (The Rise of the Creative Class) in their own ways talk about how the structure of the city environment can shape people’s behavior. Paul Graham has talked about this as well.
(and of course there’s stuff like interior design, architecture, using music to affect mood, grocery shops laying out aisles in ways that exploit our psychology, etc)
I also think that Richard Dawkin’s The Extended Phenotype is an example of this general viewpoint—the viewpoint that these bounds bewteen what’s inside of us and what’s outside of us isn’t as simple as is assumed. He argues that the phenotypic effect of genes can properly extend beyond the bounds of creatures’ bodies -- Beavers’ dams being one simple example.
I’ve written a bit more about some of these topics on my blog.
del
This is pretty much the heart of akrasia, right there. What choices come to mind (if they’re even perceived as choices) are a function of the environment + current internal representations. In the absence of specific executive direction (willpower) or self-priming (imagination), we pretty much go with the flow.
Willpower doesn’t actually change the flow, either, although the actions stemming from it can. Self-priming, OTOH, can change the flow, once it reaches monoidealism (no other thoughts). This can take time and patience, however.
(Of course, you can also change the “design” of the “machine”, since our automatic responses are “soft”. I’m speaking here only of “real-time” options for changing behavior, as opposed to things you can do in advance.)
del
I didn’t use the phrase; you did. ;-) I was replying to the rest of the sentence.
I don’t get this. The idea of a “next step” is a bit more parsimonious than trying to describe the details of the feed-forward prediction hierarchy, even if it’s less technically accurate. But perhaps you have a different meaning for “next step” than I do. I just mean our brain’s prediction of “what happens next” in the world, which may be a self-referential and self-fulfilling prophecy.
These predictions are generated by a combination of external state information + internal state information (current goals). So for example, if one level of my current internal goal state is to walk across the room, and I’m currently sitting down, my prediction at one level of “what happens next” is the prediction that I’m going to stand up… which then cascades down to lower-level predictions triggering the motor actions to do so. Meanwhile, my internal state is still holding the subgoal of getting across the room.
We don’t necessarily have simultaneous awareness of all subgoal states, nor are most of these goals and subgoals consciously chosen. To initiate an action that runs counter to our active predictions, we have to change the entire subgoal stack, hence the time-and-patience part.
For example, if I’m not currently in the mood to work out, I can simply sit and visualize working out, and deflect all the objections that pop up in my head—objections that are essentially my brain’s prediction that, given my current state, I don’t want to exercise. I ignore them, and continue visualizing, until the objections are exhausted and all active subgoals are purged, at which point I suddenly notice that somehow or other, I’ve already wandered over to the treadmill and started it up while I was still busy visualizing.
(Note: YMMV if you try this; the actual technique has enough potential “gotcha” details that I’m teaching a live workshop today for my group so there can be practice and Q&A to handle them.)
That depends on what you mean by “central” and “control”. ;-) If you mean conscious central control, then yeah, there’s pretty much none. The goal stack (heap?) isn’t really under conscious control. it just seems that way because consciousness is triggered by exceptional conditions in the control flow. It’s like consciousness is a quality control consultant who thinks he’s in charge of running the business, because everyone comes to him with their problems. ;-)
That being said, consciousness seems to have a little bit of goal stack of its own, the ability to direct the attention as a whole, and the ability to suppress actions. These three abilities, when combined, make it possible for us to influence the rest of the system.
(This isn’t something I’ve focused my work on much in the past; my main focus has been changing the memories that drive the contents of the goal stack—i.e., what predictions will occur in what internal+external state combinations. While it’s still a conscious intervention, it amounts to changing the programming in advance, rather than making changes “while the program is running”. I’ve branched out into this area partially as a result of LW discussions, and partly because I’ve gotten to the point where the constraint on my productivity is no longer my old bad programming, but the lack of new good programming.)
In this particular context, hypnosis and LoA have vocabulary and practice regarding monoidealism, going back about 100 years. Unfortunately, both fields also have a huge amount of utter nonsense theories floating around.
I don’t see the discussion as superficial, but then that’s because I’m viewing this stuff as extensions to my basic model of behavior as prediction-driven. (See the wikipedia page on the memory-prediction framework for the cog sci part.)
My work has been focused on changing memories (via Reconsolidation approaches) to change the predictions, and thus, the behavior. The somatic marker hypothesis also plays a role, in verifying successful updates to cached predictions.
I don’t talk about the details of these things much in my work, because it is not essential to know how they work in order to make successful behavior changes. To date I’ve focused primarily on disruption of dysfunctional predictions, rather than the creation/addition of useful ones, but that has started changing this year.
Actually, “fighting” akrasia on an ongoing basis is a dumb idea, based on a confusion about what it is. It’s more efficient in the long haul to change your stored predictions, than to spend all your time working around them.
Well yeah, changing the environment changes your behavior. But since you don’t describe how these changes relate, the practical advice comes through as a suggestion to inject more randomness in people’s lives, and somehow that is supposed to help with akrasia...
This may be related … a friend once said that when she half-wakes from a dream, if she rolls over and goes back to sleep, the dream ends. But if she doesn’t roll over, the dream continues.
I believe this works for me, too. Perhaps it’s not just a change in sight, but in touch too. Or perhaps in any of the senses.
I was planning to write about something similar myself when I have more free time. The embodied/dynamical approach to behavior is similar to Brooks anti-representationalist approach and I think its best statement is philosopher Alva Noe’s recent book Out of Our Heads. His account of perception, which builds on the ecological psychology of Gibson, is particularly excellent (he goes into more depth about it in his earlier book Action in Perception; you can also find most of his papers online). I have yet to find anything that isn’t better explained by the embodied/dynamical approach and it has the advantage over representationist/computationalist cognitive science of being biologically-plausible.
Embodied/dynamical is not orthogonal to representationist/computationalist. It’s presented that way in the literature, because a good story has to have conflict.
In my experience, behavior-based programming breaks down above about 100 behaviors, which Brooks etc. never get anywhere near in their robots. The number of interactions between n behaviors scales as n-squared. This is bad.
I find accounts that try to reconcile the two approaches unconvincing. Andy Clark tries to do this and his books and articles suffer greatly for it. There’s no reason you couldn’t combine the two, of course, but the problem is coming up with a reason why somebody would hang on to representationalism if it’s no longer “the only game in town.” Representationalism/computationalism is unmotivated by evidence, creates more problems than it explains and is biologically-implausible. If you have alternative explanations without these problems then I fail to see why you wouldn’t use them.
Hmm, in what alternative approach do the 100 behaviors not interact?
Indeed, I reorganized my office a bit this week so that I can easily clear my desk off onto the shelves behind me, and the mood/attitude impact of having a regularly clean desk is noticeable.
However...
Some things are easier to change from the inside than the outside, though. For example, if you have an aversion to something that you actually need to do, it’s probably easier to get rid of the aversion than to get rid of the need to do it.
Also, the thing you have an aversion to might be something you have to do a lot, so if you do something about the aversion you’re making future iterations easier. Of course, if you can remove multiple future needs all at once that might be even better.
I agree with your point, but might I add the caveat here that a lot of people feel they “need” to do things that aren’t actually necessary?
Oh sure, in fact that’s often the source of the aversion. ;-)
Like what? What definition of “necessary” do you mean to use?
Necessary as in working toward accomplishing some deliberate goal, as opposed to, for instance, behaviors trained in from childhood that take the form of random, arbitrary obligations that serve little purpose.
Can you give some examples of behavior that is unnecessary in this way?
A typical example is people who were trained to always clear their plate of food rather than waste it; if you’ve eaten your fill you don’t need to keep eating.
Another common one is an excessive focus on tidiness—putting random objects back in place isn’t all that necessary, especially for things that you’re likely to be using again soon. Right now I have some papers and a few books sitting out on my desk—I don’t need to put them away right now, because it would serve no apparent goal that I value, but I’ve known individuals who wouldn’t leave anything sitting out for more than a couple hours if it wasn’t being actively used.
There are programs for behavior modification that attempt to optimize your cleaning/tidying time for you which recommend exactly this behavior. (My girlfriend did this for awhile.) It reduces one’s need for lengthy cleanup sessions. So some of the people you observe doing this are serving a conscious goal.
Clearing one’s plate may well be an optimization for populations that experience famine. In the US, there’s little utility. In Houston, Texas doing this when eating out will almost certainly result in overeating.
The issue is not whether it’s a good way to tidy, it’s whether keeping things tidy is actually a deliberate goal vs. a trained behavior or perceived obligation.
If it is a deliberate goal that’s fine, but orthogonal to my point.
Perhaps it’s orthogonal to your point, but it calls to question one of your examples. The other one seems quite solid.
Perhaps the title should be ‘Outward change obviates inward change’?
Good suggestion, but I’m not sure. Reportedly Arthur Conan Doyle radically increased his attractiveness to women after his stint as a seaman. You will get some inward change into the bargain too.
I’m confused by that example.
Let’s say, by increased attractiveness you mean he started talking more attractively, then that is an outward change, but then the question is whether it was brought about by an inward change.
If the change happened without him thinking about it and only because of his surroundings as a seaman, - which is the point of your post - then it’s surely not an inward change.
But if he changed upon reflection of his experience at sea and consciously changing his behaviour, then your robot analogy breaks.
Help!
I meant your second option. You’re right, “inward/outward” was an inaccurate wording—English is my second language and I sometimes get carried away with the sound of words instead of their meaning, as is customary in Russian. Would’ve been better to say “change on the outside drives change on the inside”. I will mull it over a bit and maybe change the title.
I don’t quite see the connection to control theory, but yes, I’ve been practicing stimulus hygiene and controlled self-priming by intentionally exposing myself to “desirable” stimuli and limiting exposure to “undesirable” ones (my internal name for this is “prime-steering”—as opposed to uncontrolled “prime-drift”), and I like the results so far.
(BTW, one of the tricks I found for quickly and cheaply changing my “immediate surroundings” is simply changing my desktop wallpaper.)
Yep—using the wallpaper trick too. Also, forcing myself to go somewhere in the evenings instead of sitting at home pays off a lot.
How do you choose the “somewhere”—do you have a purpose in mind, or you’re just drifting?
The purpose is to stop thinking and start reacting, the necessary ingredients are people and music.
Could you expand a bit on the nature and effectiveness of the results you’ve gotten?
Elimination of two my primary addictions—PC gaming, mostly TF2 and WoW AH moneymaking (100% elimination for both so far), and aimless internet surfing (now reduced to about 10% of its former glory.)
Significant reduction of procrastination time in general (ironically, I’m procrastinating right now—I’m posting at LessWrong :) -- but I have an excuse this time, I had a 16-hour workday yesterday (I know, not something to be proud of, but it’s a fact.)
Being able to actually act according to my current better judgement, with verifiable results in the real world. This is an addiction in itself.
Note of caution: the results are just about 3 months old, so I can’t say yet that the tricks work reliably. Also, I wasn’t under any serious stress or pressure, so I can’t say whether these tricks will work under such conditions. And finally, these are tricks, so perhaps they work only on myself, and I have no deep theories to explain why they work.
(Cousin, looks like we’re turning your thread into an akrasia trickfest.)
Wow that’s amazing Vladimir, well done. The obvious next question is.… how did you do it? Please give an example of at least one of your tricks if possible.
Edit: I’ve made a top-level post for sharing anti-akrasia techniques—go ahead and share your techniques as well. Let’s continue the discussion there.
A very quick outline (I’ll post a detailed version later).
Determine what is your current better judgment. This is critical—I noticed that I hesitate to trick myself into doing anything I don’t consider to be relevant to my goal.
Constant asking myself: “is the activity I’m doing at the moment advancing me toward the desired state of reality”? If the answer is “no”, know that I’m procrastinating.
80⁄20 elimination, Tim Ferris / Pareto style (I’m skeptical about the rest of Tim’s book, but the Elimination chapter is pure gold).
Parkinson’s law (work expands to fill the time allotted). Again, Tim has some advice on it—basically, it boils down to scheduling most important things (in the 80⁄20 sense) first, with aggressive deadlines.
PJ Eby’s secret meaning of “just do it”. He considers the article to be outdated, but its key paragraph worked wonders for me. Basically, “just do it” = “don’t do anything else”. In its pure form, “not doing anything else” is too macho for me, so I leave a line of retreat for myself—I permit myself to eat, think about anything (not just the task), walk, have sex, but no surfing unless it’s on-topic, no doodling on paper unless it’s on-topic, etc.
Self-priming—I try to expose myself to stimuli related to my current task, and to shield myself from irrelevant stimuli, no matter how pleasant (e.g. I run away from my toddler daughter, because prolonged exposure to cuteness tends to totally ruin my ability to work efficiently :)
Begin. For example, if you need to do some stuff in Excel, just open an empty spreadsheet and type in the table header. Just stupidly staring at this makes you better at spreadsheets and your task—your mind pulls linked concepts to the fast-access cache, without your consent, and you don’t need to do anything.
Mindless repetition of things like “I want to do/make/design/create/code the best X in the world” (X stands for a short-term todo item, not some far-off goal) -- I’ve been doing this for at least a decade, and it seems to work (don’t know how, but perhaps it’s related to Cached Selves).
Sorry if I forgot something—gotta run (will edit later.)
(Obligatory note of caution—these are tricks, they are not proven by experimental results, I’ve been using them for just about 3 months, they aren’t tested under serious stress / pressure, I have no deep theories to explain why they work, and the evidence I offer is purely anecdotal.)
Perhaps any further discussion would be better served by encouraging Vladimir to make a top-level post on the subject? This seems to be veering slightly off-topic.
I agree. Go Vladimir!
Done.
The post is still in your drafts—it’s not submitted to LW proper, although it looks to you like it is. (You can verify this by signing out and looking at the new posts.)
Uh, how do I submit it?
Edit: nevermind, found it.
Suggest some good wallpapers.
Superb nature photos, pastoral porn and some Hubble:
http://www.wpdb.de/
High-quality dual screen / widescreen backgrounds:
http://www.mandolux.com/
Assorted stuff (different topics, sizes, quality):
http://www.smashingmagazine.com/2008/12/10/50-most-amazing-dual-screen-desktop-wallpapers/
However, my wet dream is http://www.micronaut.ch—they have the most stunning microscopic photography I have ever seen, but unfortunately they don’t offer their imagery as wallpapers or microstock (they offer prints on demand).