= 27d567bc2d48d9f40e9ccc20ea370b15)
Esar comments on The Power of Reinforcement - Less Wrong
You are viewing a comment permalink. View the original post to see all comments and the full post content.
= 783df68a0f980790206b9ea87794c5b6)
Loading…= b715d02e85f7b3b4ae65ca3d3e450868)
Subscribe to RSS Feed
Powered by Reddit
= f037147d6e6c911a85753b9abdedda8d)
You are viewing a comment permalink. View the original post to see all comments and the full post content.
Comments (467)
That's a fair point: I may have been treating a conditional like a bi-conditional. I think my sense of the matter is this: if a friend told me that he spent a lot of our time together thinking through ways to positively reinforce some of my behaviors, even to my benefit, I would become very suspicious of him. I would feel that I'd been treated as a child or a dog. His behavior would seem to me to be manipulative and dishonest, and I think I would feel this way even if I agreed that the results of his actions were on the whole good and good for me.
Do you think this sort of reaction on my part would be misguided? Or am I on to something?
I agree with you that your autonomy is threatened by the manipulations of others. But threats only sometimes turn into harm- distinguishing between manipulations you agree with and disagree with is a valuable skill.
Indeed, there's a general point that needs to be made about human interaction, and another about status, but first a recommendation: try to view as many of your actions as manipulations as possible. This will help separate out the things that, on reflection, you want to do and the things that, on reflection, you don't want to do. For example:
Emphasis mine. The reaction- of calling his behavior manipulative and dishonest- feels like it punishes manipulation, which you might want to do to protect your autonomy. But it actually punishes honesty, because the trigger was your friend telling you! Now, if your friend wants to change you, they'll need to try to do it subtly. Your reaction has manipulated your friend without his explicit consent- and probably not in the direction you wanted it to.
So, the general point: human social interaction is an incredibly thorny field, in part because there are rarely ways to learn or teach it without externalities. Parents, for example, tell their children to share- not because sharing is an objective moral principle, but because it minimizes conflict. As well, some aspects of human social interaction are zero sum games- in which people who are skilled at interaction will lose if others get better at interaction, and thus discourage discussions that raise general social interaction skills.
The status interpretation: generally, manipulation increases the status of the manipulator and decreases the status of the manipulated. Resistance to manipulation could then be a status-preserving move, and interest in manipulation could be a status-increases move. What articles like this try to do is lower the status effects of manipulation (in both directions)- Luke proudly recounts the time Eliezer manipulated him so that he could better manipulate Eliezer. If being molded like this is seen more positively, then resistance to being molded (by others in the community) will decrease, and the community will work better and be happier. As well, I suspect that people are much more comfortable with manipulations if they know how to do them themselves- if positive reinforcement is a tool used by creepy Others, it's much easier to dislike than if it's the way you got your roommate to finally stop annoying you.
This, with extra emphasis!
I'm confused, not only by the beginning of this comment, but by several others as well.
I thought being a LessWronger meant you no longer thought in terms of free will. That it's a naive theory of human behavior, somewhat like naive physics.
I thought so, anyway. I guess I was wrong? (This comment still up voted for amazing analysis.)
Autonomy and philosophical free will are different things. Philosophical free will is the question "well, if physical laws govern how my body acts, and my brain is a component of my body, then don't physical laws govern what choices I make?", to which the answer is mu. One does not need volition on the level of atoms to have volition on the level of people- and volition on the level of people is autonomy.
(You will note that LW is very interested in techniques to increase one's will, take more control over one's goals, and so on. Those would be senseless goals for a fatalist.)
Thanks for clarifying that. I should note that I am very interested in techniques for self-improvement, too. I am currently learning how to read. (Apparently, I never knew :( ) And also get everything organized, GTD-style. (It seems a far less daunting prospect now than when I first heard of the idea, because I'm pseudo-minimalist.)
I still am surprised at the average LWers reaction here. Probably because it's not clear to me the nature of 'volition on the level of people'. Not something to expect you to answer, clarifying the distinction was helpful enough.
I think it's misguided personally. You're already being manipulated this way by your environment whether or not you realize it.
Well, I'm claiming that this kind of manipulation is often, even characteristically, unethical. Since my environment is not capable of being ethical or unethical (that would be a category mistake, I think) then that's not relevant to my claim.
I was referring though to the case of your friend using reinforcement to alter your behavior in a way that would benefit you. I just have a hard time seeing someone trying to help you as an unethical behavior.
It does depend on whose definition of 'help' they're using.
Good point. Do you think it would be ethical if they were helping to fulfill your preferences?
Usually, yes, though there are several qualifications and corner cases.
Agreed, there probably are.
That's fair. I should tone down my point and say that doing this sort of thing is disrespectful, not evil or anything. Its the sort of thing parents and teachers do with kids. With your peers, unsolicited reinforcement training is seen as disrespectful because it stands in leau of just explaing to the person what you think they should be doing.
In my experience, telling other people how I think they should behave is also often seen as disrespectful.
Often it is, we agree. But it's the 'telling' there that's the problem. A respectful way to modify someone's behavior is to convince them to do something different (which may mean convincing them to subject themselves to positive reinforcement training). The difference is often whether we appeal to someone's rationality, or take a run at their emotions.
I agree that there are respectful ways to convince me to do something different, thereby respectfully modifying my behavior.
Many of those ways involve appealing to my rationality.
Many of those ways involve appealing to my emotions.
There are also disrespectful ways to convince me to do something different.
Many of those ways involve appealing to my rationality.
Many of those ways involve appealing to my emotions.
So, by 'appealing to someone's rationality' I mean, at least, arguing honestly. Perhaps I should have specified that. Do you still think there are such examples?
Do I think there are disrespectful ways to convince me to do something different that involve arguing honestly? Sure. Do you not?
Well this runs into the problem of giving unsolicited advice. Most people don't respond well to that. I think it's probably difficult for most rationalists to remember this since we are probably more open to that.
Not really. Rationalists are just open to different advice. There's lots of advice rationalists will reject out of hand. (Some of which is actually bad advice, and some of which is not.)
Everyone believes themselves to be open-minded; the catch is that we're all open to what we're open to, and not open to what we're not.
This feels like an equivocating-shades-of-grey argument, of the form 'nobody is perfectly receptive to good arguments, and perfectly unswayed by bad ones, therefore, everyone is equally bad at it.' Which is, of course, unjustified. In truth, if rationalists are not at least somewhat more swayed by good arguments than bad ones (as compared to the general population), we're doing something wrong.
It amuses me how readily my brain offered "I am not neither open-minded!" as a response to that.
Well I agree that none of us is completely rational when it comes to accepting advice. But don't you think rationalists are at least better at that than most people?
But your environment includes people, dude.
This shouldn't be a puzzle. Reinforcement happens, consciously or subconsciously. Why in the name of FSM would you choose to relinquish the power to actually control what would otherwise happen just subconsciously?
How is that not on the face of it a paragon, a prototype of optimization? Isn't that optimizing is, more or less-consciously changing what is otherwise unconscious?
Oh, you're definitely on to something, and it's something important.
That said, I don't think what you're on to has to do with whether and when it's ethical to manipulate people's behavior.
So what am I on to then?
Roughly, that we often respond to others' ability to cause us harm (whether by modifying our behavior or our bank accounts or our internal organs or whatever other mechanism) as a threat, independent of their likelihood of causing us harm.
So if you demonstrate, or even just tell me about, your ability to do these things, then while depending on the specific context, my specific reaction will be somewhat different... my reaction to you knowing my bank PIN number will be different from my reaction to you knowing how to modify my behavior or how to modify the beating of my heart or how to break into my home... they will all have a common emotional component: I will feel threatened, frightened, suspicious, attacked, violated.
That all is perfectly natural and reasonable. And a common and entirely understandable response to that might be for me to declare that, OK, maybe you are able do those things, but a decent or ethical person never will do those things. (That sort of declaration is one relatively common way that I can attempt to modify your likelihood of performing those actions. I realize that you would only consider that a form of manipulation if I realize that such declarations will modify your likelihood of performing those actions. Regardless, the declaration modifies your behavior just the same whether I realize it or not, and whether it's manipulation or not.)
But it doesn't follow from any of that that it's actually unethical for you to log into my bank account, modify my heartbeat, break into my home, or modify my behavior. To my mind, as I said before, the determiner of whether such behavior is ethical or not is whether the result leaves me better or worse off.
Breaking into my home to turn off the main watervalve to keep my house from flooding while I'm at work is perfectly ethical, indeed praiseworthy, and I absolutely endorse you doing so. Nevertheless, I suspect that if you told me that you spent a lot of time thinking about how to break into my home, I would become very suspicious of you.
Again, my emotional reaction to your demonstrated or claimed threat capacity is independent of my beliefs about your likely behaviors, let alone my beliefs about your likely intentions.
This seems very implausible to me. I often encounter people with the ability to do me great harm (a police officer with a gun, say), and this rarely if ever causes me to be angry, or feel as if my dignity has been infringed upon, or anything like that. Yet these are the reactions typically associated with finding out you've been intentionally manipulated. Do you have some independent reason to believe this is true?
Yes, but no reasons I can readily share. And, sure, I might be wrong.
I don't think I would be suspicious of him, as long as I agreed with the behaviours he was trying to reinforce. (I don't know for sure–my reactions are based only on a thought experiment.) I think I would be grateful, both that he cared enough about me to put that much time and effort in, and that he considered me emotionally mature enough to tell me honestly what he was doing.
However, I do think that being aware of his deliberate reinforcement might make it less effective. Being reinforced for Behaviour A would feel less like "wow, the world likes it when I do A, I should do it more!" and more like "Person X wants me to do A", which is a bit less motivating.
Really? So say I tell you that all those times that I smiled at you and asked how you were doing were part of a long term plan to change the way you behave. The next day I smile and ask you how you're doing. Has my confession done nothing to change the way you think about my question?
I'm saying that things like smiles and friendly, concerned questions have a certain importance for us that is directly undermined by their being used for for the purposes of changing our behavior. I don't think using them this way is always bad, but it seems to me that people who generally treat people this way are people we tend not to like once we discover the nature of their kindness.
Like I said, thoughts experiments about "how would I feel if X happened" are not always accurate. However, when I try to simulate that situation in my head, I find that although I would probably think about his smile and question differently (and be more likely to respond with a joke along the lines of "trying to reinforce me again, huh?") I don't think I would like him less.
Anyway, I think I regularly use smiles and "how are you doing?" to change the way people behave...namely, to get strangers, i.e. coworkers at a new job, to start liking me more.
Well, I guess I'll tap out then. I'm not sure how to voice my position at this point.
Your position is that you have a certain emotional response to knowing someone is trying to modify your behaviour. My position is that I have a different emotional response. I can imagine myself having an emotional response like yours...I just don't. (Conversely, I can imagine someone experiencing jealousy in the context of a relationship, but romantic jealousy isn't something I really experience personally.) I don't think that makes either of us wrong.
Well, my position is that doing things like asking how someone is doing so as to reinforce behavior rather than because you want to know the answer is ethically bad. I used the example of the friend to try to motivate and explain that position, but at some point if you are totally fine with that sort of behavior, I don't have very much to argue with. I think you're wrong to be fine with that, but I also don't think I can mount a convimcing argument to that effect. So you've pretty much reached the bottom of my thoughts on the matter, such as they are.
I'm curious about whether your reasons for considering this kind of behaviour "unethical" are consequentialist (i.e. a world where people do X is going to be worse overall than a world where no one does X) or deontological (there are certain behaviours, like lying or stealing, that are just bad no matter what world they take place in, and using social cues to manipulate other people is a behaviour that falls into that class.)
Ah, I'm not a consequentialist or a deontologist, but I do think this is a case where intentions are parcticularly important. Doing this kind of reinforcement training to someone without their knowledge is characteristically disrespectful if you just do it to help them, but it may also be the right thing to do in some cases (I'm toning down my claim a bit). Doing it with the result that they are harmed is vicious (that is, an expression or manifestation of a vice) regardless of your intentions. So that puts me somewhere in the middle.
I wouldn't necessarily say that. Doing it when you know they don't (or would not) want you to is disrespectful.
This definitely seems false. It is the expected result, given information that you have (or should be expected to have) that can indicate viciousness, not actual results. For example, I could reward my children such that they never Jaywalk (still not quite sure what this is) and only cross the road at official crossings. Then one of my children gets hit by a car waiting at a crossing when they would have been fine crossing the street earlier. I haven't been vicious. My kid has been unlucky.
It the general case it is never the result that determines whether your decision was the right decision to make in the circumstance. It is the information available at the time. (The actual result can be used as a proxy by those with insufficient access to your information at the time or when differing incentives would otherwise encourage corruption with 'plausible deniability').
Can you express your personal ethics explicitly and clarify where it comes from?
I'd be happy to try. Do you want a brief account specific to this topic, or something more general?
If you could trace your ethics backward from "it's unethical when people consciously use punishment/reward system to modify my behavior to their liking" to some basic ideas that you hold inviolate and cannot further trace to anything deeper, I'd appreciate it.