Sunday, December 18, 2011

Negative Feedback Loops

Ten years ago, almost to the day, I took what was at the time the most difficult exam of my entire life; the 6.002 final. It was later eclipsed by the 6.046 final, but at the time the test I wrote in Johnson was one of the most trying experiences of my life. It had 5 questions worth 40 points each, and I solved the first two completely in less than half an hour. My final score on the exam was in the 90s (out of 200), which means in the final 2.5 hours I managed to score something like 15 out of a possible 120 points. And here's the kicker; my grade on the final was "a strong B." Fucking ridiculous. Anyway, that class was super difficult, and to be honest it wasn't one of the ones that really reached you deeply and changed the way you think. 6.046 did that for me, as did 6.034 (which wasn't actually that hard) and 6.041. Even 8.01 did, in that it set up the rhythm of "how to beat an MIT class" for me. At the time he limped out of 6.002 with a B all that 19 year old Jesse was really able to think about was "Holy shit did anyone get the license plate?" Since then, however, I've realized there was a very important lesson to be learned in the class, if you could just manage to see it through all the differential equations. It had nothing to do circuits or electronics, but was taught through the class's goto device; the MOSFET. The basic idea wasn't all that interesting at the time, but since then I've realized that it's one of the key concepts you need to utilize to keep your life in order; the negative feedback loop.

You hear references to the negative feedback loop (or its evil twin brother, the positive feedback loop) all the time. Have you ever heard the expression that "a football team is either getting better or worse"? That's a positive feedback loop. I guess at this point I should maybe define some of the terms I'm just throwing around, so here we go. A negative feedback loop exists in any process where the act of parameters moving outside of acceptable boundaries actually causes those very parameters to move back into the acceptable range. I don't remember the details, but this is how the MOSFET worked; when the current (or voltage, or whatever) crossed a certain boundary something changed and it went back to what it was supposed to. What's important to realize here is that it just happens; the parameter getting out of whack causes a change in the system that moves the parameter back in range. A positive feedback loop is just the opposite. In a positive feedback loop when a parameter starts to get out of whack it causes a change in the system that causes the parameter to get MORE out of whack.

I can hear you now..."OK Jesse, that's vaguely interesting, but what on Earth are you talking about?" Well hopefully I can try to explain it a little bit with some examples from my life. An obvious one that comes to mind is Danielle's ability (or inability) to keep areas clutter free. For her the entire process is a positive feedback loop; once an area gets "cluttery", be it dishes in a sink or clothes on a floor or whatever else you want to imagine, she views the space as "dirty" on the "clean/dirty" binary scale and therefore no longer sees any reason to make any effort to keep it "clean". So what happens? The number of dishes in the sink should be zero, and Danielle has no problem keeping the number AT zero. But once it jumps to one, the positive feedback loop rears its ugly head and boom, one becomes three and three becomes seven and that's ballgame. For me the exact same process is actually a negative feedback loop (and in this case it's because I do not ascribe to the binary "dirty/clean" categorization system). For me the correct number of dishes in the sink is also zero, but once that number jumps above zero in my world the process is governed by a negative feedback loop. Eventually (and it may not be right away) I do something (open the dishwasher, insert dish) to move the parameter back inside the acceptable range. Some other intuitive examples are the process of going on tilt at a poker table; this is definitely a positive feedback loop cycle. Your mental parameters (stress, focus, rage, etc) have acceptable ranges and once you go on tilt they get out of bounds, and them getting out of range causes them to get MORE out of range. And that's basically how tilt works. The Tommy Angelo way of doing things basically suggests you set up your "process" in such a way that you make it a negative feedback loop; when you feel your parameters getting out of bounds you make a conscious decision to do things (take a break, breath deeply, be quiet, etc) to put them back in range. It's all based on the derivatives; if the derivatives of a parameter stay constant or even grow as a process advances, it's a positive feedback loop. If the derivatives change sign (first become positive, which then actually causes them to go negative), it's a negative feedback loop. Addictive behaviors (drug abuse, alcoholism, gambling problems) are positive feedback loops.

After that extremely long winded explanation, hopefully the point of this post has become pretty obvious. In your life you have literally dozens of process and procedures, and if you want to keep everything under control you need to set up most of them to be negative feedback loops. Or, at a bare minimum you need to be aware of any positive feedback loops that you have and be vigilant not to let their parameters get out of bounds. A perfect example for me these past few weeks has been my process of coffee consumption. I never drank coffee really before this year, and all of a sudden for some reason at work I decided to let myself try a small Starbucks for $1 every morning. And (big surprise here) I really liked it. For a while the process stayed in control, with me just having a single small coffee in the morning. But eventually I started getting the occasional refill; then it became a refill everyday. Then sometimes it was two coffees with one of them being a medium, or even (gasp) two refills. And just like that, boom, I have a day where I ingest close to a gram of caffeine (I mean really...that's like drinking THIRTY cans of diet coke) and (another shocking surprise) I feel like absolute shit. Positive feedback loop: 1. Jesse: zero. Another example for me of a positive feedback loop is "going to the gym". If I go to the gym, I tend to go more. If I don't, I tend to go less. The right number of gym visits per week for me is probably like 4; but I seldom hit that number. I have a positive feedback loop that I need to actively manage in order to keep things flowing along smoothly.

So I guess that's kind of it. If you want your life to go smoothly and stay manageable, make sure you're aware of any positive feedback loops that you may have and be vigilant about keeping their parameters in bounds (at levels that don't produce further change). To go a step further, try to set up your processes in such a way that they are governed by negative feedback. Even if it doesn't work, you'll be happy you thought about it, I promise. In short, everything I needed to know about life I learned from Gerry Sussman and a MOSFET.


TheSuperUser said...

Thanks for your post. I have the 046 Final tomorrow and I google "6.046 final". Now I'm truly terrified!!

-Drew D. MIT Course 6.2 '13

TheSuperUser said...

^grammar fail. Mind. too. tired. to. think.

jesse8888 said...

Best of luck with the final, Drew. I just checked and this post shows up as the 5th result for that search, which is either a massive fail on Google's part or a monstrous win for me. Actually it's probably both.

avoidthe9to5 said...

epic post. tyvm for your thoughts!

Dave said...

6.002 was my least favorite class at MIT. Why I didn't take it freshman year (so it would be pass / no record, and leave the C off my record, I have no idea). 6.046 was one of my favorites. I'm only writing this to piggyback on your Google rank.

Captain R said...

Who puts decimal points in their course numbers? IMR.002.

Matt said...

This post is one of your best recently. Nicely put, great examples, inherently useful.