With many thanks to my wife for listening to me complain about people being wrong, and for her wisdom about ways to be right.
Alfie Kohn wrote Punished by Rewards over 30 years ago; it was published in 1993. In large part, it’s a direct response to the behaviorism of B.F. Skinner, and the book starts with that. Reading this book from 2026, it’s hard for me to understand everything that Kohn is reacting to. I think a lot of behaviorism has either fallen away or become so omnipresent as to be invisible. This makes the book a bit like reading a history of a long-ago war: I find both belligerents slightly alien, but I do hope to learn something about their philosophy and strategy.
Behaviorism as described by Kohn strikes me as slightly Buddhist: there’s no such thing as an internal “I”, just the environmental contingencies and conditioned responses. Skinner apparently went on to say that he himself had no agency. I find this pretty wild, and Kohn seems to find it funny. In spite of how weird it is to me, I want to take it at face value to avoid the typical mind fallacy.
The kind of psychology created by someone with no sense of self would obviously not focus on any kind of self-definition or agency. Behaviorism is all about responses that people have to stimulus. If you can train a dog or an LLM through reinforcement learning, surely you can train a human in the same way.
Alfie Kohn is a person who is incredibly driven by his own moral compass. I opened Punished by Rewards expecting a book about how to incentivize behavior, and what I got was mainly a polemic about the evils of manipulation and control. The book does have a lot of details about how rewards and punishments can fail, but even when they work Kohn will attack them as immoral.
If BF Skinner is someone with no sense of self who sees education as a training program of operant conditioning, then Alfie Kohn is his mirror. Kohn is a person with an extreme sense of self who sees education as a way infusing students with the right kind of self. I imagine that Skinner would say it doesn’t matter if someone enjoys doing homework as long as they do it, while my imagined Kohn says that whether they enjoy it is the only thing that matters.
Kohn’s book is not simply a rejection of punishments and rewards, nor a celebration of personal autonomy and selfhood. It tries to show a path to a new way of organizing life. He explicitly wants to call into question our most foundational institutions. He wants to push people to live in a totally different way that more fully embodies the virtue of caring for others. He wants us to reject any kind of interaction that doesn’t serve that goal. In arguing for this, he suggests that we throw out many of the things that allow a pluralistic society to function.
My Purpose in Reading the Book
I have a few goals here:
- Argue that Alfie Kohn is wrong in his rejection of institutional pressure to motivate people.
- Sift through Kohn’s book to find the ways that rewards actually can be useful, according to his description of their function and my definition of being useful.
- Discuss the nature of virtue, and encourage people (you!) to be more virtuous in ways that include what Kohn cares about.
I have two kids, and I want to raise them as well as I can. I want to do everything I can to make their world as caring, wonderful, and interesting as it can be. I want to have a large impact on their lives, and I want for them to come to value the same things that I value and to be good people. I believe that necessarily means I want to change the way their brains work, but I want to do this in the most kind and least coercive way that I can. This seems like something Kohn desperately wants to support, and the last chapter of his book is all about how to do it.
I want to support my kids’ childhoods in such a way that they grow up to be people that I would be proud of and would like to know.
When I started the book, I was hoping to get a better handle on how to structure rewards to inform but not manipulate. I was hoping to understand when rewards degrade intrinsic motivation and when they don’t. Basically, I wanted to sharpen a tool in my toolbox. I had a small expectation that maybe the book would convince me to give up rewards for something new, if it argued well enough that rewards were counterproductive.
After a few chapters I realized that the book was arguing that rewards would work sometimes, but I should never use them. I realized that it was making serious moral claims that I didn’t always agree with, and doing it in ways that were sometimes manipulative in their own right.
The Moral of the Story
Punished by Rewards is a book about morality. It doesn’t use these words, but my take on the book is that Kohn sees behaviorism as akin to deontology, and Kohn advocates for virtue ethics.
Deontology just means “ethics is following the rules”. Different deontologies have different rulesets, but fundamentally they all just agree that you should submit to the ruleset to be a good person. If that’s what you think ethics is, then behaviorism makes a lot of sense. Reinforcement and conditioning can get people to follow rules, so you’re done.
Virtue ethics is more complex. It’s not about “following the rules” so much as it’s about “being a good person internally.” In deontology, it doesn’t really matter why you follow the rules. In virtue ethics, motivation and internal mind state are the most important things. Kohn spends a lot of time talking about how rewards might get people to do something, but they don’t make someone into the kind of person who would want to do the thing outside of the reward context.
The first part of Kohn’s book is dedicated to showing how rewards and punishments are manipulation, and are omnipresent in society. Kohn is both anti-coercion and anti-hierarchy. It’s such a huge thing for him that I took a break from the book at one point to look up whether he was an anarchist (while he did give interviews on anarchist podcasts, he claims to be a democratic socialist). Behaviorism, according to Kohn, is inimical to democracy and is all about controlling others.
Kohn’s morality is very strict. He will rail against the slightest whiff of manipulation, coercion, or hierarchy. I find this kind of infuriating, because his book is itself incredibly manipulative. As far as I can tell, Kohn never lies and never stretches the meaning of a study. What he does do is argue everything in the most manipulative way possible. He’ll make up examples about how pay-for-performance is wrong because someone suffering from a compensation plan at work might come home and abuse his child. He’ll come right up to arguing for something that’s not supported by the evidence, then gesture at it as though it must be true. When Kohn doesn’t have solid evidence against something, he’ll often just say there are “serious doubts” about it and switch topics.
I find what Kohn does to be much more manipulative than what he’s arguing against. Instead of changing payoffs for my actions, the book presents information in a way that seems intentionally geared to confuse me about what the payoffs of my actions might be. If I follow Punished by Rewards, then I would no longer be able to make my own decisions about what to do. Instead I’d be making Kohn’s decisions and I’d be often confused and surprised about the outcomes.
Kohn’s arguments generally assume the worst of anyone he’s arguing against. He also seems incapable of dealing with tradeoffs. Either something supports his virtue-based agenda, or it is bad and should be removed. The idea that something could be useful for a different but also good goal is totally absent from the book.
This is really frustrating because I am so sympathetic to the underlying arguments about virtue and agency. Kohn wants to help kids develop agency. He wants them to be good and decent people outside of following a set of rules. Kohn sees people as having the “capacity for responsible action, the natural love of learning, and the desire to do good work.” He wants to teach people in a way that fosters those virtues, instead of smothering them. That is all wonderful and I want to support it, but Kohn’s manipulative arguments and confusion between moral and pragmatic choices make that difficult for me.
Supporting Trust
Let’s talk about computer security. Actually no, let’s talk about trust. If you work in computer security for long enough, you end up thinking a lot about what it means to trust someone, and what it means to support people in performing tasks while preventing them from abusing infrastructure. Bruce Schneier is a security researcher who started writing cryptography books decades ago and has slowly migrated to writing books about society and trust.
Bruce Schneier’s book Liars and Outliers is about security and trust. It makes the argument that society needs enormous amounts of trust to function, and that trust is fostered through a variety of means. Schneier sorts the ways that society supports trust among its members into four different “pressures” that push people to act well. Even though Kohn’s book was written decades before Schneier’s, Punished by Rewards takes a strong stance about Schneier’s pressures.
- At the base is moral pressure: people do the right thing because they know it’s right. People have an idea of what it means to “be a good person” which they adhere to. Kohn wants all interactions with all people to be fundamentally based around this.
- Next is social pressure: people do the right thing because they want to be liked. If you jump off a bridge or donate to charity because all your friends are doing it, it’s social pressure that pushed you into it. Kohn thinks this is manipulative and wants teachers and parents to avoid it.
- The third thing impacting how people behave is institutional pressure: the rules, rewards, and punishments that constrain behavior. If you think about cheating on your taxes but decide not to because the fine when you get caught would be too high, then institutional pressure has done its job. Most of Kohn’s book is written against this layer of trust maintenance.
- Finally, if none of the above layers cause somebody to do the right thing, security is the set of tools we have to physically prevent someone from doing harm. Kohn is ok with this layer when it comes to preventing toddlers from running into the street, but it’s not clear how much further he’s willing to go in accepting this.
If I squint a little, I can map virtue ethics to Schneier’s moral pressure and deontology to Schneier’s institutional pressure. What I really like about Schneier’s book is that he acknowledges the use of and need for all of these things. Obviously we want everybody to do the right thing for moral reasons, but we can’t depend on it. We need the other layers in order to enable the kind of society that we have. If you interact regularly with people you don’t know closely, all layers of this stack are critical.
Punished by Rewards is about raising children and motivating adults in the workplace. It makes sense that it focuses mainly on Schneier’s moral pressure layer, since it’s about interacting with people you know well. Where Kohn’s book goes wrong is in rejecting the other three layers as immoral because they are controlling. Those layers allow a pluralistic society, and it turns out that schools and the workplace are two very pluralistic institutions. We want our kids to learn how to interact well with all of these layers.
Kohn is a black and white thinker. He identifies that kids need to be taught virtue. He sees that rewards and punishments don’t do that. He concludes that rewards and punishments are therefore bad. What he is not seeing is that these other methods of supporting trust are valuable on their own. They form a backstop to virtue, and help people interact with each other in safe and sane ways all the while everyone is at different places in their virtue-learning journey.
Rewards, Punishments, and Consequences
Let’s get one thing out of the way up front: it can be hard to define what is a reward. It seems like it should be easy, but the edge cases are thorny and depend on a lot of context and unobservable thought processes.
Imagine this: a parent and their child are in their living room. Next to them is a pile of LEGOs on the floor. The parent says, “if you clean up your LEGO mess, then we can go get ice cream.”
Depending on context, this could be reward, a punishment, or a consequence. Let’s see how:
- Reward: The parent wants the kid to clean up, and is explicitly thinking about ways to convince the kid into doing it. The parent intends ice cream as a reward. The ice cream is only going to happen because the kid cleans up.
- Punishment: The parent is upset that the kid made a mess. They had a plan to go get ice cream, but now the parent is threatening to withhold it. They want to cause a tiny bit of anguish in the kid in order to convince them to clean up.
- Consequence: The parent and kid have been working on personal responsibility. The kid knows that if they make a mess it’s their responsibility to clean it up. The parent doesn’t want their child to suffer in any way, and also doesn’t want to bribe them in any way. The parent just wants ice cream, similar to the kid. The statement the parent makes is simply informing the kid about what order the events are going to happen in.
What makes the above more complicated is that the kid may have a different view of what’s going on than the parent. I think it might be pretty common for a parent acting in a “consequences” mode to be interpreted in a “punishment” mode by the kid.
I’ve seen a lot of people on social media commenting that the word “consequences” is just a smoke screen for punitive parents to use to get away with abuse. Kohn not only agrees with this, his book might be the source of the argument. While I expect that some parents and teachers do punish under the guise of consequences, I actually think that many times parents (and maybe teachers too) are just exasperated and not sure how to convey the responsibilities that they are trying to instill into their kids. Correctly conveying if something is a reward, a punishment, or a consequence is subtle and changes dynamically as the kid grows and learns.
Kohn doesn’t see this subtlety. Instead, Kohn defines punishment as “an intervention that forces someone to do something she’d rather not do (or prevents her from doing something she wants to do) as a way of trying to change her behavior.” Kohn uses this definition to argue that all consequences are punishments.
One of his examples of a consequence that’s actually a punishment is “Food is taken away from children who persist in not washing up before dinner.” This hits close to home for me, because one of my kids went through a phase where he didn’t want to wash his hands after using the bathroom. Our entire family was getting sick way more often than normal, and we didn’t know why. At some point, I noticed that my kid’s hands smelled like poop while he was grabbing noodles with his hands and slurping them up. We discovered that he had been turning on the sink and then turning it off, without actually washing his hands.
The immediate consequence of my kid not washing his hands was that we all got sick all the time. We reinforced the rule that he had to wash up after using the bathroom. We made it clear that it was a rule about health, and that we were getting sick from germs. He still showed up to dinner with poop hands regularly. My wife and I made a new “consequence”. He had to wash hands, and they had to not smell like poop, before we’d let him eat dinner. This even resulted in us washing his hands for him a few times.
Kohn doesn’t expand on his argument that preventing a kid from eating with poop hands is punishment. He does have an argument about a similar type of health and safety activity though:
If a child tips her chair back too far, she will fall over. That is a “natural consequence”—and the fact that it qualifies for that label offers no argument for letting it happen; caring adults go out of their way to prevent many such consequences from occurring. If, by contrast, a child who tips her chair back is forced to stand up for the rest of the period, that is a punishment.
For my family, the natural consequence of our kid’s behavior was that we were all getting sick all the time. That sucked, and an artificial consequence helped turn that around. In fact, the artificial consequence was the way that prevented real negative consequences from occurring. What’s Kohn’s advice for how to get a kid to change their behavior without consequences? He has no specific advice beyond talking with the kid (though more on that later).
I find this confusing. Kohn wants to raise kids who can be responsible and self-directed. Kids sometimes make poor choices because they don’t look forward as far as adults do to see certain natural consequences. Parents can help prevent natural consequences by instituting artificial consequences with a shorter time horizon. Helping kids understand natural consequences via artificial consequences is helping them to become more self-directed.
If a natural consequence is near enough, and small enough, then maybe it makes sense to just let the kid learn about it directly. But we don’t let kids run into the street because the natural consequences are too big. We shouldn’t be letting them get themselves and then their whole family sick just to avoid any hint of “control”.
In fact, that’s exactly what removing a girl’s chair does if she keeps tipping it back. The point of a consequence is that the parent or teacher decides some activity is too risky (not washing hands, tipping a chair). Then they do what’s necessary to prevent the risk (prevent sickness, prevent a tip from turning into a fall). If the girl doesn’t have a chair, she can’t tip it. It’s not about making the girl suffer, it’s about preventing the risky behavior. The teacher should also explain the rule and motivate it, so the girl can learn why it can be dangerous instead of having to rely on institutional pressure to save her skull. Kohn doesn’t seem to see this. His list of consequences that are actually punishments is a mix of things that I’d consider to be obviously one or the other.
Kohn apparently didn’t have kids when he wrote the original book, and an afterward written years later does express the “consequences” formulation like I showed above in a more generous light. Then he goes on to say that he wouldn’t change a word of his book after having raised his own kids.
One additional point should be said about consequences in the form of evaluations. If a kid is performing some action that they could improve at, getting honest and detailed feedback will help them improve. Parents and teachers can give them this feedback, but it’s easy for feedback (consequence) to turn into praise (reward).
Kohn hates praise. He thinks it’s manipulative, and is against praising kids. Personally I like to let people know when I think what they’ve done is cool. Kohn advises parents to carefully consider a whole checklist of things before telling a kid that they like something. I think it’s fine to just say “I really like what you did there!” without having to interrogate your own emotional state to see if you might be subconsciously manipulating your kid. I do think it is worth checking to see if you are only praising your kid in order to reinforce certain behaviors, or if you’re just honestly sharing your opinion. If you are honestly sharing your opinion, I think the rest of Kohn’s checklist can suck it.
Some things that are often couched as “feedback” may not be consequence. Grades are a great example of this. Grades (in today’s schools) are a target. You get a good grade as a reward for doing what the teacher wants, and a bad grade as a punishment for failing. While meant as high level feedback, I don’t think they serve that purpose well. Kohn seems to agree, though I’ll talk more about that later.
Rewards and Payoffs
Alfie Kohn defines rewards as being manipulative. Rewards are the goodies that people dangle in front of others to get them to do something. They exist on Schneier’s social and institutional pressure layers.
This can make it a bit confusing to talk about why people do things. As an example, people sometimes argue that charity is selfish because when people give to a good cause then they experience the reward of feeling good about themselves. In other words, living according to your idea of virtue makes you feel good, and that good feeling is a reward. I find the actual argument tedious, and I only bring it up to focus on the wording used here.
This argument uses the word reward differently from how Kohn would use it. In this case it just means benefit, and it’s not necessarily about anyone external trying to change your actions. When discussing economics or game theory, the word reward is much more general than how Kohn uses it (though Kohn might disagree and say economists are all manipulative).
I’m going to use the word “payoff” to collectively represent all of the benefits (or costs) of an action. This might be the moral pressure payoff of feeling good for doing the actions of a good person. It might be the social pressure payoff of somebody being friendly towards you after you help them. It might be the institutional pressure payoff of getting a piece of candy for doing homework. All of these positive outcomes of an action are payoffs, but not all of them are what Kohn would call a reward.
When Rewards and Punishments Fail (and When They Succeed)
Kohn has several chapters dedicated just to research on rewards in different situations. He will willingly admit that they work well for a variety of tasks.
Want your kid to improve at multiplication? Reward will probably help there. Token economies, where kids perform behaviors to earn tokens redeemable for goodies, work quite well as long as the tokens keep flowing. Kohn says, very explicitly, that they can work quite well to get kids to do stuff. He just doesn’t care about when they work, because according to him “rewards must be judged on whether they lead to lasting change.” Obviously, I disagree. Rewards may work very well at exerting institutional pressure without being good for moral pressure. And that’s fine.
Kohn has a lot to say about when rewards fail. I’ll summarize it:
- Rewards work when they are present. If you take rewards away then the behavior will go back to normal.
- Rewards work when people can’t cheat them. If people are self-reporting to get rewards, you won’t get better performance.
- Rewards don’t make people more creative.
- Rewards don’t improve the best someone can do, and may slightly degrade peak performance.
Kohn often cites studies that compare people who do things for rewards to those who do them without rewards. He then compares the quality of the work that each group does (or sometimes compares learning rate). I don’t think this is particularly useful. Obviously if someone will do something without a reward then there’s no need to reward them for it! Even Skinner would say this, as rewarding for things that are already happening is just a waste of resources. I guess rewards for high performance were a big thing in the 80s, and he felt a strong need to argue against it.
What matters to me personally isn’t the performance of two groups doing the same thing for different reasons. What matters to me is whether rewards cause people to do things they otherwise wouldn’t. The times that I use rewards with my kids, it’s generally when they have a huge emotional aversion to doing something. Yes, I do want to address that emotional aversion directly, but that honestly seems like it’s going to take months for a lot of issues.
One of my kids used to get really uncomfortable and upset about his own writing. While I could have told my kid he doesn’t have to do any homework for the months it takes to understand why he didn’t want to write, things seemed to work better when we bribed him to write. For a long time, we were sitting right next to him and giving him one chocolate chip for every word he wrote. His homework got done, he got writing practice, and his skill at writing actually improved. That did lower the barrier to writing without rewards.
I agree that the writing my kid does when we give him a chocolate chip for every word is not high quality. It’s often not legible, or he intentionally writes words in a very silly way that takes up an entire page. That also doesn’t matter to me at all. I just want him to get used to writing, both emotionally and physically. Some magical world where he did writing without reward is not an option I actually had available to me. I tried literally dozens of things to understand the emotional block that he had to writing. I asked him, I told stories, I suggested possibilities, I tried changing the ergonomics of his pencils. My kid didn’t want to address it or share his struggle. Over the course of several months, the literal best thing to get him to write was direct rewards.
What’s happening here is that the chocolate chip rewards are institutional pressure. There’s a rule (write one word, get one chip). The rule changes the payoffs for doing homework. My kid does homework even though the moral pressure for it isn’t there. This makes doing homework less stressful, because my family can focus on the moral virtue of writing or learning independently of an assignment that’s due in two days. We do actually have to address the moral domain here, as the chocolate chips have no impact on it.
Kohn warns about this when he says that rewards don’t require finding out why the student isn’t doing the work in the first place. I think one of the major societal problems that the book points at is that people confuse institutional pressure for moral pressure. A lot of the failures of rewards come down to people expecting moral reform from institutional change. Instead of seeing this as effort on institutional pressure displacing effort on moral pressure, Kohn sees it as a direct undermining of the moral pressure. I don’t think his book is convincing of that.
In hindsight, I think one of the main reasons that the chocolate for writing plan worked so well for us is that writing words helped make my kid more comfortable with writing. The thing that we were rewarding was directly beneficial for him to be doing. I think this also explains why rewarding multiplication practice helps make people better at multiplication. The reward causes the practice, and the practice just is learning.
Kohn suggests that rewarding quantity not quality seems to work, but I find this particular point confusing. The parable of the pottery class is a story about a teacher splitting their class into two groups: one graded on quantity and the other on quality. It turns out that the group graded on quantity also ends up producing higher quality, with the moral of the story being that one should practice a lot to get good. I realize this is just a single example, but I think the reason it resonates so much with people is that a lot of folks have direct experience of doing something a lot and thus getting better at it. Just like with my kid and writing. It makes me wonder if the studies Kohn cites don’t last long enough to show this effect, or if something else is going on.
From Kohn, I think there are two high-level reasons that rewards fail. The first one is that people are not rewarding the right thing. This risk should be obvious to any programmer who’s ever implemented a reinforcement learning algorithm. It is Goodharting. For those non-programmers and non-economists in the audience, Goodhart’s Law says that “when a measure becomes the target, it ceases to be a good measure.” In the case of education, that means that if we start rewarding people for e.g. doing homework, then the amount of homework done won’t correlate as highly with how much people learn.
Kohn has a lot of examples of how rewards diminish learning. Offering rewards for work completed leads kids to:
- do easier work
- be less creative
- pay less attention to things unrelated to the rewarded work
None of that is at all surprising if we understand that kids are doing what they are rewarded for. If a kid is being rewarded for A, then that provides no additional incentive for them to do B. In fact, the opportunity cost for them to do B is higher, since A has more benefit than normal. Just because a teacher thinks B looks like A doesn’t mean kids will care. Kids are really smart! They can figure out what’s truly being rewarded.
While Goodhart’s Law dates back to 1975, it doesn’t appear explicitly in Punished by Rewards. This is unfortunate, because Kohn could have summarized at least a chapter of his book with just that one law.
The nice thing about Goodhart’s Law is that it points to the places where rewards can work. “What gets measured gets managed” is a corollary of Goodhart’s law. If a teacher can reward the specific behaviors they want, then it doesn’t matter that those behaviors become the target. This helps explain why rewarding my kid with one chocolate chip per word written worked. I wasn’t trying to reward quality or effort or word length. I literally just wanted him to write words.
The corollary to Goodhart’s law implies that we can incentivize the behaviors we want as long as we actually reward for that thing in particular. Should we actually do this, or are there other problems with rewards? This brings us to the second failure mode that Kohn talks about: external rewards undermine other reasons for doing things.
It’s common to know someone who decided they hated math because of school. Some even hate reading books because of school. Not many hate music due to school, but I wouldn’t be surprised if many did dread the idea of playing music themselves due to school. We simply have to address the fact that, while schools may improve people’s ability to do certain things, they also may cause a person to hate or fear doing those things. Is this because of the rewards and punishments used in school?
Closer to home, how worried should I be about bribing my kid to write using chocolate chips? After a few weeks, he seemed less averse to writing than when we started. Is this going to backfire in the long term and make him hate writing or homework overall?
Why do external rewards sap intrinsic motivation? Kohn suggests that rewarding people for something causes them to think that thing is less valuable. A student might think something like “if they have to pay me to do math, then math must be inherently worthless.” Rewards also encourage kids to focus on how well they are doing towards getting the reward, rather than on the activity being performed. This can distract them and keep them from noticing that the thing they’re doing actually can be fun. Both of these possibilities are concerning, but I don’t think they are insurmountable.
Many of the other arguments that extrinsic motivation degrades intrinsic motivation are fundamentally behaviorist. They boil down to the argument that rewards operant condition people into associating the activity with being controlled. What happened to Kohn’s celebration of human agency? Do we expect operant conditioning to work, or not? If so, then all we need to do is figure out the right way to apply rewards. If not, then why should I buy these arguments about intrinsic motivation being undermined?
Kohn has a whole chapter that’s just a laundry list of the limitations of rewards. I agree with some and disagree with others. I agree that rewards don’t usually address the root cause of a misbehavior. I don’t really agree that rewards are automatically and intrinsically a threat of punishment. I don’t agree that rewards are intrinsically destructive to relationships.
Overall, reading this book did convince me that rewards are more limited than I had thought. They really do only work on the level of institutional pressure or social pressure, and we do need to help people to develop intrinsic moral drives about engaging with their world. Rewards won’t do that, but that doesn’t mean they’re bad or useless. They can be a great institutional backstop to support other efforts at resolving problems on a deeper level.
Just a few weeks ago, my kid started doing his homework without chocolate chips. It’s no longer a huge slog to get him to write each word, he just sits down and does it. Family homework time has been transformed.
His mom asked him what changed, and he said that he “used to only do homework if it was so easy it could be done in one second, or if it was fun.” Now he does it even if it’s hard. This is a change at Schneier’s moral level. It wasn’t caused by the chocolate chips, but they didn’t prevent it either.
People Popping M&Ms into their Own Mouths
It’s a little funny to me that Kohn spends so much time railing against certain reward methods as intrinsically controlling and manipulative, because I have seen dozens of people trying desperately to use those methods on themselves. While I’ve never been to CFAR myself, enough of my friends have gone that I’ve learned a lot about their techniques from osmosis. I was also fairly active in the rationalist community for a number of years (right up until I had kids, ironically). The rationalists have a very strong interest in using the ideas of psychology to improve their own lives, and they are happy to use behaviorism to do it.
Kohn describes “popping M&Ms into your own mouth” as simply another way others try to control you. If a teacher gets you to give yourself a reward, she doesn’t have to watch you all the time. This exact behavior is literally something that I’ve heard rationalists recommend in order to fight procrastination. Other techniques suggested by the rationalists are things like setting up larger rewards after tasks, gamifying tasks, and setting up automated fines for failure. These all embrace behaviorism directly. I’ve tried them all, and generally think Kohn wouldn’t be surprised by my experience.
Beeminder’s fines for failure did help me back when I was in grad school to keep on track. Habitica’s gamified XP for task completion reminded me regularly to do flash cards, go grocery shopping, etc. Offering myself rewards like short vacations for finishing major projects did help me buckle down and grind. I didn’t enjoy any of it though, and I don’t do it any more. I don’t particularly think any of those behaviorist interventions changed my long term preferences or tendencies. They worked to get me to do stuff, but I didn’t then become more likely to do that stuff without reward.
That was useful when I really needed to study. Threshold effects on education are real. Putting in a bunch of effort to get my Master’s degree did open up lots of opportunities for me. I did learn a lot in the process. Putting in the time and doing the work was good for me, even though I had to threaten and reward myself into it at times.
Jonathan Haidt views the human mind as like an elephant with a rider. The elephant is the mind’s unconscious. It is large, powerful, and cares about immediate concerns. The conscious mind is a rider sitting on the elephant’s back that can see farther away and make longer term plans. The rider can’t push the elephant around. The best the rider can do is give small directions, and if the relationship is good between the two then the elephant might listen.
The behaviorist interventions attempted by rationalists are ways that the rider can influence the elephant. When the rider is right about their desire being good for the elephant too, this leads to a really good outcome for both. I wonder if Kohn would say that the rider is trying to control the elephant here, but since they’re both the same person I also wonder how useful that framing is.
I’ve presented behaviorism as favoring deontology (good is rule following) and Punished by Rewards as favoring virtue ethics (good is embodying principles). The rationalists generally favor utilitarianism: good is whatever gets valuable outcomes.
Utilitarianism doesn’t fit very nicely into Schneier’s 4 layer dip of trust. If someone is sitting outside of that context and just choosing what to do, they may end up choosing techniques from all the different layers.
School
Kohn says that kids’ interest in learning is the primary criterion against which schools should be judged. I, personally, think we should judge schools by how well our students learn. I do not think that interest is enough, I think we actually need good pedagogy and teachers that teach. Learning can be hard. It can be a slog. I know this because I have tried to learn some very hard things as an adult. I feel motivated to do this, but I also sometimes complain about it. I don’t think that’s bad.
In fact, Kohn claims repeatedly in his book that kids are naturally interested in learning, curious, and hardworking. If he believes this, and he believes that the primary measure of a school is whether kids are interested in learning, then I’m surprised he doesn’t go all the way and propose doing away with schools. Some folks do, and there’s a whole collection of unschooling aficionados who let children do whatever they want with the idea that they’ll eventually want to learn. (Kohn is skeptical of a similar trend known as free-range kids.)
I’m sympathetic to unschooling after seeing my own kids hoover up non-fiction books, science comics, number blocks, chemistry kits, and electronics gadgets. I also think that some parts of learning are effortful and annoying, and that people sometimes do need to buckle down in order to get really good at stuff. I suspect that some of the benefits of unschooling are drafting off of the infrastructure of the normal educational system, and that if everyone started unschooling then things might break down.
Kohn talks a lot about how to help individual students without rewards and punishments. He doesn’t talk much about the group interactions of a classroom, beyond noting that teachers sometimes use it as an “excuse” to reward and punish.
School should be an environment where most people can learn, and that does mean that individuals shouldn’t be disrupting class. Any discussion of how to organize class needs to take this into account. Just talking about reforming existing power structures is too easy. It lets the reformer off the hook for doing the hard part of saying how they would accomplish the good parts of the existing structure.
Kohn asks if “we want only to control short-term behaviors, or do we want to help children become responsible decision-makers?” He comes down squarely on the latter, but it seems obvious to me that both of them are actually very important in school settings. If every student had a dedicated 1:1 teacher, then maybe we could put all effort on long term responsibility development. That’s not our world, so we actually do need solutions that work in the short-term for large groups.
If a student is disrupting class and preventing others from learning, we need a way to deal with that. Rewards and punishments are an institutional pressure that helps here. There are of course moral ways of doing this too. Kohn lists some of them, such as involving kids in coming up with classroom expectations. I support those completely, and I think that rewards and punishments can serve as a backstop.
Kohn is against that tired old slogan “I want to see what you can do, not what your neighbor can do.” But that’s stupid to be against! It actually matters if my kids can do math or write. I don’t care about the group performance of their second grade class, I care if they are learning!
I don’t want to go too far here. Group work is great. It’s a good way for kids to share knowledge, to get social payoff for knowing things, to help each other, to grow. I think there’s a huge place in school for group work. I also think it’s possible for groups to end up lopsided, with one or two kids doing most of the work. Solo tests (“I want to see what you can do”) are there to act as a signal about who’s learning what.
Kohn might push back on me here and say that grades are punitive, but I don’t care about the grades. I care about whether my kids are learning, or if the classroom environment is set up such that my kid doesn’t get the opportunity because another kid is doing the whole project (or because the curriculum isn’t working, but that’s a bigger problem). There needs to be some signal there that is indicative of each individual child’s learning.
Kohn does note that grades are often a parent’s only view into what is happening at school. I care a lot about what my kids are learning, and I don’t quite trust schools to do a good job on their own. Grades aren’t a great solution to this, because if I don’t trust the school to do a good job teaching my kid then why would I trust them to do a good job evaluating? Grades aren’t great, but they are better than nothing and they can give warning signs. Kohn says that parents will need alternate sources of information on their kids’ progress when grades are done away with, and I’m glad he emphasizes it.
I’m a big fan of mastery learning, and one of the main features of that style of education is that kids don’t move on until they really actually know the material they’re learning at the moment. That means the teacher needs to know what the students can do. Kohn gives a big long list of reasons for grades, but not one of his reasons is so that the teacher can know how the student is doing. He buries that particular use like six paragraphs down and hedges it by saying that a competent teacher would just “have a good sense” for how the kids are doing.
How does the teacher get that sense?! If my kid’s teacher isn’t sure how he’s doing, should I just assume she isn’t competent? Like everyone else, I would like to reduce class sizes down one or two students per teacher so that the teacher can focus on each individual student. That doesn’t seem likely to happen, and I still want my kids to get the most out of school. To me that means that “I want to see what you can do” is critically important.
Other than that, I basically agree completely with Kohn about grades. They suck. Kohn goes on to say he’s against sorting kids, and I’m less sure about that. One reason I like mastery learning is that kids can go as fast or as slow as necessary for them to really learn the material. I think that will naturally lead kids to learn things at different times, and there’s no need for teachers to push that. I think it’s also important that teachers not push against it, and it’s not clear Kohn would agree with me on that.
Kohn has a set of specific recommendations for improving schools. These boil down to:
- Let kids learn via discovery by being open to mistakes, eliciting curiosity, and giving reasons for assignments
- Don’t make learning competitive. One student learning something faster/better shouldn’t take points away from someone else. Allowing kids to complete project-based assignments in different ways can reduce competition.
- Do a lot of group work. Let kids share, discuss, and rethink.
- Contextualize new information by tying it to topics kids are already interested in.
- Make learning hard, but not too hard. Kids love a challenge.
This all makes sense, and I mostly agree with it. Here’s what I think is missing:
- Deliberate practice on specific skills.
- Memorizing critical information.
Kohn doesn’t like these (he’s explicitly against “rote” learning), so I’ll explain more about why I think they are so important.
First of all, memorization is just knowing things. If you memorize it, you know it. If you don’t know it, you haven’t memorized it. Memorization gets a bad rap because it’s effortful and sometimes boring, but memorizing the right facts allows people to use those facts to understand new information and build new knowledge.
I also think memorization gets more hate these days than it deserves. We know how to help people memorize something: spaced repetition. Using flash cards with exponential forgetting curves lets people minimize study time and maximize memorization. I personally do anki flashcards every day for about 5 minutes (often while drinking my coffee), and it has helped me to memorize a huge number of things that I find valuable. I have thousands of flashcards, but I only do a few each day due to the way spaced repetition only presents cards when they’re about to be forgotten.
We should be teaching kids how to memorize, and then we should be helping them to memorize useful knowledge.
This brings me to deliberate practice. Memorization is effortful. It’s not just chatting with friends about a topic, or reading a book. You have to actually try, test your memory, and try again. There’s a lot of failure, because you’ll forget something and then need to re-learn it. This effortful deliberate practice is very useful, and we should help kids to learn the perseverance to do it well.
We should do the same with other skills, like reading, writing, and math. These are things that just take a lot of practice to get good at. Doing that practice, whether via homework or schoolwork, is important. When people have these skills as adults, it genuinely makes their lives easier. It’s worth supporting kids in learning it for that reason, even if love of learning never makes it to the moral pressure level for them.
Today’s school problems
The years between 1993 and 2026 have seen a lot of changes to schools, many of them for the worse. I often see complaints about schools closing gifted programs, no longer teaching advanced math, or just requesting parents to not teach their kids. I also see complaints about grade inflation and poor pedagogy. People have started collecting lists of these failures. Some of what I hear complaints about seem like things Kohn would also dislike, but he might celebrate other changes.
I benefited a lot from accelerated learning programs when I was in school. I was doing high school algebra and geometry in middle school. I did Running Start in high school, where I spent junior and senior years at a community college taking classes for college credit. These things were good for me. I liked learning, but I absolutely hated high school. High school (with a few exceptions for specific teachers or classes) made me miserable. Escaping to community college was a huge benefit to me.
This experience makes me really worry about degradations in advanced programs. They were so helpful to me and to my friends, and I worry that districts who dismantle them are harming their students. I’m not really sure what Kohn would say about it though. He doesn’t really talk about having high standards for learning or for skill growth in the book, so he might not care about a lack of performance. He does care a lot about control, and reducing options available to kids is more controlling, so maybe he would support accelerated learning programs on those grounds.
It’s also hard to know what Kohn would think of grade inflation. He hates grades, and thinks that they should be abolished. Instead, he wants direct feedback that has been leached of any hint of reward, punishment, or praise. I think he would hate that we still have grades at all. That said, one of the things he suggests is replacing the A through F grades with only two: A and incomplete. I actually really like this idea in theory, since it lines up with mastery learning. I do ultimately want all kids to get really good at what we’re trying to teach them. The A/incomplete idea only works if teachers have high standards though. If kids are getting As before they actually know the material, then those kids are being betrayed by the educational system.
The last major change in schools that I want to touch on is pedagogy. Teachers are teaching the same subjects in different ways now, and that’s preventing kids from learning. The poster-child for this is reading: teachers stopped teaching phonics. Now that some states are moving back to phonics based teaching, it’s become obvious how bad the last couple decades have been for American literacy
I’m not an expert why 3-cuing replaced phonics for a while, but one story about it is that teachers were feeling controlled by the phonics curriculum because it prescribed specific learning procedures. When another curriculum offered more choice to teachers, they jumped on it as a better option even though it was dramatically worse for the students.
Kohn’s book doesn’t touch on high standards or performance over time, but he has a blog post that does. His post argues that people have been complaining for decades that kids these days are performing more poorly than they used. Kohn says that if people in the Eisenhower era were complaining about schools on the decline, then there was no golden era of great teaching. While I agree about there never being a golden era, I think the argument doesn’t actually show very much. Mississippi’s phonics turnaround happened in less than 10 years. It seems very feasible to me that a cycle of education degradation and rebuilding can happen in between major complaints (perhaps caused by them), so the idea that people have been complaining about education for decades is meaningless on its own.
Kohn’s book emphasizes freedom and choice. Enhancing choice is one of the main reasons that he rejects rewards. He even suggests that if we control teachers (such as by enforcing a specific teaching method) then the teachers will therefore try to control (and thus mistreat) students. He says explicitly that teachers should be given freedom to decide their own curricula. The literacy losses since his book was published show the limitations of his idea.
Fundamentally, teachers have a job. That job is to help kids learn and grow as much as possible. Requiring them to use the best teaching methods available may be institutional pressure instead of moral pressure, but doing so is worth it for the benefit to the students. Of course, we also need to worry about whether we are requiring a good curriculum or not. What did the people of NYC think when they required reading to be taught by 3-cuing, and thereby destroyed the reading ability of their children?
Work
The main goal of teaching is to help each student do their individual best. The main goal for employees is to help the team produce as well as they can. These are different goals and therefore we need to treat them differently. I disagreed with a lot of Kohn’s suggestions for students because he suggests only focusing on group achievement. My comments on his advice for workers are different.
A lot of Kohn’s complaints about rewards in the workplace can be explained through Goodhart’s law, so I won’t belabor it much here. Suffice it to say that rewards in the workplace can cause people to focus on the wrong things.
One more interesting point about rewards in the workplace is that they don’t seem to improve performance. This was true for kids as well, but my use of rewards for kids is mostly limited to getting them to do things they otherwise wouldn’t. Performance improvements didn’t enter into it.
It seems that in the 80s, pay-for-performance incentive plans were common. Kohn talks about situations where people get cash rewards for doing a better job. He cites a lot of studies showing that paying somebody to improve their performance doesn’t work very well.
There’s a saying in martial arts that motivated me to train harder: “You don’t rise to the occasion, you sink to the level of your training.” The idea here is that if you aren’t able to do something in the training hall, you won’t be able to do it in real life. High pressure situations make it harder to think and people will default to whatever they already know.
In the workplace, offering rewards for performance increases pressure. That means the peak performance a worker can achieve will be lower. That might be fine if the reward moves someone from not doing something to doing it, but it won’t benefit you if they’re already doing it and you just want them to improve how well they do.
Kohn’s advice on performance matches my own intuitions. If we want to improve performance, we need to focus more on training and less on direct reward for performance. My company does this by offering direct reimbursement to employees for training that they do outside of work (I buy a lot of books and online classes this way).
Kohn quotes several studies as saying that pay is not an important motivator. It generally ranks behind things like “interesting work” and “impact.” I don’t question these results, but I do question Kohn’s interpretation of them. Most people are not optimizing their lives around income, they are satisficing on income by finding a job that pays “enough” and sticking with it. I think that’s good, but it doesn’t mean income won’t change people’s decisions.
Several large-scale national surveys have found that people who were unhappy with their jobs pointed to reasons like the lack of variety or challenge, conflicts with coworkers or the boss, and too much pressure. Salary simply was not a major issue.
I think this means the opposite of what Kohn is implying here. Salary isn’t a factor for the people unhappy with their jobs because when salary is the reason for unhappiness then people change jobs. Being unhappy with your salary will get you to look around for a new job. Being unhappy with the level of challenge will get you to kvetch at the neighborhood barbecue. I know this because this person is me: I left my last company for two reasons, and one of them was money. My current job offers too little challenge and too little impact, and as much as I talk about leaving for something better, I am still there.
One of the studies in that survey […] found that people who were paid on a piece-rate basis did not even turn out any more work than did those paid only for their participation.
Whyte reported that piecework systems typically fail because workers “set a quota on what constitutes a fair day’s work and refuse to go beyond this amount even when it is well within their ability to do so.” This is not, he continued, a function of a poorly designed or administered incentive system but of the underlying theory of motivation.
This seems fine though. Remember that we don’t care about maximizing individual performance in the workplace, we care about maximizing performance of the team. If each individual decides how many pieces to make in a day, and we pay them fairly for those specific pieces, that’s fine. Increasing the amount we pay for a piece may not cause people to make more pieces. It’s still worth it if it causes more people to enter the market to produce those pieces, increasing the size of the “team.”
Kohn argues that work is not a means to an end, it’s something people do because it fulfills them and helps give life meaning. I totally agree with this for the workers. Money and payment aren’t the reason that we work, and plans to improve performance only through changing salaries are not going to be as effective as other plans. People want to work, money just changes what they work on.
Capitalism
At first I thought Kohn’s book was just arguing against bonuses for performance improvements, but Punished by Rewards leans very hard towards an equal pay for all workers model. As Kohn puts it in the book:
After I offered a critique of incentive plans at a management conference not long ago, one businessman exclaimed, “Well, isn’t that communism?”
This is a bit of a throw away line in the book. The next paragraph immediately pivots to the history of management. Kohn eventually returns to this line near the end of the book, but again just gestures at “calling into question our most fundamental institutions” without actually making an argument. I want to linger on this comment though. Kohn doesn’t come out and say he’s a communist, but he also doesn’t say he’s not (in other writings he says he’s a democratic socialist). My feeling reading this particular anecdote is that Kohn is looking at us, the audience, as though expecting us to implicitly understand that communism is better.
I don’t particularly think that communism is good, and this line helped crystalize my complaints about the “work” section of the book. I can accept that pay for performance doesn’t work very well, and that pay for creativity is doomed. I don’t think that means that we should pay the same for everything.
In particular, I think Kohn is making a common mistake for academics, of thinking that work is for the benefit of the worker. It makes sense that academics would have this bias, because school is actually for the benefit of students. The main purpose of work is to provide stuff to people (to society, to your community) that they need.
It’s easy to lose this in the modern era, because we all live in the middle of such luxury. Very few people in modern America need to worry about whether they’ll be able to afford to heat their house in the winter. Very few people need to worry about whether they’ll literally starve. Some still do, and that’s a tragedy, but wholesale famine is a thing of the past. It’s a thing of the much more recent past for communism.
Famines in the 20th century were primarily in communist countries. The Holodomor, the Great Chinese Famine, and the North Korean Famine – these were events where a country’s organizational system was unable to provide enough for people. Hundred of millions couldn’t get enough food. Tens of millions starved literally to death (though “only” 500,000 for the North Korean Famine in the 90s).
That’s why I get nervous when people start implying that communism is so obvious that they don’t even have to argue for it. Look through famines in the 20th century and they are mostly communist countries or war torn countries. The productive output of an economy literally defines if the people in that economy get to live. Famine isn’t the only reason I don’t want to live under communism, but it’s a big one. I think capitalism has a lot of problems, and I also think it’s resulted in a great decrease in human suffering over the past few centuries.
I’m not trying to say Kohn wants famines. I’m not even trying to say Kohn supports communism (he only gestures at it, remember). What I’m trying to say is that the cost of ignoring effective institutional methods of organizing people is enormous. Those enormous costs are not visible in the individual motivation of people. They’re visible in the large scale output of groups.
If I’ll agree with Kohn that monetary rewards for performance don’t improve an individual’s performance, why then do I think that capitalism works? When capitalism works, it’s because it encourages people to move around and do new things, not because it encourages people to work harder.
Consider an economy in a steady state. Workers are all paid enough to be happy where they’re at. They produce enough that needs are mostly satisfied. Now imagine some external shock: a flood that destroys several farms, for example. Suddenly the same number of farmers in the economy is producing less. Raising the wages of the farmers won’t help them to produce more, but it will attract other people and get them to start farming. Those who were making an amount in between the old and new farmer’s wages are likely to switch, leading to more food being made.
The money that workers are paid is a signal about how many more workers are needed to do that job. When more workers are needed, the wage will rise. When fewer workers are needed, the wage will fall (blah blah inflation adjusted). It’s not really about how much value the worker produces (though if they produce less than they’re paid then they’re the customer and not the worker).
I know wages work as a price signal because I used that signal to change careers. I studied electrical engineering in school, and I worked for about ten years doing circuit design, schematic capture, PCB layout, etc. It was fun and I loved it, but at some point I realized that I needed more money. Getting married and starting a family made life suddenly feel a lot more real and precious.
Over the course of about four years, I got a sequence of jobs that moved me from electrical engineering to firmware development to software development. I took a bunch of classes in the evenings to bump my skills and did a bunch of hobby projects to test them. I now make significantly more than what I was making before this career shift.
I agree with Kohn that if someone had offered me performance bonuses to do software work while I was working as an EE, I wouldn’t have been able to do it. It literally took years for me to be able to do the work that I knew I needed to in order to get paid how I wanted. Kohn is right that short term performance won’t improve if you offer people more money to do a better job at what they’re already doing.
Where Kohn is wrong is thinking that pay for performance is all there is to our work-pricing system. We don’t pay for performance, we pay to incentivize people to move around.
Motivating Employees
Let’s move on from my extended capitalism digression. Regardless of whether you agree with me about capitalism being a good way to figure out where to allocate workers, I think we can all agree that the inside of a firm is not capitalist. Nobody pays their coworkers to help them with a problem. Nobody uses supply and demand to figure out what parts of their project to prioritize on a day-to-day basis.
If capitalism lets us incentivize people to shift companies, why not pay everyone at a company (or maybe in a given role at a company) the same amount? Then if the company needs more of those people they pay everyone more. This brings us back to Kohn’s main question: how do we get the best out of the people that we hire? How do we manage them?
Kohn complains that pay-for-performance schemes are just a way to avoid doing the hard work of management. Again, I don’t know what life was like in the 80s, but if all managers were doing was trying to figure out how much to pay people that seems terrible. Maybe this was a huge revelation when the book was published. I’m not sure.
What I am sure of is that all of the big companies I have worked for are basically doing the things that Kohn wants managers to do. I am lucky enough to work as a software developer, a job for which there’s a cultural understanding that people need the freedom to excel. Kohn wants managers to “set up conditions that will maximize the probability of [workers] developing an interest in what they are doing and remove the conditions that function as constraints.” That’s what my managers at big companies have focused on doing.
In order to set up workers for success, Kohn wants companies to look for problems and blockers and help people solve them. He wants managers to listen to worker concerns and try to see things from their point of view. He wants managers to provide informational feedback. He wants them to facilitate more teamwork, knowledge sharing, and camaraderie. He wants them to help assign work to workers that is interesting and novel and helps the worker learn new skills. I don’t think managers are always good at this, but it does seem like exactly the thing that they’re trying to do in my experience.
When it comes to performance reviews, Kohn wants to decouple them as much as possible from compensation. He complains that performance reviews only serve to show how a manager thinks a worker is doing, but at the big companies I’ve worked all perf reviews have been 360. I’ve gotten way more official feedback from peers than from managers.
Kohn also thinks promotion cases assume “someone’s achievement in his current job is a reliable predictor of how successful he will be in another, very different, position.” That’s not how promo cases work in big tech. I’ve been told over and over that you get promoted into a role when you’re already doing the job. As a senior software developer, I had to already be handling most of the responsibilities of a staff software developer before I would be assured of promotion. My manager and my more experienced peers worked really hard to give me the opportunities to show that I was doing that, and now that I’m more senior I strategize every week about how to give junior devs the opportunity to grow their responsibilities.
When it comes to accountability for poor performers, the traditional way of handling it is the Performance Improvement Plan (PIP). The stereotype is that the PIP is just an excuse and once someone has been PIPed they’re going to be fired. That hasn’t been my experience seeing it from the outside. I’ve seen people quit rather than going on a PIP. I’ve seen people quit halfway through rather than keep going with it. I’ve seen people succeed at PIPs and go back to normal performance. I think I’ve only seen it result in somebody being fired once.
The way a PIP works is that there is a concrete set of things that someone has to do to reach “adequate” performance. The list is very explicit and objective, with as little room as possible for managers to fudge the outcome one way or another. At the end of the PIP, the worker can point to the list and say “I did this, I did that, I did the other thing.” Sometimes there’s a quality component, but even there I’ve seen it written to be as objective as it can be. The PIP works as an institutional pressure, providing explicit rules (do this) and consequences (or else that).
Yes, I think the end result of a PIP is a consequence and not a punishment. When I’ve seen people on PIPs, their managers have often been pretty stressed out by it. The manager doesn’t want their employee to suffer, they don’t want the employee to be fired. My own manager now has gone to great lengths to prevent people from being put on PIPs (like making skill-up plans for employees who are struggling so they know how to avoid the PIP). Teams exist to accomplish certain work-related goals, and if people can’t support those goals even with major institutional backing then the consequence is that they aren’t allowed to contribute to the team anymore.
And what about money? Kohn describes his own method of negotiating payment for talks as: negotiate for maximum fair compensation, then decouple performance from money. That’s not quite how a job at a big tech co is, but it’s close. There’s not much talk of salary or compensation when people are actually working. People talk a lot about how to negotiate a salary and stock going into the job. Every year around perf review time and vesting time people will talk about raises and ratings. People do get bigger raises when they have a better perf rating (as rated by both peers and management).
That brings me to the main way that big tech companies diverge from Kohn’s advice. Every one I’ve worked at uses a rating scale for all employees every year. Often this is a 5 point scale, where 3 is “does all the responsibilities of their job” and is how most people are rated. Those people who are doing a lot of responsibilities at a higher level than their job might be rated 5, and those who are struggling to accomplish anything might be rated 1. These grades are not informative – they often come long after the informed feedback of managers and peers has been delivered.
Workers (myself included) definitely choose the work they do with an eye towards what rating they want. Managers talk to employees about what new scope they’ll have to take on to reach a 4 or a 5. Kohn would hate this, and would say that it shouldn’t work well to motivate employees. To me it seems like the company is trying to figure out how much they want to pay the person to stay. If capitalism is about constantly revaluing tasks so people move around, then companies need to figure out whether they want to compete to keep employees when compared to wages offered elsewhere. I’m not certain about the motivation aspect to it.
Talent
There’s another option that people pursue to build effective companies. Tyler Cowen calls it finding talent. In this paradigm, you aren’t looking for ways to motivate or manage employees. You are looking for employees that have enough talent and drive that they don’t need you. Find the right people, give them what they need, and let them cook.
This might work to start new companies. It might be what you want to do if you only care about hits. Fund 100 people who might be talents. If 95 of them crash and burn but 5 of them come out with medical miracles, that’s a win for the whole world.
This is not a very useful paradigm for deciding what to do in a school. It’s not very useful if you already work in a company with employees and need to interact with the existing systems. It’s not even clear how much this paradigm can be scaled. At some point your biotech startup needs a lot of techs to run your experiments, right? Or do we just wait until those get replaced by robots? Is the plan to eventually transform the entire working world into “talents” and robots, and everyone else just goes home to play with their kids?
I mostly like the idea of talent as a way to differentiate more between work and school. School is where you develop talent, work is where you deploy it. We should be helping as many kids as possible develop their talents as far as they can. We should help as many workers as possible deploy their talents as much as we can. Inasmuch as Tyler can identify talent and allow it grow, that’s great.
Talent acquisition doesn’t dissolve the need for management or training, and I think this mindset doesn’t speak much to the ideas of punishment and reward in general.
Choice
There’s a just-so story about checklists. Pilots adopted checklists early because if they made a mistake, they would die. Surgeons rejected checklists for a long time because if they made a mistake, they wouldn’t be the one who dies.
I don’t know how true that is as a historical description, but it gets at my feeling about choice when it comes to work. There are better and worse ways to accomplish tasks, and if someone chooses to do something a worse way then that person is responsible for their failures. This is true even if the effective method is “less fun” or offers “less creativity” than whatever alternative the person selects.
I touched on this a bit earlier when I talked about teachers who stopped teaching reading using phonics. Kohn wants to maximize autonomy and choice, and I fear what that will do to the effectiveness of our society.
This is not to say that I think choice and autonomy aren’t valuable. I also want to increase the autonomy that people feel. I especially want to increase autonomy for kids! Kids in our society are often prevented from doing a huge number of things, which I think is very bad for them overall. I just also think we need to be sure that giving people choice doesn’t let individuals break things for everyone.
In the software engineering world, people love creativity and autonomy. One of the common motivators for people to get into programming as kids is that it lets them have complete control over something, often for the first time. This complete control can be dangerous for workers in a big company, as a mistake on prod could bring down a whole site and cause thousands to millions of dollars of lost revenue.
How do companies like Amazon deal with software developers’ love of autonomy and the high risks of breakage? They use Schneier’s 4th layer of trust enforcement: security. Between git merge limitations, PRCI automated testing, and prod deployment checks, it is actually very difficult to do something that breaks a web site at most large companies. The systems in place prevent breaking changes from landing, providing a safe sandbox for people to mess around and exert choice. It’s only when people’s choices would impact the overall product that they start running into guardrails.
Every company I’ve worked at has had rules and institutional pressures to support code quality, but those are generally less important to continued high quality than the security guardrails that prevent failures from leaking through.
Teachers, police, and surgeons don’t have those guardrails. The nature of their job means that they are always working “on prod,” so they always need to be performing well. This is where tools like checklists come into play (I wish Kohn would review The Checklist Manifesto so I could know his thoughts on it). Teaching and similar jobs are where “controlling” features like institutional reward and social praise can help. It’s where people’s pride in a job well done can help. We need to recruit as many of Schneier’s pressures as we can to help people in these jobs to choose to use the best techniques available to them.
With all of that said, I do agree with a more general exhortation to maximize autonomy and choice. I do want people with direct experience at their job to participate in decision making about how to do it. I do want people to feel in control of their own work and for their decisions to make a difference. I want to maximize choice and autonomy as much as possible without sacrificing effectiveness.
Status
Kohn talks about how money isn’t necessarily the biggest motivator at jobs. Some bigger motivators are things like impact and group culture. In other words: social pressure.
One reason that impacts on others is so important when motivating people is status. It’s weird to read a whole book about how to motivate people and not see a discussion on status. Money and physical rewards are obvious, but I feel like social status and the positive regard of your peers is a huge part of why people do anything. Is this just obvious to me now because everyone wants to be an influencer?
Kohn talks a little bit about status when he says that rewards create an atmosphere where we want to impress the person who hands them out. That’s not automatically bad to me. Wanting to impress others may not be as noble as wanting to discover the laws of the universe, but it’s also a very human desire. Turning that human desire to impress people towards learning seems good actually.
For a book that explicitly values group work and collaboration so much, it’s weird that it doesn’t interrogate the social aspects of motivation very much. Schneier’s book Liars and Outliers is the thing to read for more info on status as a trust mechanism for supporting society.
We Live in a Society
In the section on school, Kohn doesn’t talk about the impacts one kid’s behavior has on a whole class. In the section on work, he ignores the systemic impacts of wages as a pricing signal. I think the book has an overall blind spot in thinking about how to motivate or incentivize people when their behavior has immediate impacts on others. Kohn only thinks about the impacts of rewards and punishments on the recipient, not on the people the recipient impacts.
When it comes to discussion of how to raise kids, it makes sense to focus mostly on them. Most books about raising kids or being a parent focus on kids: what they need and how to raise them well. I think maybe that emphasis has grown too strong in recent years though.
I don’t know what life was like for parents back in the 80s, and I barely remember what it was like for me as a kid in the early 90s. Kohn reports that spanking was used by the “overwhelming majority” of Americans, but I’ve met exactly one person who admitted to using it these days. Maybe it’s good that Kohn’s only emphasis in 1993 was on how kids experience actions designed to shape their behavior. I do want to drive spanking down to 0. I also think that a focus on the child can go too far if it results in an abnegation of the parent.
Personally, I really benefited from reading Hunt, Gather, Parent. I forget if the argument was explicit, but I came away from that book with the understanding that my preferences are actually important. Hunt, Gather, Parent blew my mind with the argument that if a toy really bothered me, it was OK for me to get rid of it. I could throw out the toy because of my own preferences, not out of a cost/benefit calculation for what was good for my kid. While I can’t speak for other parents, it was very useful for me to be reminded that my kids are in a relationship with me and I’m not a robot whose sole purpose is to train my kids.
Kohn has a throwaway line in the book about how we shouldn’t potty train kids until they’re ready for it. To me this is a perfect example of Kohn neglecting the needs of parents. Potty training is such a cultural thing, with age varying hugely around the world. Waiting for kids to be ready for it is not nearly as relevant as doing it in a way that’s appropriate for the kid’s age. One of my Malaysian friends was doing some preliminary potty training with his 6mo old, but that looked very different than the kind of potty training my wife and I did with our kids at 2 years.
I’m not against child-led milestones. What I am against is the idea that anything other than child-led is a problem. The needs of the rest of the family do impact when kids should be learning new skills or taking on new roles. It could be harder for the kid, sure, but we’re not here just to make things easy on them. We’re here to raise them well. We have responsibilities to them, and they have responsibilities to their own family. Yes, they have responsibilities even at two years old. I fervently believe that the responsibilities a kid undertakes should rise smoothly as they age, and that means they need to start very small when they’re very young.
All else equal, I want to treat my kids in ways that raise them up to be virtuous, hard-working, thoughtful people. Sometimes all else is not equal, and I need to take actions that piss them off in order to keep them or others safe. Other times I need to take actions that seem like punishments to them, but are really just me protecting my or my wife’s sanity after a long day.
Ultimately, I want to raise kids that I want to be friends with as adults. One of my greatest hopes in life is that when I’m 60, my kids and I share a deep bond of affection and respect that goes beyond just being parent/child. I want us to hang out, to share struggles and to share triumphs. I want to go grab a drink at the pub with them on the weekend and hear about what they’re working on.
My kids are only 7 now, and we already have fun on hikes and doing little hobby kits on the weekends and playing D&D. Those are things I might do with my adult friends. The tenor of the interaction is different, but I can see a path towards the future I want. This is part of why I react so negatively to Kohn. He advises parents to go through an enormous amount of introspection about how their behavior impacts the child before they even compliment the kid.
I know people who are my parents’ age, with kids my age, who don’t have any friendship with their kids. Their kids don’t go to them for advice or comfort or help. Their kids don’t go to them with triumphs. It’s a holidays and vacations type of relationship only. My observations of the boomers is that it’s rare for them to have an actual good friendship with their kids. Some of them seem confused about why it didn’t happen, others seem like they don’t mind or don’t notice. That’s not what I want for my family.
I do acknowledge that there’s a lot of subtlety here. While my kids are kids, I am their parent first and foremost. There will be times (a lot of times) where I just can’t be their friend. What I try to do is make it really clear when I’m acting in one mode or the other. I obviously also have to make sure that I’m the kind of person they’d want to be friends with when they get to be adults. This is a place where I’m still learning and growing, and it’s a place where Kohn’s advice seems likely to be counterproductive. All he says is to “develop a caring alliance of openness and trust,” but he doesn’t say how to do that.
Virtue
In spite (or perhaps because) of hanging out with so many rationalists, I’m not a utilitarian. Perhaps utilitarianism is fundamentally correct in some sense. Even if it is, using utilitarianism to decide what to do is like using quantum physics to design a bridge. It’s so inefficient as to be impossible.
In a very deep way, I’m a virtue ethicist. I might be more virtue-ethics aligned than anyone else I’ve actually met. I think it’s crucial to develop a strong internal character of responsibility, self-determination, charity, courage, and creativity. I think a lack of those virtues is having a noticeable negative impact on our society today.
Kohn seems to share my view on the importance of character (though maybe not in the details). This is why I kept reading the book, even though much of Kohn’s pedagogy and even some of his morality are objectionable to me. I am looking for better ways to instill virtue in my kids.
We recently enrolled both of our kids in a martial arts class. Martial arts classes for kids are not really a self-defense thing; regardless of your threat model, a kid should be running or deescalating from every encounter. Instead, kids’ martial arts are about exercise and about self-discipline.
I did Tae Kwon Do from when I was about 10 until I was 16, when I got my first degree black belt. The physical conditioning was hugely beneficial, but not as beneficial as the discipline and the virtues that it very explicitly focused on. My dojang had a list of virtues painted on the wall, and we had to know them for our belt tests. I still sometimes chant “ha-myeon doen-da” (goals we set are goals we get) to myself when I start something hard. I’m hoping my kids pick up some similar virtues from their own martial arts experience.
It’s strange to watch my kids in their martials arts class while reading Punished by Rewards, because martial arts is the one place every week that has the highest amount of what Kohn would call punishment. Forget your belt? Do 20 pushups before the class starts. Fool around in class too much? Do pushups instead of getting to kick the targets. Make a hand juggling motion? Do 67 pushups. None of this seems bad intentioned to me, and the class does have a lot more discipline from the kids than I’m used to from kids that age. Punishment’s presence is stark when compared to the rest of my kids’ non-punishment oriented life, even though punishment is generally rare even in martial arts class.
The main focus of the class is focus. The kids practice doing hard things with intention. They practice doing a lot of kicks. They practice holding their breath. They practice balance. They hold stances until their muscles strain. At the end of it, the instructor emphasizes that they did that. They did a hard thing, that seemed hard, but they focused and exerted themselves and succeeded. I think this is great.
I don’t know if it’s the exercise (schools don’t have nearly enough of this) or the focus on focus, but starting martial arts has also had a great impact on my kid doing homework. Rewarding him with a chocolate chip for each word written did help a lot over the past year, and after he started martial arts it seems like the homework aversion might be fully resolved.
I wonder what Kohn would think of this. He emphasizes over and over that we need to raise kids to have internal strength of character: agency, self-determination, some kind of internal fire of motivation. Now my kids seem to be getting a lot more of that, and they’re getting it from an organization that’s a lot more open to minor punishments than the rest of their life is.
Instilling Virtue
Kohn thinks the best way to instill virtue into people is using the 3 Cs of Content, Collaboration, and Choice.
- Content is what you’re asking folks to do. What kind of virtue are you asking them to embody, and how reasonable is it for them specifically.
- Collaboration is how you involve folks in the decision making process for what’s allowed and how it will be done. Involve them as much as you can given their age and skill levels. When there are problems, help them to understand why and to reason through what should be done on their own. This is the true “teachable moment”.
- Choice is how you let people do what’s right on their own. In order for people to learn moral virtue, they need to do it independently of social or institutional pressure. They need to experiment with it, fail at it, and try again.
Kohn gives some examples about how to instill the virtue of care within kids using his 3 Cs framework. This was super valuable, and for me this chapter was definitely worth reading the rest of the book for. It was still limited though. Alfie Kohn focuses on only the one virtue of caring. He mentions courage, but only in the context of being able to care better. I find this a bit flat.
In Ben Franklin’s autobiography, he describes 13 virtues that he wants to embody. These include justice, industry, and frugality. Franklin doesn’t include care as a virtue explicitly, though many of his virtues speak of doing good to others. He doesn’t talk about courage at all. Nonetheless I find his list to be more humane than just a focus on care itself. It’s more accepting of the complexity of life and value.
I also really like Franklin’s virtue of frugality. He phrases it as “waste nothing” but to me it represents an idea of effectiveness. We only have so much time. We only have so many resources. We only have so many chances to help our kids and our workers and ourselves. Let’s make the most of them. Kohn wants to judge schools by whether kids like learning, but the virtue of frugality leads me to judge them by whether kids actually learn.
Let’s go back to Schneier’s 4 ways to get people to do the right thing: moral pressure, social pressure, institutional pressure, and security. Kohn argues that any action made to improve social, institutional, or security will not impact the moral realm. He says we can’t teach morality by changing social pressure. I’m mostly aligned with this (though unlike Kohn I don’t think that means the other layers are useless or bad).
Kohn also argues that institutional or social pressure can undermine morality. I remain pretty skeptical of this. I think it’s more likely that people perform social or institutional pressures and then think that acts as moral education. If a parent is using social and institutional pressure (rewards) and that displaces moral education, that’s the problem that can be solved by just doing the moral education that was displaced. There’s not an intrinsic tradeoff here except in time, and I generally want to spend more time with my kids.
Mysterious Wizards and the Demands of Modern Life
Robert Kegan’s book “In Over Our Heads” is about how people deal with the world by changing their own concept of themselves. It’s complex and difficult to read, but does a pretty good job of explaining how people’s mode of thinking and conceptualizing their own self impacts their ability to succeed in our modern world.
Kegan arranges modes of thinking into a hierarchy that people move through as they age. These modes describe how people primarily represent their own self-concept. Each level defines the main way that the person is driven: how they figure out what’s valuable and what to do. I argue that these levels are what it feels like from the inside to be most constrained by Schneier’s different pressures.
- Level 1: impulse or reflex (infants and toddler years). When you’re primarily being limited in action by security precautions (fences, git integrations, parents who don’t let you put marbles in your nose) this can feel confusing and arbitrary.
- Level 2: needs and desires (kids, teenagers). When you’re primarily choosing actions based on the institutional rules and what the explicit rewards and punishments are, this can feel isolating and zero-sum.
- Level 3: social relationships and roles (most adults). When social pressure guides your actions, problems can feel like they reflect on your fundamental worth as a person. Someone critiquing your performance can feel equivalent to an attack on your relationship with them.
- Level 4: a self-authored set of values (some adults). If you are choosing actions out of an understanding of your own moral values and agency, it will feel like you are able to chart a course through your own life while managing the systems that you come into contact with.
- Level 5: self-actualized and unbound (mysterious wizards). When you are a mysterious wizard, life feels like an infinite game where you are expressing that which is good through your every action. Or something.
I’ve heard it said that someone at level 2 shows up to work on time because they don’t want to get fired. Someone at level 3 shows up to work on time because they know their coworkers are depending on them. Someone at level 4 shows up to work on time because that’s their job.
Somewhat confusingly for us, Kegan names level 4 the “institutional” mind because that’s the level where people are able to generate and manage institutions, whereas Schneier calls his level 2 institutional because it is external institutions which mostly constrain someone at that level. Kegan’s book spends almost all of its time talking about how people make the transition from Level 3 (socially driven) to Level 4 (values driven). This is largely because modern life presents a lot of challenges that are easier to manage when you are able to work at a higher level.
In Punished by Rewards, Kohn constantly pushes educators, parents, and managers to support people in acting out of their own values. I think Kegan would approve here. Kegan, like Schneier, suggests that people need support at all the levels of the hierarchy. We should support people and help them to exist in society, even if they struggle at the moral level. Kohn is right that we should be trying to minimize how we limit people via security, institutional pressure, or social pressure. Kegan and Schneier are right that these other levels are still necessary and good. We just need to be sure they don’t suffocate the people who are still primarily living within them.
I first discovered Kegan’s work because rationalists really love it. It’s a highly structured system in which being at a higher level means you’re more capable. The levels are murky enough that it can be tricky to tell what level you mostly interact with the world through. It also has a super powerful level 5 where things just get easier. If you squint at your life, maybe you can pretend that you’re a mysterious wizard. Or if not, you can at least aim at one day being a mysterious wizard.
I’m going to be honest, I read In Over Our Heads a long time ago and the Level 5 stuff was pretty confusing to me. I don’t think I got a good sense for how Kegan’s idea of institutional thinking (his level 4) differed from his idea of inter-institutional thinking (level 5).
Kegan and Schneier have these complementary hierarchies that fit together very nicely. Kohn has no hierarchies. Given how much he hates hierarchies in human interaction, it’s possible he would hate the idea of these hierarchies as well. I’m putting his ideas into these other frameworks because I think they fit well and provide a good window into the limitations of Kohn’s thinking.
What Kohn does really well though is describe something that I now think fits the difference between Kegan’s Level 4 and Level 5. Kohn describes a kid learning caring as them coming to internalize it as a value, but he suggests that there are different types of internalization.
When a person first learns a value, they “swallow the rule whole.” This kind of values driven action is what Kohn calls introjection. One who is driven by introjected values can chart a course through their own life, but the values they’re expressing aren’t fully their own.
The level beyond introjected values is integrated values. When someone has integrated their values, they have made the rule their own. They are able to make decisions about which values and rules to embrace. They can choose which virtue to embody at any given time, based on what serves the good in their world for that specific situation.
Who makes our society?
I love America. This country has it’s problems, but it is still so wonderful. I think our constitution, our Bill of Rights, and our history have been great forces for good throughout the world.
The constitution is just a piece of paper though. Each generation of Americans has to reify it to preserve what it really stands for. We can’t just accept the constitution into our hearts and execute it. Instead, we need to grapple with it, to understand its strengths and its limitations. When necessary, we need to amend it. If we don’t, our country withers from within.
We can look back on Ben Franklin, Clara Barton, Martin Luther King Jr., and all the other heroes of our country and say that they did good. They helped to create institutions that endure to this day. Those institutions shape the rules and roles that we live our lives within. But those people also expressed the virtues that made our country great. We can’t just take the rules they produced and live by them. We have to embody the virtues that they embodied. Only then can we truly re-instantiate the institutions that they created.
I worry that a lack of virtue, of moral education, has harmed our politics greatly. We have politicians today who imprison innocent people, tell bald faced lies to reporters, shut down criminal investigations they don’t like, threaten war in order to extort good trading terms with other nations. We have staffers that lie to cover up the cognitive impairments of elderly officials. Some of my friends think these are costs worth paying for certain political ends that the administration is also pursuing. I think that is farcical. A leader without virtue will never deliver on promises unless it benefits them personally, and that puts all of their followers at the mercy of their individual whims. That is not American.
Sometimes appeals to the American virtues lead people to point out all of the ways America has fallen short of virtue. We say “grabbing people off the streets for not having their papers is un-American” and they respond “just look at these specific civil rights violations throughout our history”. They are right, but the discussion is conflating what is with what should be. Societies, like people, will never fully embody their virtues. We nonetheless need to always push towards those virtues.
Whether we know it or not, we are always collectively reinventing our society. We of the modern era didn’t miss America’s founding, we are living it right now. We can either reinvent it to be a better version of what it already is, or we can let it degrade. I’ve focused on virtue here because I think it is most lacking in our politics now, but I actually think we are collectively reinventing our society through moral, social, institutional, and security pressures. We need to reinvent all of those, and improve them all, every generation.
While Kohn doesn’t write about patriotism and I assume he’d be put off by my own, I think he would agree with the fundamental insufficiency of accepting a ruleset defined by previous generations of people. No matter how good it is, we need to re-author it and act through it as a part of our own agency. This applies at the national level and down into the family level.
Kohn isn’t against rules. His last chapter talks a lot about bringing kids into the process of creating rules, and how to balance applying rules to them from the outside with the need to involve them in rule creation. This seems crucial to me, and I basically agree with everything Kohn is saying here.
In order to help write the rules, people need to be able to understand them. That means that the laws and rules that form the institutional pressure of a society need to be in principle learnable. Obviously our world is complex and we need rules that are detailed enough to deal with that complexity. I think that needs to trade off against the fact that the more complex our rules become, the more people are excluded from democratic participation.
We need to lower the mental demands of modern American life, so that people at every level are better able to benefit from it and contribute to it. We need to ensure that the institutional rules we have apply to everyone equally. We need to ensure that social pressure is reasonable and caring. We need to help people develop virtue, so that they act well in the world.
