Epistemic status: Original work, explanation of a mental model that I developed for a few years that brings together knowledge from existing fields.
Is all communication manipulation? I hear this sentiment frequently expressed and want to explain in this article that there’s nonmanipulative communication by using protein folding as an intuition pump.
It is common knowledge within molecular biology that proteins fold into their native state. That native state is the folded shape that possesses a minimum of free energy. Finding global minima is however a hard problem. For bigger proteins, it's at the time of writing - still impossible to calculate the shape.
Even in vivo protein folding is a hard problem. Cells are densely packed with many different molecules that push against each other. Frequently, resources are wasted when a protein misfolds into a shape that's not its native state.
Nature is clever and developed a way to help proteins fold into their native state. Cells produce chaperones. A chaperone surrounds an unfolded protein to protect it from outside influences to help the protein to fold into its native state. A chaperone doesn't need to know the native state of a protein to help the protein fold into that state. Instead of manipulating the protein like a sculpture, it holds space for a protein to be safe from outside influences, while it folds into its native form.
This allows a chaperone that works in an uncomplicated way to achieve a result that very complex machine learning algorithms currently don't achieve. The machine learning algorithm tries to figure out the best way for the protein to fold while the chaperone just lets the protein find this way by itself.
The psychologist Carl Rogers advocated that good psychologists act in the same way nonmanipulative with their patients. In his view, it's not the job of the therapist to solve the problem of their patient by manipulating the patient into a healthy form. A good therapist isn’t like ta sculptor sculpts a sculpture. The job of the therapist is rather to hold a space for the patient in which the patient is safe from certain forces that prevent the patient from finding their healthy authentic native state.
I don't intend to argue for nonmanipulative communication from a moral perspective. In cases where you know how to fix the problem of the person you are talking with and are confident that the other person will follow your advice, go ahead. If you don't know what will help a person, taking a nonmanipulative approach is often more effective than giving the person advice that they have already heard a dozen times.
If you tell an obese person that they should lose weight again, you add additional stress which can make it harder for them to think about the issue. In the Rogerian model effective change isn't about creating enough pressure by telling the obese to lose weight till they finally get it. For an obese person who feels shame for being obese, it can be hard to clearly think about the issue when they are alone. Providing the person a space where they can speak about their challenges in a way where they aren't feeling judged can help them to make progress for themselves.
There's a mystic quality to being nonmanipulative. Even Carl Rogers, who proposed the ideal, that all interactions should be nonmanipulative, sometimes fell short of it. For practical purposes it's often more useful to do what makes sense in the moment and what helps the other than to live up to an ideal of being perfectly nonmanipulative.
On the other hand, having a mental model of what it means to be nonmanipulative can be very helpful to understand communication practices like Rogerian psychotherapy, Gestalt Therapy and Circling.
I invite you to explore communicating in a way that holds the space for others to find themselves.
Your title and opening sentences make me think you want to convey the idea that the phrase "non-manipulative communication" means exactly what the literal words the phrase is made up of mean. I do not think you made the case that that is so.
These sentences seem to be trying to put tension between the machine learning algorithm and the chaperone. However, it is not clear to me that the result achieved by the chaperone is the same as the result machine learning algorithms attempt to achieve.
Does the chaperone "know" in what way the protein folded itself? Can we interrogate the chaperone to learn about the protein? I think not. Neither the chaperone nor the protein has an inkling about the other...nor could they even if we grant them magical sentience or agency.
A physical process that emulates the result a ML algorithm is going for would seemingly encompass much more than just the chaperone. To me, if you really wanted to analogize chaperones to something somewhat apropos, it seems to be more analogous to some small component of some ML algorithm than it is to the ML algorithm itself.
Unlike humans, when it comes to agency and intent, the protein and chaperone do not have any.
For these reasons, this does not seem like an intuition pump that gets me to an understanding of the type of communication you're talking about and I do not think you've made an argument that "non-manipulative communication" is non-manipulative. I think you completely sidestepped what your opening seems to promise an elucidation of.
I want to note that I haven't made any claims about whether or not "non-manipulative communication" actually is or is not a literally correct phrase. I've given almost no thought to it, which is why I was interested to read this post when I saw the headline on my RSS feeds.
The following is more of an aside or addendum that is unrelated to the previous part of my comment:
Even if all communication actually is manipulative, we may want to, almost tautologically, define the phrase to mean the type of communication you're describing. This is sometimes a useful thing to do. I agree that the type of communication you describe is good and useful and something we should have in our toolbox.
I actually think I've got a pretty good grasp on what is meant by "non-manipulative communication", and I think it's an important and useful mode of communication for humans. As already mentioned, I've not really given the subject any thought, but as of right now, I don't think that phrase is a literally correct usage of the words "non-manipulative" and "communication".
I also think that's OK.
Does the chaperone "know" in what way the protein folded itself?
That's part of the point that the chaperone intervenes in a way that requires no knowledge about the folding or the outcome of the folding and still provides a good result. That's similar to nonmanipulative therapeutic interventions. While a good manipulative therapeutic intervention requires knowledge of what you want to change nonmanipulative interventions don't.
Actually, by your description I don't think the chaperone intervenes with the protein at all. There does not seem to be any communication from the chaperone to the protein. The chaperone intervenes with the environment surrounding the protein.
The closest analogy I can think of that seems to match, is a therapist communicating with everyone around their patient without actually communicating with the patient and keeping it a secret from the patient that they did so.
I'm not sure that is a useful definition of non-manipulative communication.
The chaperone intervenes with the environment surrounding the protein.
No, the chaperone is basically the full enviroment surrounding the protein while it folds. In the moment in which the protein folds the chaperone is it's enviroment just like the therapist sets the enviroment during a session with the patient.
It's a concept from which useful distinctions are drawn in some areas of therapy.
Some areas of osteopathy seem to use the same notion of nonmanipualtion which might be independent from Rogers.
No, the chaperone is basically the full enviroment surrounding the protein while it folds.
Perhaps you can expand on this because I do not see how it's functionally different from what I said. It becomes the full environment by intervening with the protein's environment. It cannot become the full environment without intervening with the protein's environment.
In the moment in which the protein folds the chaperone is it's enviroment just like the therapist sets the enviroment during a session with the patient.
...and thus I do not see how it's "just like" what a therapist does...at least if we're talking about the ways in which the therapist communicates with the patient.
I understand the intention of the therapist is to be like the chaperone. But your analogy seems to be between the chaperone and what the therapist actually does.
This is not to say that the therapist can or cannot communicate with the patient without manipulation, only that that the therapist actually does communicate with the patient and the chaperone does not.
This might be true. However, your post seems to be making the argument that the type of communication a therapist participates in is literally nonmanipulative and I do not think that is the same argument you make with this sentence.
Perhaps you can expand on this because I do not see how it's functionally different from what I said.
You spoke about the equivalent would be the therapist talking to people in the enviroment of the patient that are external to the therapist. A chaperone doesn't change things in the enviroment of the protein that are external to it to make the enviroment interact with the protein in a good way.
There are reasons why the phrase holding space is frequently used to describe this kind of communication as something that the therapist does.
There are things in the field of alternative communication that are hard to communicate. I'm not sure whether there's much more that I can say at this point if what I have already written doesn't bring the idea across.
therapist participates in is literally nonmanipulative
Words have a bunch of literal meanings. The literal meaning that it's about is (3) at webster: ": "to change by artful or unfair means so as to serve one's purpose". In this case the therapist doesn't have a particular purpose towards which they want the patient to change.
I'll make this last comment to clarify my position and if you want to reply, I'll let you have the last word (I say this with sincerity and not in a passive-aggressive manner!)
First of all, I feel like you're continuing to defend the idea of nonmanipulative communication. To make it clear, I'm not questioning whether it exists or is useful or anything at all. I'm questioning the idea that the chaperone-protein analogy is actually analogous to any sort of communication.
I don't feel as if that's exactly material to the point at hand. The main point is that the chaperone doesn't interact with the protein in any way. It's impossible for a human to be like the chaperone and for the human to communicate with the "protein".
However, I will point out that I don't mean to claim exactly what you seem to think I mean to claim. My only claim is that the therapist interacting with people other than patient, without interacting with the patient, would be somewhat analogous to the chaperone. That is as far as it goes. That doesn't go far enough to become a useful analogy because the chaperone - protein relationship is not equivalent to any sort of communication.
There are reasons why the phrase holding space is frequently used to describe this kind of communication as something that the therapist does. There are things in the field of alternative communication that are hard to communicate. I'm not sure whether there's much more that I can say at this point if what I have already written doesn't bring the idea across.
I think you're still sidestepping the point here. "Things in the field of alternative communication" have almost no bearing on the point of my comments.
My whole point is that the chaperone-protein "relationship" is not communication at all. There is no special type of communication that is not communication.
(You can probably make the argument that the protein communicates one-way with the chaperone. How does the chaperone "know" where to be? I do not know. However, this is impossible to analogize with the type of communication you're analogizing with.)
In this case the therapist doesn't have a particular purpose towards which they want the patient to change.
Sure, I agree.
My comments do not attempt to dispute that. My point is that, I do not think you made the case for this definition of (or any of the definitions of) "manipulative" because 1) the chaperone is not analogous to communication of the type you describe and 2) your post largely hangs on this analogy.
If you take away the analogy, your post amounts to the assertion that non-manipulative communication exists.
Right. I guess my point is that that seems to make comparing the chaperone to the ML algorithm a non-starter.
While I wasn't making this point in my comment, I also think it doesn't seem likely a good analogy to nonmanipulative conversation since the participants in the nonmanipulative conversation are never in a similar state of ignorance. Even if you're talking to a complete stranger and trying to be nonmanipulative.
You might be able to emulate such a state, but your post makes no argument to that effect.
Do you have an operational definition of "manipulation" that can help me understand what you're asserting here?
The chaperone manipulates the environment to let the protein fold into a lower-energy state than it would without the chaperone. Is this not manipulation of the end-state of the protein? I interpret the cultural norm that one is to be "nonmanipulative" as "please only cause change in ways that we approve of", not as "have no impact at all".
I don't believe there is any communication which doesn't change the future state of the universe, so I think I'm unclear what "manipulate" means.
Manipulation is when you have a specific outcome in mind and exert power on a system to move into that outcome. Chaperones don't have any idea of how the want the final shape of a protein to look like.
It's possible for a protein to fold into the lowest-energy fold without a chaperone, it's just that frequently the pressure that exist inside the intercellular fluid get a protein to misfold. For proteins misfolding has a straightfoward definition, it's derivation from the lowest-energy fold and that definition is not dependent on the existence of chaperones. Early on in evolution there were no chaperones and cells had to deal with having more misfolded proteins.
Nonmanipulation is about allowing self-regulation processes to determine the resulting shape instead of a therapist working towards the patient changing into the shape that the the therapist things would be better for the client.
Not sure I follow - chaperones don't seem complex enough to have intent at all, so by that definition they are non-manipulative in the same sense that rocks are - it's a concept that doesn't apply to them, not something they could do and choose not to.
That's a big contrast with human communication - there is definitely intent behind every communication. For this kind of action, the selective removal of forces seems near-indistinguishable from the selective addition of force in order to enable/influence some change. It feels like there's a naturalistic fallacy going on - some underlying belief that what happens in a vacuum is better than what happens in a real equilibrium.
What about "communication" with a program?
On this site, concerns about the (theoretical) manipulative abilities of superhuman AI seem to be mentioned fairly often. (Facebook algorithms/worry about twitter (usually more about observed results/mechanism design-ish) come up less often, but are mentioned.)
What do you want out of the algorithms you interact with, say underlying social media?
Thanks, this analogy is interesting.
(Unless they are bad, bad chaperones who went all prion on everybody else. Not that that makes their action manipulative, they simply... let things take shape the way they secretly want to be.)