Earlier this year we completed a lovely Beeminder book club to read behavioral scientist Katy Milkman’s new book, How To Change. The discussion all happened in the amazing Beeminder forum but as a private group of 18 of us, so we could trash talk the book guilt-free (or just to be able to talk more freely about our own changing, I suppose). Of course the cost of entry in the book club was to beemind reading the book. We made a little gallery of our Beeminder graphs if you’re curious.
In case you’re not sure yet how much you care about this, I’ll start with an overly brief summary of each chapter:
- Fresh starts, as in temporal landmarks, are powerful.
- Temptation bundling is a clever and useful lifehack.
- Commitment devices are the best of all, especially monetary commitment devices, though sadly, for Beeminder in particular, most people are not sophisticated enough to use them. (You’re rolling your eyes at me but I’m pretty sure this is a fair characterization of Milkman’s chapter on commitment devices!)
- Trigger-Action Plans!
- Nudging and defaults and the power of tracking.
- Believe in yourself or whatever (there are some actual useful things in this chapter too).
- Conformity and social pressure are influential. Everyone knows this.
- Akrasia is like diabetes — you don’t take insulin for a month and assume you’re cured. You’ve got to apply and maintain these systems and tools your whole life.
My consternation at how little Milkman talks about Beeminder in the book was abated by the fact that it’s practically a Beeminder paean. Or, another way to say that is Milkman is at the cutting edge of the science of behavior change, and so are we, and we’ve done our best to incorporate it into Beeminder. So when Milkman describes the practical lessons from all that science, she’s kind of describing Beeminder. Which makes this an easy book for us to review. But there are plenty of Beeminder-orthogonal ideas in the book which I’ll get to as well.
Let’s dive in.
Milkman sets the stage in the introduction with a fun anecdote about Andre Agassi’s tennis renaissance (tennaissance?). The lesson is that it takes a whole bag of very personalized tricks to effect personal change. She comes back to this in the final chapter.
Chapter 1 on the power of Fresh Starts
Temporal landmarks like round dates or momentous/memorable life transitions are good excuses to initiate life changes and they offer at least some modicum of motivation. This is the Fresh Start Effect. Also there are more such temporal landmarks than you’d think (every Monday can count as one). Getting vaccinated felt like a pretty good one when our book club was happening. And now New Year’s is around the corner.
Of course Beeminder lets you turn on the Fresh Start Effect like a faucet. Just create a graph! (Having trouble getting yourself to do that? Like you need a meta-Beeminder to get you to use Beeminder? We have a tip for that.)
Continuing the theme of the book being a Beeminder infomercial in disguise, Milkman also talks about how fresh starts can be poison. If you’ve built up a good habit, it can fall apart after a vacation. This is why you schedule breaks in your Beeminder goals and don’t just try to remember to resume your routine when you get back.
My biggest beef with this chapter is just that it fails to use Milkman and her students’ original term, “temporal landmarks”. She has at least one paper about this in which she uses the term “temporal landmark” 40 times and “fresh start” zero times. I guess she changed her mind on the best term to use. We shall allow it.
Chapter 2 on present bias and temptation bundling
Here’s a bit of incentive alignment wisdom from Mary Poppins and her spoonful of sugar: Make things you need to do be things you also want to do, in the moment. Milkman makes this concrete with some nice examples of what I consider her biggest claim to fame, temptation bundling:
- Get a page-turner of an audiobook that you can only listen to while at the gym
- Get a pedicure while reading a boring work document
- Watch a particular Netflix show only while doing housework
- Eat at your favorite burger place only while mentoring a problem student (probably she can’t use this one anymore now that she used the example in a best-selling book?)
- Only drink wine while cooking meals
She rounds out the chapter with some discussion of symbolic awards and praise being motivating, and how gamification backfires if you’re not bought into it. Fair enough.
Chapter 3 on commitment devices
It would be weird if this chapter didn’t mention Beeminder, and it does, barely. (Mostly the book mentions StickK, since Milkman is friends with the founders of StickK. Fair enough again.) But of course this chapter is just all about why things like Beeminder are the unchallenged king of behavior change. Which is nothing new for this audience. Our favorite part of this chapter is the story of Nick Winter and his banana-pants collection of commitment devices that he details in The Motivation Hacker.
Milkman goes on to talk about softer commitment devices like public pledges: they’re better than nothing but don’t hold a candle to cash commitment contracts.
She emphasizes the value of smaller and more frequent commitments. This is Beeminder’s whole schtick, turning a long-term commitment into a continual, daily commitment.
The tragic conclusion of this chapter is that, as far as commitment devices go, the world is divided into sophisticates and naifs. Most people are the latter and are just not cut out for Beeminder. We made this point in “The Type Bee Personality”. We agree it’s a minority of people who can successfully use commitment devices — particularly nerdy ones like Beeminder — but it’s a significant minority! 
Chapter 4 on forgetfulness and cues
This chapter mostly talks about trigger-action plans. I call them trigger-action plans (or TAPs) because that’s what people in the rationality community call them. Milkman doesn’t use that term. She refers to cues rather than triggers. Same difference. Also these are called “implementation intentions” in the psychology literature but I think we can all agree that that’s a horrible name and move on.
(Actually we’ve mentioned implementation intentions on the blog a couple times. Like Bee’s weight loss tip #3 in her guide to beeminding weight loss.)
My [favorite trigger-action plan]( https://blog.beeminder.com/should/ ” “I Resolve Not To Resolve; Or, The Anti-Resolution Resolution” ) — serving me very well for over 6 years and counting — is the following:
IF I hear myself saying “I/we should do X” THEN I take some immediate action that gets me slightly closer to X happening.
At our house, especially with our kids, we’ve been getting more and more mileage out of a meta-TAP, planning to forget. When you catch yourself planning to remember to do something (pick something up after work, etc) or making a request prefaced with “remember to”, that’s your trigger to create a specific TAP that will prompt you to take a specific action at the specific time.
Which is another point from this chapter: that timeliness is key. Timely as in T minus seconds. If you remind someone about something that’s even minutes in the future, it flies right out of their head immediately.
Another thing that helps is vivid, weird, or catchy cues. Milkman also makes the point — another very Beeminder-y one — that making plans (trigger-action or otherwise) makes you get specific and break big goals down. That alone is one of the most powerful change-achievers in the book, she says. Finally, as any pilot will tell you, checklists are lifesavers.
So I liked this chapter and I’d especially recommend it to anyone not already familiar with TAPs. Some of our book club participants (especially Alys) were just a little annoyed by how many of the examples could’ve been better solved by consistent use of a calendar with alerts set at appropriate advance intervals. And/or just putting sticky notes on your keys in the evening.
Chapter 5 on nudges and tracking
This chapter ends with something that might’ve made more sense in chapter 4, since it’s a nice corollary of the TAP concept. Namely, chaining habits together. Piggybacking, Milkman calls it. You can use an existing habit as the trigger for a new habit. Maybe you’re trying to get yourself to floss regularly so you put the floss on top of your toothbrush in a way that you’re prompted to pick up the floss when you go to brush your teeth.
The chapter begins with all the usual stuff about nudging and defaults. See our old post on Cranial Silicosis and Paths of Least Resistance for how we learned some of this the hard way. As we say when discussing UI design, defaults are the most powerful force in the universe.
Next comes elastic habits and the so-called limits of the power of routine. There’s a story of Routine Rachel vs Flexible Fernando. I was unimpressed with all this. Beeminder doesn’t need to care about your routines. You just commit to staying on track overall and then you can balance routine vs flexibility however you like. With the commitment in place, you don’t have to stress about those tradeoffs. (Much) more on this in the next chapter.
After that comes the so-called Seinfeld hack and the power of tracking. (Milkman loses points here for credulously repeating the myth that this has anything to do with Seinfeld.) This is another case of being the babiest baby step towards Beeminder’s power. See our old Seinfeld Hack post. And tracking, I mean, look at us.
My favorite bit in this chapter (already likely familiar to the uterus-having half of you) is the birth control pill placebo hack: You don’t need to take the pills during the week of your period but you don’t want to get out of the habit of taking the daily pill so there’s just a week of placebos in there every month. (I hope mentioning this isn’t an infohazard!)
Finally, there’s one thing in this chapter that I’m not sure I understand or appreciate. Milkman emphasizes that you need to drill habits until they’re automatic, like a paramedic drilling a life-saving protocol. Does she mean to literally schedule time to walk through the motions of habits you aim to cultivate?  She doesn’t give examples of actually doing this so I’m uncertain whether there’s a specific behavior change protocol here or if she’s just emphasizing that habits, both good and bad, can be, well, habituated. This is pages 122-127 if anyone can better extract the wisdom here.
Chapter 6 on confidence and advice and believing
This was a very self-helpy chapter. The first bit of ironic meta advice is that giving advice helps the advice-giver but backfires for the advice-receiver. (This is related to something we’ve discovered with customer support. Upside-down support, we call it.) Taking that idea to its logical conclusion leads us to Advice Clubs. Normally, unsolicited advice shakes a person’s confidence, but also giving someone else advice helps you by creating cognitive dissonance if you aren’t following your own advice. An Advice Club gets the best of both worlds. All the advice is solicited, by virtue of being given in Advice Club, and, most importantly, advice-givers are getting a commitment device. You feel like a hypocrite if you don’t follow the advice you’re giving someone else.
(Side tip that I picked up somewhere in the rationality community: Instead of offering advice to someone, pose an equivalent thought experiment. So instead of “you should do X” you can say “what do you think would happen if you did X?”)
So that was all pretty solid. Also in this chapter is what psychologists call, with an apparently straight face, the what-the-hell effect. Nothing earth-shattering here, just the observation that sometimes when you screw up with, say, your diet, you throw up your hands in dismay and upend your whole refrigerator directly into your gaping maw. Milkman’s conclusion is that you should set ambitious goals but allow mulligans. This is related to elastic habits, from chapter 5.
I hated that in chapter 5 and I hate it here. Mulligans are unbeemindery. Or, rather, Beeminder has something better. In fact, multiple better ways of getting the advantages that Milkman claims mulligans have. (There’s a study that Milkman’s mulligan claims are based on that we tore to shreds in the book club and which I’ll gloss over here. More about Milkman’s credulous citing of bad studies at the end.)
- In Beeminder you can aim to stay in the green and eat into that safety buffer for emergencies.
- You can cap your pledge at something affordable and choose to pay for the derailment and subsequent respite sometimes.
- You can schedule breaks or shallow spots in your red line if you plan ahead a bit for times when you expect more disruption.
- You can even schedule a random flat spot for no reason other than to give yourself breathing room if you’ve been skating the edge.
Beeminder is all about flexibility.
The idea of aiming to do something 7 days a week but excusing up to 2 days a week does seem psychologically powerful. But if you’re very akratic then it degenerates to just doing it 5 days a week and you might as well be clear that that’s the actual commitment.
This is Beeminder’s Anti-Leniency Principle. Anything that makes your goal easier in the short-term backfires. Grace periods, mulligans, 3-strikes policies, the “wrong lane” of the old yellow brick road — it’s all a bad idea. If you want more leniency you should set your bright red line more conservatively — maybe with a long initial flat spot — but it is and must be bright. Things like mulligans fuzz it up. It amounts to trying to psychologically trick yourself into maintaining safety buffer by pretending the bright line is closer than it is. This works for a while and then eventually fails.
What mulligans are really trying to accomplish is to specify two rates: This is how much I ideally want to do and this is the bare minimum. Beeminder focuses on the bare minimum and doesn’t give you much guidance on how you’re doing with respect to your aspirational rate. A workaround is to have two Beeminder goals: one for the ambitious rate with a low pledge cap and one for the bare minimum with a high pledge cap.
And Beeminder may streamline that in the future. But I think part of the appeal of mulligans is similar to the appeal of setting your watch 5 minutes fast. It’s a psychological trick you’re trying to play on yourself, hoping that your future self won’t wise up to it. Beeminder refuses to play such tricks. It’s all about an unambiguous bright red line.
Deep in self-help land
I mentioned this was a self-helpy chapter. There’s a fair bit of woo in it. Self-affirmation, etc. I’m clearly the wrong person to review anything like that, so if you like such things, assume I’m skipping over some good stuff here. But it did lead to one insight that I appreciated: Overconfidence is a huge cognitive bias but that’s because it’s so vital to ambition and getting things accomplished. Underconfidence is disastrous for agency and has no upsides. So you want to fight underconfidence and err on the side of overconfidence.
Oh yes, and growth mindset, which Scott Alexander has been critical of, but maybe mostly critical of the hype and exaggerated claims more than the common-sense notion of focusing on hard work and grit over the fragility that comes with focusing on natural talent. That’s a fine message.
Housekeepers and magic milkshakes
Let’s move on to my least favorite part of the whole book, starting with a study on hotel housekeepers who magically lose weight by believing that their jobs count as exercise. Or an even more implausible study claiming that if you believe your milkshake is low-calorie, your body produces different gut peptides or something. Milkman cites all of this completely non-sarcastically. It’s pretty bad. There’s no way this stuff replicates, right? (More on that at the end.)
Post-publication, on an EconTalk podcast, she seems to walk it back considerably. Generously paraphrasing her: “I mean, it’s just the placebo effect but that counts for something. Well, maybe not for much, and I wouldn’t expect the results to last for more than like a month or anything.” So, yeah. I was disappointed that she didn’t give those qualifications and caveats in the book!
The one anecdote I liked here was the (true, I checked!) story of George Dantzig. He was a student who copied down a couple unsolved math problems thinking they were homework. Not knowing they were major open problems obviously not solvable by a student, he proceeded to solve them.
Chapter 7 on conformity and social pressure
My “everyone knows this” in the preamble was a joke but I guess I’m also a bit serious. I’ll bullet-point what I distilled from this chapter so you can skim it to see if there’s anything you didn’t already know:
- Surrounding yourself with amazing peers makes you more amazing. Verified by Science.
- Beware the false consensus effect aka the typical mind fallacy.
- Copy-and-paste other people’s life hacks, but ideally things you seek out for yourself, not silver platters.
- And specifically ask successful people for tactics to copy.
- Social influence is influential. Social norms can be marketed. “Most other hotel guests reuse their towels, you wasteful waster.” See also our diatribe about fake data.
- Social influence doesn’t work or backfires if you’re too different from your peers or if the desired action is too abstract or long-term.
- Milkman is pretty excited about a social influence intervention that gets an 8% increase in voter turnout. I get that that’s a big deal for voter turnout but it doesn’t seem worth this much fanfare in the context of personal behavior change.
- Tell people your goals! Social accountability! Like we’ve been saying.
- Paying people $1 for gym visits in sync with a friend motivates them far more than $1 paid unconditionally for gym visits. Who’d have guessed?
- People will donate more if you give them an excuse to show off their generosity. Related trick: public sign-up lists for volunteering.
Chapter 8 on sustaining change
I already led with the main message of this chapter and the diabetes analogy. There’s not much more to say. Akrasia is a chronic condition! Even after ingraining something for years, when you take away the treatment, the effect decays pretty fast, says Milkman. So don’t do that.
That’s why most Beeminder goals are open-ended. Really systems not goals. Things like getting 10,000 steps a day or practicing piano for 20 minutes or writing 500 words per day. Every day, forever.
But does it replicate?
I mentioned that my least favorite part of the book was Milkman’s citing of what seem to me like horrible studies. More generally, I’m disappointed by her lack of appreciation of the replication crisis. She acknowledges it when asked but makes zero mention of it in her whole book, citing probably hundreds of social science papers. I checked and there’s only one instance of the words “replicate” or “reproduce” in the book: mentioning that a study from the 1800s has been replicated. That’s kind of worse than never mentioning replicability. It implies that you need to mention that a study from the 1800s replicates but you don’t need to mention it for more modern studies. Someone who didn’t know would assume that that means that modern studies generally replicate! It feels almost dishonest to blithely cite all these studies to back up your claims, knowing that the ones that haven’t replicated are maybe literally more likely than not to be bogus.
Here’s a trimmed transcript of Milkman responding to a question about the replication crisis on the EconTalk podcast:
I didn’t put in any studies that I believe are unreliable into my book and in fact, took some out as I learned things that made me question them. So, I tried to be pretty careful — which doesn’t mean that it’s all perfect. But, that was really important to me.
I will say, I’ve sort of been at the center in some ways of watching the replication crisis unfold. I’m in a department at Wharton with [Simmons and Simonsohn, studying p-hacking, etc]. I think it’s so important — I’m so glad that we’re getting things cleaned up. I guess I will say, I think there are plenty of things that aren’t true. But, there’s also lots of things that do replicate and are really robust.
For instance, present bias. That replicates. Commitment devices are valuable. That replicates. And, it’s not shocking that some of the things replicate that do. Because, honestly my sense is a lot of the things that don’t replicate are the things that looked like magic in the first place. So, when you constrain yourself, when you incentivize yourself to be differently, better things happen.
So, I tried really hard when writing the book to stay away from anything that felt or smelled like magic. I’m sure there will be something in the book that doesn’t replicate because almost certainly there’s always something. There are a lot of studies in there. I think about a seventh of the pages are my reference section. But, I stand by it. In general the themes are so robust and the things I focus on — many, many, many studies have shown them to be true. So, hopefully there’s not much in there that people should be worried about.
I think she’s absolutely sincere there but she’s ultimately saying that unless an unreplicated study smelled fishy to her, she assumed it was legit. That’s insufficiently appreciative of the breadth and depth of the crisis.
Even though Milkman believes in everything she cites (except maybe the magic weight loss stuff in chapter 6?), I think intellectual honesty should require more explicit asterisks when making claims based on social science studies that haven’t yet been replicated. It’s not enough to filter out studies that have failed replication.
(But it’s nice how the one thing she points out definitely does replicate is the most Beemindery stuff — present bias aka akrasia and commitment devices!)
Ok, requisite replication rant written. I’m still an academic at heart so I was compelled to say all that but most of you are reading this for perfectly pragmatic reasons. Like strategies for, I don’t know, how to change? Even if we wanted to be maximally cynical and assume that every study we don’t know replicates doesn’t replicate — as I largely did in the bulk of this review — the book is still full of important insights. For those already deep in the Beeminder community there won’t be too many surprises. But for everyone else, Milkman delivers.
 If we condition on people who read footnotes of Beeminder blog posts (or people who know what “conditioning on” means) then the posterior probability is far above 50%. I.e., we can pretty safely say that the minority of people who can successfully use Beeminder includes you.
There’s also a good chance you can successfully use Beeminder if you’re reading this footnote but mostly think of shampoo and butts when we say things like “conditioning” and “posteriors”.
 UPDATE: I failed to mention when this went to press that whatever Milkman means by deliberate drills of habits, plenty in the Beeminder community mean it very literally and explicitly. Check out The Hammertime Club in the Beeminder forum.
Image credit: Faire Soule-Reeves