Bayesian Storytelling

I launched a newsletter yesterday. So far, the feedback has been good. A few readers said that they felt drawn in by the writing. In any case, the purpose of the first few posts is simply to get myself warmed up. Any extra flutter the posts generate is a bonus. Amit Varma recommends not looking at the stats for a couple of years.

It taught me quite a few things. Particularly that I need to be smart about the data analysis. It’s important to remember that this isn’t a work project. It’s meant for a general audience, so there is such a thing as too much detail. For example, detailed handwritten notes on every single table in the HCES isn’t important. You could have dealt with each table independently. Only focusing on the tables needed for the problem in question should have sufficed. On the other hand, cleaning and denormalizing the data and releasing it on GitHub was a good idea; and the tweet announcing this has 10 reposts, 69 likes, 52 bookmarks, and has gained me 12 new followers. This is what Austin Kleon calls “showing your work” (Mahima Vashist says “your journal is your art”). Perhaps one needs to think carefully about what shareable and useful assets can be created in the service of a larger, transcending work.

In the same spirit, I think I’m close to discovering a storytelling framework for Being Bayesian. When I wrote the first few drafts, it was pretty exhausting. I’m lucky enough that writing has flown out naturally quite a few times, but this wasn’t one of those times. Several aspects of the post seemed to take on a life of their own, and I found it hard to weave them together. It was easier to break down the steps independently, and given them a name and a purpose.

The first step is the Prior, the default: the deepest assumptions and beliefs of the writer. Surround the reader with them, make it claustrophobic. Be like Charles Dickens writing about the carriage in London. The purpose of this is to draw the reader in. Embed them into your experience. Make them share the precise frame of reference that you have. Ensure that they have the same probabilistic “priors”.

The next component is the Trigger. This is what the story actually reacts to. This is what makes you write the post. This is what you ridicule or debunk. This is typically the first chronological step of the story, but don’t write about it first. The first step is to always establish the prior, and then throw in the trigger. The trigger will either be an outlier to the prior, in which case it gets branded “ridiculous” or “outrageous” to begin with; or it will sit well within the prior, in which case something else - the smell - will have given it away. Either way, slip it into the prior in a way which isn’t explicit or tactless, but the audience nevertheless sees it as a fly in the ointment. Satire, sarcasm and absurdity are useful tools here.

Then comes the Skepticism: a simple, surface-level, common sense-based examination of the trigger. It’s a back of the envelope calculation. No data or stats yet, just logic. This is why the trigger felt off, or why you smelled the smell. At this stage the OP still has the benefit of doubt. At this point, Hanlon’s razor still applies. Maybe they didn’t fact-check. Maybe they didn’t do a quick Google search. They might have made an honest mistake, it happens to the best of us, after all. Your role right now is simply that of a helpful assistant - one who’s interested in correcting the trigger, not demolishing it just yet.

But if you’re any good at finding BS, you’ll soon realize that the trigger is beyond redemption. Even after allowing for Hanlon’s razor and reasonable doubt, it turns out that the trigger was so horrendously off that you have to pull out the big guns. The guns are high quality data and solid analysis beyond reproach. Don’t forget that this will be exhausting. Do it first, before writing anything, if it helps. This stage is the Investigation. Once you’re in this phase, show no mercy and give no quarter. Demolish every single erroneous aspect of the trigger. But be extremely vigilant. The more brutal your attacks become, the more vulnerable you leave yourself to a counter-attack. Exceptional claims need exceptional evidence. This applies to you too. Again, the investigation must be beyond reproach. If you’re not confident enough about something, don’t say it. Even if the smallest, most irrelevant bit is found incorrect, you will lose credibility. Ask for as much help and feedback as possible, and rely only on high quality data. Double-check all the analysis. When that’s done, write it down in as cold and bland a manner as possible. Remember that you’re reporting facts, and facts don’t have feelings. Those come next.

The last step is the Posterior. By now it should be clear to the reader what this is - it’s the takeaway, the insight. It’s the opposite of the prior. It’s the negation of your beliefs. It’s something that shakes your world-view. But even to the most rational of your readers, this alone may not be enough. You still have to make the journey from insight to impression. For the insight to leave an indelible mark, for it to truly stick, you have to use more than a p-value. So now, tell one more story. A story which is the opposite of the one told in the prior. Everyone knows and believes Daniel LaRusso’s story. Now tell the story of Johnny Lawrence. But remember that the prior was quintessentially yours. And so too must the posterior. It’s a story that establishes that your change of heart wasn’t purely mathematical. It was personal, too. The trouble, of course, is that you may never have deeply experienced an opposite of your prior. So then, do it vicariously. Write about a movie scene which moved you. Write about a piece of fiction you might have read; which, while fictional, still reinforces the posterior. Or go find someone to talk to. Tell their story. Pathos is a powerful tool, but it means nothing without sincerity. This, then becomes the coup de grâce of the entire story1.

None of this is going to be easy. Sometimes satisfactory writing just flows out in a smooth, coherent stream - which is usually the result of having mulled something over and over in your head ad nauseam. Sometimes, even after having written for days, you still struggle to connect the pieces. Sometimes the story comes full circles, and sometimes it ends abruptly, anticlimactically.

All of this makes the whole process really draining. Be prepared for the mental toll well in advance. Then, write each stage as a separate, standalone essay. Worry about the flourishes, the segues and the centrality at the very end. Till then, be okay with imperfection. No amount of time is enough to craft a perfect sentence, or to draw the perfect chart which everyone will love. Time-box all stages.

Finally, remember that this is a template, a framework. It’s purpose is to make things easy for you. If it makes things harder instead, drop it like a hot potato.


  1. Matthew Dicks recommends that you should begin at the end, i.e. the starting of a narrated story must be the opposite of the ending. We’re doing exactly this, except we reverse the order - in our stories, the ending is the opposite of the beginning. Dicks’ recommendations are for oral narratives, and it’s well worth the effort to adapt them to writing. ↩︎