Feynman’s Razor

Sometimes it feels like the trend online right now is to design everything to make you stupider, on the theory that if it asks less of you will go along with it more easily. For instance, Google results seem more "least common denominator" every month. No longer can I search for something subtle and get subtle results; their algorithm pushes me towards and unhelpful answers. Maybe it's because lots of people do respond better to this, so it shows up better in the data. Anyway I hate it. I would like to gradually learn more as I interact with things, and engage with the complexity in systems, not have it hidden from me.

I agree this is true if your goal is to inform. If, however, your goal is to increase clicks, decrease support tickets, and get engagement, I'd be surprised if being technically descriptive and accurate is better than dumbing it down to the point of inaccuracy.

None of that is to say that I agree with the goal of engagement over conveying information.

Funny thing is I think this about Feynman’s own writing for “the layperson”.

I think things like Feynman’s “little arrows” descriptions in QED only muddied and added to the mystique and mysticism of the physics he loved.

Which is interesting because his written lectures[1], though, in their breadth and complexity require effort, seem as if they are intended for experts while being approachable to ”the layperson“.

My only real complaint about those lectures is that even when I understood I rarely felt I had actionable tools for that new knowledge.

The best descriptions of physics I feel that can sufficiently inform “the layperson” are ones that implements the physics in code[2], or through numerical methods.

[1] https://www.feynmanlectures.caltech.edu/

[2] https://www.lpfp.io/

Feynman confused pedagogical skill with expertise in a given field. He insisted that experts could explain things. No, that's what a teacher does. Feynman was good at both and apparently had no one to challenge him on his claim that they were the same.

It's possible to be a great teacher without being an expert on given subject matter. And it's possible to have the most expertise but be inscrutable.

In the case of this quote, Feynman wasn't talking about expertise at all, but about explaining a thing with enough relevant details to be understood. Without those, a layman would not understand either.

That isn't a razor...

I struggle a lot with this in my work. I can’t assume that my readers are fluent English speakers, nor that they understand the topic. I can’t even assume that they care to understand. I have adopted a plain and unambiguous writing style that hopefully accounts for it. I am now experimenting with formatting that allows them to skim everything and still get the gist of it. In my opinion, nhs.uk is the best example to learn from.

People are not stupid, but they are busy, tired, lazy, or simply not that interested in the finer details. There is a balance between respecting their intellect and respecting their time.

I don't quite think the message-doesn't-exist example is caused by "dumbing down" too much. To me, the message "it no longer exists" is not the work of someone who, in trying to make layman understand, overdid it.

Instead, it's the typical kind of mistake made by writers who forget about context.

From the programmer's perspective, I speculate, the error _is_ indeed a case of something being non-existent. The code is written to fetch something from the server, the server says it's not there, and well, the application should tell the user about it.

Within the very narrow context of fetching it from the server, "doesn't exist" is correct. However, in the context of the user who's sitting there looking at the message, "doesn't exist" is simply false.

The message needs to be re-contextualized, such that it makes sense. "The message doesn't exist on the server, it exists on your computer at the moment, but just for a short while, so you better copy it".

It doesn't matter whether the user understand what a "server" is. Simply knowing that the message doesn't exist [at some place], is enough to make the whole sentence sensible.

Like if I go to a car mechanic, who just points to the whole car and say "it's gonna be replaced", it would be confusing. But if he says that "xxx is gonna be replaced", that's perfectly comprehensible, even if I don't know what xxx is.

My favorite comment was from lisper...

Cool! Someone noticed me!

:-)

(I find this noteworthy because I've been putting all this effort into writing a series of blog posts about the scientific method and I've gotten very little feedback on that. But this little throwaway comment, that gets attention! Life is funny sometimes.)

To be great is to be misunderstood, and Feynman was a great man. I'm still not sure who decided to abuse some quotes and invent this "razor".

I don't believe Feynman ever meant to overconstrain and corrupt information based on (likely incorrect) assumptions about his audience. There's no need to "dumb down" anything ever, not even for children.

The information is either there or it isn't regardless of who can comprehend it. Losslessness is non-negotiable. The real problem is how you present that information, not whether you should leave it out. The only time you should leave something out is if it has little to no relation to the rest of what you're trying to say.

This message has been deleted from the mail server, but Outlook still has it in its temporary cache on this device. You can copy the message contents, or discard it from the cache, at which point it will be permanently deleted.

Oh yeah also let's solve the "puzzle" from the middle of the blog post in Microsoft-y language everyone is familiar with by now.

"This message was deleted due to your organization's retention policy, spam filter, or administrator. It temporarily remains on this device, but is at risk of being lost forever. Do you want to save a copy?"

I believe in the real world Outlook doesn't even bother and has you and your message eat shit. If it was deleted it will never tell you about a cached copy and to the layperson it's just gone. The road to hell is paved with good intentions and indecision only makes communication harder.

Nah I like New Scientist and Quanta Magazine etc… to give me an insight into a complicated topic. It isn’t adequate for a professor to understand the subject to write a thesis sure, but there is no harm in more people kinda understanding something. This is especially true of the sort of dumbing down the big short film did, when the impact on society means everyone should sort of understand what happened Nd why the economy tanked.

Funny, I always thought Feynman's Razor was along these lines (from 1979 QED lectures in New Zealand, response to audience question):

Q: "Do you like the idea that our picture of the world has to be based on a calculation which involves probability?"

A: "...if I get right down to it, I don't say I like it and I don't say I don't like it. I got very highly trained over the years to be a scientist and there's a certain way you have to look at things. When I give a talk I simplify a little bit, I cheat a little bit to make it sound like I don't like it. What I mean is it's peculiar. But I never think, this is what I like and this is what I don't like, I think this is what it is and this is what it isn't. And whether I like it or I don't like it is really irrelevant and believe it or not I have extracted it out of my mind. I do not even ask myself whether I like it or I don't like it because it's a complete irrelevance."

As far as complexity and how to explain things to people without technical experience of the subject, the rabbit hole always goes deeper. Here's a nice quote from the rotation in space section of the caltech lectures:

"We shall not use these equations in all their generality and study all their consequences, because this would take many years, and we must soon turn to other subjects. In an introductory course we can present only the fundamental laws and apply them to a very few situations of special interest."

The original message is enough information to deduce what’s probably going on as an expert and not feel alien to the customer base.

What’s the point of conveying root causes to the user? In this case the error tells you what you need to know: copy the message. Send it again if you want.

The point of communication isn’t always education. This isn’t a razor. It doesn’t do anything to help you consider audience. It will lead you to bad conclusions.

If something can be explained without technical jargon in a way that satisfies novice and expert, what use is the jargon?

I feel the example given in the article is bad because it isn't useful to either person, and jargon doesn't improve or lessen it. It's simply a poor explanation, which makes the argument weak.

I miss Norton Disk Doctor, not because I work on drive recovery but because it had an interface that respected the user’s intelligence.

Every other storage-related system I’ve ever used either had inscrutable gibberish like:

   “Mode 5/7?” [Y]

    Are you sure? [N]

Norton meanwhile had several paragraphs of text explaining what every decision meant. It explained concepts inline. It provided an explanation of the benefits and risks. It let you make an informed choice instead of just hitting Y over and over and hoping for the best.

This kind of respect for the reader / user applies to all forms of technical writing such as manuals, user interfaces, blog articles, and API docs.

Instead of trying to dumb down the text for a “lay” audience, try to educate all audiences to become more technical after having read and understood what you wrote.

[1] That is verbatim from a Hitachi SAN array that was holding all of the data for a government department. The manual helpfully explained that this option toggles between mode 5/7 being either on or off. If you choose wrong the array will erase itself and kill your cat. Or neither of those things. Who knows?

The "This message can't be saved" text is not that bad, actually. The important thing is that the second part tells the user what to do. The first part could be rewritten "An error occurred" and that would be fine, too, since the details are not likely to matter to the user -- the important thing is what the user ought to do next.

A driver does not need to know the reason for a detour. All that matters is whether the alternative route is clearly indicated.

what the hell does this mean? The message doesn’t exist, but I can copy its contents? If I can copy its contents…why can’t I save it?? If it doesn’t exist why do I have to discard it??

Maybe it should be even less descriptive.

"This message will soon no longer be available. Make sure you copy the contents of the message before you discard if you want to use them later."

Talking about the cache doesn't help the user. Really the solution is to add a button at the end of the message that says "save message contents to device." Then the message is clear and easily actionable.

Something I’ve been doing recently is declaring a variable, writing a comment explaining clearly what the variable actually does, then renaming the variable to reflect the comment. If I have to explain the variable name too much then it’s not a good name.

The author introduces a distinction that I hadn't fully considered before: difficult writing is not all equally problematic. Being misunderstood is a problem, but how and why you are misunderstood matters.

Funny that the author himself failed to reproduce Feynman's quote: it should be "And there are now sixty two kinds of particles", not "And there are not sixty two kinds of particles".

Why this obsession with "all words must be known"? People are perfectly capable to understand the core of the message even if it contains unknown words. It happens not every time, it doesn't happen for example when each one of the words is unknown, but the message:

"This message has been deleted from the mail server, but Outlook still has it in its temporary cache on this device. You can copy the message contents, or discard it from the cache, at which point it will be permanently deleted."

can be understood. Even if the first sentence looks like a technical garbage, the second gives the clear instructions. And I don't believe that the first sentence is a complete nonsense to a layman. It can be understood as "message is halfway through the process of the deletion".

While there really is a trend to overly dumb down things, I'd just like to speculate that the "message no longer exists" part is simply these when someone adapted the reasonable error for a delete message, and forgot to actually check if that phrasing work anymore.

Small changes creeping up into something stupid is a common thing as well when everyone just does a tiny isolated part.

There's an association here to another topic which came up recently: a review of Matt Strassler's new book, Waves in an Impossible Sea [1]. Professor Strassler says a big motivation for writing this book was what he calls 'phibs': 'explanations' of physics so bowdlerized that they are not just uninformative, but at least misleading and sometimes outright wrong. The particular phib which got him started was a commonly-repeated attempt to give an idea of how the Higgs field gives rise to the masses of certain particles [2]

[1] https://news.ycombinator.com/item?id=40607307

[2] https://profmattstrassler.com/2024/04/16/why-the-higgs-field...

Simplicity is key, but over-simplicity is a killer.

“This message will self-destruct after reading.”

Yeah computing day to day sure is mired in euphemisms for math.

Cache is a variable. Server is a function. Anyone who graduated high school should know that much.

Hopefully we’ll start to move away from 1900s ignorant business machines euphemisms for applied physics

In UX, the right way to think about your users is not that they are dumb, but that they are very smart but very busy, and don't have time for your app's bullshit. It puts you in the right frame of mind when designing interfaces. It's not about dealing with dunces, but about being efficient with peoples time and attention.

The case of the original error message does not really match what Feynman was concerned about. The original error message was not too simple, but its first part was some hocus-pocus that made (almost) no sense -- the less, the more the reader is a layperson. Only an expert had a chance of guessing the explanation. In the quoted passage, Feynman instead was concerned about popular accounts of science that are not useful for the reader -- not because they are unintelligible or wrong, but because they are too simplistic and provide too less information.