Showing posts with label QWERTY. Show all posts
Showing posts with label QWERTY. Show all posts

Thursday, November 14, 2024

Tech Check-In - How Good is the Latest A.I.? - Part I

Introduction

During the ‘90s, all a company had to do to get funding was throw around the word “Internet.” Then the dot-com bubble burst, and the venture capital tightened up, but during the aughts a company could still generate a lot of excitement by using the word “cloud.” The effect of that word wore off by the teens, when tech companies had to toss about terms like “disrupt,” “transformation,” and “Internet of Things,” but even when used together these didn’t act like much of a magic wand. Now, in the roaring ‘20s, any company mentioning “A.I.” indicates its intent to be perceived as a cutting-edge company worthy of massive funding and universal adoration. The effect is starting to wear off, of course, since there are so many poseurs. And yet, there does seem to be something magical about A.I., and I’ve put it through its paces over the years (scroll to the bottom to see a list of my posts).

Thus far I’ve been pretty skeptical of A.I. and how much it will actually “disrupt” the workplace. But the technology is evolving rapidly, so I think it’s worthwhile to periodically check in on its progress. Since I last blogged on this topic, Google has increased its generative A.I. capabilities, and OpenAI has upgraded its ChatGPT engine from GPT-3.5 to GPT-4. At the same time, A.I. technologies are facing increasing opposition from the publishing industry. In this post I will evaluate the following:

  • Google’s AI Overview and opposition to it from web publishers, myself included
  • Why the New York Times is suing OpenAI, and my own exploration of ChatGPT’s plagiarism
  • ChatGPT’s strides in taking a position and supporting it


Google AI Overview

This year, as you’ve surely noticed yourself, Google has rolled out a new feature: it distills the search results it considers most germane in order to provide a handy summary, which it displays above the search results. This is convenient for users, but is perceived as a threat by web publishers. As noted in this New York Times article, publishing executives are “worried that the paragraphs pose a big danger to their brittle business model, by sharply reducing the amount of traffic to their sites from Google.” One executive complained that “it potentially chokes off the original creators of the content,” with Google’s generative A.I. summary replacing “the publications that they have plagiarized.”

So … is this true? In a word, yes. Leveraging my own thought leadership in the realm of spelling, I asked Google, “is it kindergartner or kindergartener?” and it came back with this:


My first issue with this response is that it’s not even accurate. Both spellings (kindergartner and kindergartener) are widely accepted, with no widespread preference for one over another.

That inaccuracy aside, my main issue with this summary is that it supersedes the search results Google used to show, which put my blog post on this topic right at the very top. My post provides a far more complete answer, as everyone who’s viewed it—over 21,000 readers as of this moment—can attest. It bothers me that new readers to this post (who continue to accumulate, with 374 in the past seven days) may decrease in number over time since this query, along with Google’s new “featured snippet” and “People also ask” features, shove the link to my post down so far on the screen you have to scroll to see it. I feel like my post, as the number one “organic search” result, deserves higher placement. People wondering about this spelling are now getting a very concise, wrong answer instead of finding their way to my blog and getting the full story (and hopefully bookmarking albertnet). It’s a lose-lose.

I’m not trying to suggest Google is uniquely culpable for regurgitating publisher’s content while depriving them of readers. ChatGPT does the same thing, which gives me a nice segue to…

Why the New York Times is suing OpenAI

The New York Times recently filed a lawsuit against OpenAI and Microsoft, accusing them of using its content without permission to train AI models, including ChatGPT. The lawsuit claims that these companies have effectively taken advantage of the Times’ significant investment in high-quality journalism, turning it into products that compete with the newspaper’s own content offerings.

The above paragraph was generated by ChatGPT. I know it may seem silly of me to provide this citation, since ChatGPT isn’t a writer per se, and since its very purpose is to create text people can pass off as their own. But I personally don’t think that’s ethical, since ChatGPT is basically lifting its content from other sources without clearly crediting them.

That is to say, I find the Times grievance entirely credible. Without much trouble, I was able to see for myself how ChatGPT plunders my content, thus competing with my blog, without giving me sufficient citation. After I asked ChatGPT about the spelling of kindergartner/kindergartener, I asked it, “Is there a professional writer who has been inconsistent with her spelling of this word in a major magazine?” ChatGPT gave a very general response that included the statement, “The New Yorker provides a great example,” without actually giving the example (nor its source). To pin the chatbot down, I asked, “Can you give me a precise example of a New Yorker writer who has spelled kindergartner/kindergartener inconsistently?” It replied:

Yes, there is a known example of a New Yorker writer who has been inconsistent with the spelling of “kindergartner” and “kindergartener.” Rebecca Mead, a long-time contributor to The New Yorker, has reportedly used both spellings in her articles without a consistent preference. Her variations may reflect the linguistic blend between American and British English, as she was raised in England but has worked extensively in the United States… If you’re interested in more details on how and why both spellings are acceptable, various sources like Grammarhow provide further insights on regional preferences and dictionary entries for both “kindergartner” and “kindergartener.”

In terms of sources, ChatGPT’s text response mentions only Grammarhow, implying that this was the source of its answer, which is absolutely not the case. Only if you look above the text response and click on the down-chevron symbol to expand the list of sources will you see a link to my blog post:



A lazy or hurried user—i.e., one who wouldn’t bother to un-collapse the link listing and then click both links to review the source content—would never know who bothered to do the research to come up with Ms. Mead’s variant spellings, much less her background as a writer who has lived in both the UK and the US. Furthermore, since ChatGPT is designed to furnish copy-and-paste-ready content to so-called writers, it is enabling these users to plagiarize my (copyrighted!) blog without even knowing they’re doing it. As you can imagine, I have a big problem with this.

(In case you’re wondering why the Times is suing not just OpenAI but also Microsoft, it’s because of the partnership between the two companies. Microsoft has invested about $13 billion in OpenAI and there is a strong API integration between ChatGPT and Bing. This should answer the other question you may have had, which is why in the hell ChatGPT would use Bing—I mean, Bing, for crying out loud!—instead of Google.)

ChatGPT’s rhetorical improvement

As I have blogged about before, GPT-3.5 was barely able to form an opinion, much less support it. I asked it, about two years ago, to write a short blog post about why Campagnolo Super Record bike components are better than Shimano Dura-Ace, and it duly complied, concluding that “Campagnolo Super Record’s customizable shifting, lightweight construction, and iconic design give it an edge over Shimano Dura-Ace.” But then I asked it to write a blog post about why Dura-Ace is better, and it did that, too, with its final synopsis being, “Shimano Dura-Ace’s superior electronic shifting, durability, and modern design give it an edge over Campagnolo Super Record.” ChatGPT was basically brown-nosing me, telling me what it supposed I wanted to hear. And its supporting evidence for these arguments was verbose, vague, and overall very weak.

I see huge improvement in this realm. I asked the new ChatGPT to weigh in on a debate near and dear to my heart: the Dvorak keyboard layout vs. traditional QWERTY. I challenged it to compare an anti-Dvorak article by two economists in Reason magazine to my critique of that article in these pages. ChatGPT gave a very thorough synopsis of the positions, with far greater concision than GPT-3.5 could have achieved. What’s more, it went beyond just summarizing them: it came to its own conclusion, which suggests a semblance of actual understanding of the core issue. Here is its closing paragraph:

Albert’s argument appears stronger when considering individual typing efficiency, as he directly addresses typing mechanics and user experience, whereas Margolis and Liebowitz’s argument is more focused on economic theories of adoption. While Margolis and Liebowitz provide a macro-level view, Albert’s micro-level approach may resonate more with typists who are exploring the ergonomic benefits of Dvorak. Thus, for those prioritizing typing speed and comfort, Albert’s perspective might be more compelling, even if the QWERTY layout remains dominant due to historical inertia rather than functional superiority. 

The two (dipshit) economists would really wince at this, since they were trying to deny that historical inertia is responsible for the ongoing ubiquity of QWERTY. They were attempting to discredit the Dvorak layout since their delusional economic theory depends on them refuting the assertion that inertia (aka lock-in) can trounce the actual superiority of a product. I was so chuffed with ChatGPT’s conclusion, I took a victory lap and asked it who is the better writer: that is, which article has greater rhetorical power and more style. ChatGPT responded with a nicely organized and supported essay which concluded thus:

Dana Albert has more rhetorical power for a general audience due to his humor, personal anecdotes, and engaging style. Liebowitz and Margolis provide a strong argument for those interested in the economic perspective, but Albert’s style is more entertaining and may leave a lasting impression on readers curious about the Dvorak vs. QWERTY debate.

Compared to earlier versions, the modern ChatGPT is far superior. Rather than barfing up reconstituted content from the Internet, it really does appear to be applying judgment and performing true analysis. Notwithstanding my satisfaction at having my ego stroked by this disinterested third party, I’m actually kind of frightened by how closely GPT-4’s output resembles actual human thought. Whether or not A.I. will steal all our jobs, it does appear ready to displace lesser economists.

Tune in next week…

I had originally intended to cover two more topics in this post: A.I. advances in poetry and art. Alas, I see I am out of room (or more to the point, you are out of patience) so come back next week for Part II. It will provide a thorough examination of ChatGPT’s ability to write poetry (iambic pentameter and dactylic trimeter) and to create original art per the user’s specifications. (In case you were wondering, I did use ChatGPT to create the art you see above for this post. More on that next time.) Until then, you might check out the below links for more posts on A.I.

Other albertnet posts on A.I. 

—~—~—~—~—~—~—~—~—
Email me here. For a complete index of albertnet posts, click here.

Monday, January 18, 2016

AI Smackdown - Moto vs. Cortana vs. Siri


NOTE:  This post is rated R for mild strong language.

Introduction

I already blogged about my Android phone—not the way a professional critic would, but in terms of what’s actually interesting about it, which to me is the Artificial Intelligence angle.  Actually, “artificial stupidity” was the point:  a phone playing dumb so it can play favorites.  I’ve also blogged a bit about AI in general, with Apple’s Siri agent as a case study, but that was before I had an iOS device.  Now, I own all three platforms:  Moto/Android; Siri; and Microsoft’s Cortana.  In this post I compare and contrast them: not because I’m going to help you choose, but to try to make you laugh.  And you might have something interesting to scratch your head about later.

(I’m not going to try to differentiate between the terms Google Now, Android, and Moto.  They all meld in my mind.  If somebody protests that there are massive differences, I won’t be offended if you go read his or her blog instead.)


Cortana

I’ll start with Cortana because it can be dispatched very quickly.  If you type something into the Cortana field in Windows 10, it does a Bing search.  Bing is a little bit like using a Curad bandage instead of a Band-Aid, or using Hunts ketchup instead of Heinz, or wearing Sears Toughskins jeans instead of Levi’s.  It just isn’t done.  I don’t actually care if Bing works just fine.  It’s Bing, which means it’s not Google. “Let me Bing that for you.”  Give me a break.

Moreover, before you can get into the voice recognition stuff, you have to deal with this frightening disclaimer:


Yes, I know all this meddlesome snooping is just to “tailor the experience,” but it’s the online equivalent of your tailor saying, “To get the fit right on these trousers, I’ll have to reach in and fondle your balls.”  The explicit information Windows wants to use is bad enough, but that wide-open phrase “and other information” is just over the top.

Besides, “Cortana” sounds like a new model of Hyundai.  You know what Microsoft?  It’s over.  You lost.  You’re just a PC software company.  Stop trying to act “mobile.”

Android/Moto

I won’t go into a lot of detail about how well my Android phone responds to voice commands, because a) I already did that, here; and b) as I describe the Siri experience, I’ll compare it to Android/Moto as I go.

The really lame thing about Siri

Imagine if, before engaging with a person, you had to go push a button on the person’s chest.  In most cases, this would be absurd.  (With my kids, who never hear my commands, it would actually be an improvement.)

I think it almost goes without saying that voice response is a minimum requirement for any kind of AI.  The Siri demo I watched way back in 2012 did feature voice activation on the iPhone.  But oddly enough, for the Siri voice response to work on the iPad, the iPad has to be plugged in to an electrical outlet.  That is just so bizarre!  I mean, the iPad’s portability is the whole point, isn’t it?  What’s next for Apple:  an iPhone you plug into an RJ-11 jack?  This is ridiculous.  If I’m sitting at a desk next to an electrical outlet, I might as well be using a laptop.

There’s something else really lame about the iPad:  a limitation that hasn’t existed on an Apple product since the Apple II computer.  But I’ll get into that later.  Better to keep you in suspense.

Siri voice response:  up to snuff?

In general, Siri tries to have a bit more personality than Moto.  For example, if I ask my Droid, “Do you love me?” it shows me a song called “Do you love me?” by the Contours.  (My younger daughter, who has not seen “Her,” put me up to asking this question.)  When I asked Siri if she loved me, she responded, “I respect you.”  And when (again at my daughter’s behest) I asked Siri, “Have you ever gone to the bathroom?” she replied, “Who, me?”

Is this cheekiness a good thing?  Well, Siri’s responses may strike you as funnier that Moto’s.  On the other hand, when I asked my Droid about using the bathroom, I got a list of hits pertaining to using the wrong restroom (i.e., the one intended for the opposite sex).  It was a very funny list, linking to some amusing sites.  (You may be wondering:  have I ever used the wrong restroom?  Well, yes, once, purely by accident.  I was in there doing my business and thinking, “What kind of public restroom doesn’t have urinals?”  When the answer suddenly came to me, I hightailed it right on out of there.)

Sometimes Siri’s personality gets in the way.  For example, I asked Siri, “What time is it?” and she responded, “At the third stroke, it will be 16:26.  Beep.  Beep. Beep.”  This was more confusing than amusing, and besides, it was inaccurate:  the actual time was like 4:25:30.  I would rather Siri have a more reliable connection to the NIST Internet time servers than a zingy response.

If your desk is as cluttered as mine, being able to summon your phone by voice is very handy.  When I call out my keyphrase for my Droid, it makes a pretty loud two-tone beep to let me know it’s listening.  The iPad beep is much quieter.  Neither device responds in any useful way to the question, “Where are you?”  Siri says, “Wherever you are, that’s where I am.”  This isn’t that funny, and for most people wouldn’t even be true.  (I bring my iPad, Droid, silverware, and all other valuables with me wherever I go, so don’t bother burglarizing my house.)

 This is where these devices are inconsistent.  Their programmers need to decide if the device should have a sense of self or not.  Siri speaks in the first person (e.g., “Who, me?” and “I respect you”), but when I say, “Hey Siri, how’s your battery doing?” she has no idea what “your” means.  She replies, “My apologies ... I couldn’t find those stocks.”  Pretty useless.

When I tell my Droid, “Find my phone,” it makes this cool sonar sound continuously until I find and silence it.  When I tell Siri “Find my iPad,” she tries to make me turn on Location Services.  Look, Siri, if I could do that, I’d know where you are, and I wouldn’t be asking.

I’ve often thought that one of the most useful features of voice response would be getting help configuring the device.  So I said, “Hey Siri, turn on your flash.”  She replied,  “Who, me?”  I decided some context might help, so within the camera app I said to turn on the flash.  Siri replied, “It doesn’t look like you have an app named ‘flash.’  If you’d like, I can help you look for it on the App Store.”  I just don’t think this is that difficult a concept.  You have a camera.  It has a flash.  Turn it on.

The Droid does respond to “Take a selfie.”  It ought to say, “I can’t,” because its camera is basically its face.  But its reaction, which is to launch the camera, put it in selfie mode, and set a self-timer, is actually fairly useful, at least for the hands-free breed of narcissist.  When I tell Siri “take a selfie,” she says, “You’ll need to unlock your iPad first.”  This isn’t very helpful, and in fact isn’t even true.  As my older daughter discovered, the iPad can be used as a camera even by somebody who lacks my fingerprint and passcode.  And when I follow Siri’s instructions and unlock the iPad, it does go into camera mode, but not selfie mode.  As regards this command, Siri is fairly incompetent.

Something that bothers me about my Droid’s AI is a certain lack of resourcefulness.  When I ask it, “How do I look?” I think it should activate the camera in selfie mode, to use as a mirror.  Or It could really wow me by saying, “Your hair is a mess.”  (When you consider modern digital camera technology, which can tell if a subject’s eyes are closed, this hairdo check actually seems quite doable.) 

I asked Siri, “How do I look?” and she really stumbled.  She kept hearing, “How do I luck,” which she should have automatically revised because it just doesn’t make any sense.  One time, she thought I asked, “How do I lurk?” and replied, “I found something on the web about ‘how do I lurk.’ Check it out.”  That’s really unfortunate.  I wonder what kind of banner ads and spam I’ll get now that the Internet thinks I’m a stalker.

Finally Siri heard me right and replied, “Judging by your voice, I’d say you must be fairly attractive.”  Clever, but also kind of patronizing.  I mean, it’s bad enough asking an inanimate object such a personal question, but to be damned with faint praise ... that’s pretty pathetic.  I think Siri should be generous and say, “I would so go to bed with you.”

At least Siri respects my privacy.  When I said, “Get me home,” she replied, “I don’t know your home address.  In fact, I don’t know anything about you.”  I found this really reassuring, especially after Cortana’s attempted shakedown earlier.  (Yes, Siri did ask me to go into settings and identify myself, but didn’t require it.)

I have to say, though, there’s something a bit creepy about Siri.  When I ask something complicated, such as a question regarding navigation, there’s this little blurry light that bounces back and forth along the bottom edge of the screen, which reminded me of something sinister.  After racking my brain for awhile I realized what:  the single roving eye of a Cylon from “Battlestar Galactica.”  Is Siri some kind of kindred spirit to the AI powering the Cylons?  If so, that doesn’t reflect well ... the Cylons were really pretty stupid.  They always went down like bowling pins.

Another creepy thing:  when I said, “Hey Siri, lock my iPad,” she replied, “I’d like to, but I cannot.  My apologies.”  This almost gave me chills.  It brought me right back to the HAL 9000 in “2001 – A Space Odyssey,” when Dave says, “Open the pod bay doors, HAL,” and HAL replies, “I’m sorry, Dave.  I’m afraid I can’t do that.”  Who is Siri’s master:  me, or Apple?

Feature parity with Moto?

If I were designing Siri, or working on an update package, I’d pay close attention to what the competition is doing.  I’d make sure, for example, that anything Moto could do, Siri could do better.  This evidently hasn’t occurred to Apple, because there are all kinds of commands Moto can handle that Siri cannot.  For example, if you ask Moto, “What’s up?” it will trawl through your appointments, e-mails, etc. and give you an update.  I asked Siri “What’s up?” and she said, “I’m thinking about pie.  Mmmmmm.” 

If you tell Moto, “Talk to me,” it will announce incoming calls and texts for the next 30 minutes.  This doesn’t occur to Siri, who responds, “I’d really prefer it if you talked to me.  Tell me your hopes, your dreams, where you’d like to make a dinner reservation.”  So I told Siri, “I hope my dinner is yummy tonight.”  She replied, “I don’t know what you mean by ‘I hope my dinner is yummy tonight.’  How about a web search for it?”  Not a very good listener, since she specifically asked me to tell her my hopes!  Just lip service.  I said, “I dream of being rich and famous one day,” and got the same “I don’t know what you mean” response.

I said, “Hey Siri, play Beethoven on YouTube.”  She replied, “You don’t seem to have an app named ‘YouTube.’ We could see if the App Store has it.”  Don’t play dumb with me, Siri!

I asked Siri to zap my screen (which is how Moto is told to take a screen snapshot).  Siri kept hearing “zapped my screen” (which resulted in a web search) and then eventually heard “zap ice cream,” and—bizarrely—pulled up a Dairy Queen in Beulah, North Dakota.  I’m not kidding.

Since screen snapshots are really useful to bloggers, I kept trying:  “Hey Siri, take a screen snapshot.”  To my great surprise, Siri didn’t play dumb, but simply refused:  “That’s beyond my abilities at the moment.”  Huh?  No way is this beyond her abilities.  I managed to learn (no thanks to Siri) how to get a snapshot (pressing two far-flung buttons at once).  So it can be done.  Why can’t Siri do it?  Is she a bit ... simple?

You may be wondering how I know so many cool Moto commands.  It’s because you can say, “Get a list of commands,” and Moto provides one.  I told Siri, “Get a list of commands” and though—as you can see—she did hear me right, she decided just to show me a map of the nearest Coast Guard station.  WTF!?



Is Siri the best at anything?

Okay, I’ve been pretty harsh on Siri here.  Is she better than Moto at anything?  Well, yes.  I think her navigation is better.  I just asked Moto, “Where is the nearest pizza place?” Moto replied, “Here are the listings for ‘nearest pizza place’ within zero point eight miles.”  The nearest place—Gioia Pizzeria—was listed first among non-paid entries, but at the top of the screen was an ad for Little Caesars $5 Pizza, which a) isn’t nearby, and b) isn’t even pizza.  (I don’t know what that stuff is, but it ain’t pizza.)  Meanwhile, if I were trying to get this answer without having to look at my phone—like, if I were driving—this written response would be useless.  (At least Moto did better than when I first blogged about this, when it lied and said Zachary’s Pizza was the closest.)

Here, Siri did better.  She replied, aloud, “The nearest one I found is Gioia in Berkeley, which averages 4½ stars and is inexpensive.  Would you like to try it?”  Presumably if I’d said yes, she’d have navigated there.  Instead I said, “Actually, Siri, it’s pretty expensive.”  To which she replied, “I’m sorry.”  Well played, Sir[i]!

It’s in the realm of a more specific request where Siri really shines.  I asked, “Where’s the nearest deep dish Chicago style pizza place?”  She showed me Zachary’s, which is correct.  I’m pretty impressed, especially since when I asked Moto the same question, it showed me Giordano’s and Lou Malnati’s, both of which are in Chicago.  As you can see, Moto clearly heard “nearest” correctly, but somehow missed my meaning.


The second really lame thing about iPads

Earlier I complained about how you have to plug in the iPad to get voice activation, and promised to reveal another huge shortcoming.  I doubt you’ll immediately grasp how lame this next one is, but here goes:  Apple iOS doesn’t support the Dvorak keyboard layout, which is  more efficient than QWERTY and has been supported by Apple since the Apple IIc.  According to Wikipedia, the IIc  “had a mechanical switch above the keyboard whereby the user could switch back and forth between the QWERTY layout and the Dvorak layout.... The IIc Dvorak layout was even mentioned in 1984 ads, which stated that the World’s Fastest Typist, Barbara Blackburn, had set a record on an Apple IIc with the Dvorak layout.”
                                                               
I haven’t been able to find anything on the Internet about why Apple decided not to support Dvorak on the iPad.  I guess the default answer for their product choices—“Because we’re gods, and we can do whatever we want!”—will have to do.  It’s so frustrating, since this has got to be really simple to do in software.  It would probably take some Apple developer about five minutes.

But why should you care, since you type on QWERTY anyway?  Well, consider the security ramifications of encouraging third party developers to create such fundamental utilities as keyboard software.  After installing Fleksy, a free Dvorak-enabled app, I messed about with the iPad a little, wandered off to do something more useful, and then realized, “Duh, I’ve just done something really stupid.”  What better way to steal somebody’s keystrokes than to create an app that quite obviously has access to everything I type?

At first I told myself this was no big deal.  After all, I’m mainly using the iPad to browse the web, and I don’t kid myself that my every move on the Internet isn’t already tracked, and not just by the NSA.  (By the way, keep up the good work, guys!  Thanks for keeping me safe!)

But of course, there’s the little matter of passwords.  I felt like I’d just given away the keys to the kingdom, or at least to the two websites I’d logged into (my bank and my e-mail).  I was all set to go change those two passwords, but first decided to see how hard it is to switch iPad keyboards on the fly, so going forward I could type passwords with the native Apple iOS keyboard.  Perhaps there would be a function key right on the soft Fleksy keyboard to simplify this switch?  And then I noticed this:



It might be hard to tell, but in the first snapshot above, the cursor is in the Username field.  In the second snapshot, the cursor is in the Password field.  The same flag that tells the OS to obscure the password (i.e., showing ******* instead of what’s typed) tells the iPad to switch to the standard Apple iOS keyboard.  So those passwords I typed before? I’d typed them on the standard QWERTY  keyboard without even realizing it.  Those passwords weren’t at risk of being intercepted by the Fleksy keyboard app after all.  That’s pretty clever of Apple, isn’t it?

Of course, when your cleverness only serves the mitigate the downside of your pointless shortcoming, it’s actually a lot less impressive.  Hey Apple, why not support the Dvorak layout to begin with, like you did with the Apple IIe, the Apple III, the Macintosh, the Quadra, the PowerBook, the Performa, the iMac, the iBook, and the MacBook?

At the time of this writing, Apple is sitting on over $200 billion in cash.  Couldn’t they spend a few bucks to match the features of their own earlier products?  Hell, they could probably get an unpaid intern to do it.  I, for one, am not feeling the love ... even if Siri does claim to respect me.

Wednesday, June 24, 2009

The Case Against Margolis & Liebowitz

The Defendants

I type a lot. In 2002, plagued by hand and wrist pain and fearful of developing a repetitive stress injury, I decided to switch to the Dvorak keyboard layout. Dvorak is an alternative to the traditional QWERTY keyboard. It is named after its creator, August Dvorak, who designed it for maximum efficiency. The QWERTY layout, meanwhile, had originally been designed to keep early typewriter keys from jamming. Unlearning QWERTY and learning Dvorak was very difficult (I’d tried to convert years before and had failed), but since I successfully made the switch I’ve never suffered hand and wrist pain again.

Naturally, I’d like to encourage others to adopt this layout, and I take up that challenge in another blog post. I know from experience that Dvorak is more efficient than QWERTY, and that its benefits justify its learning curve, but I don’t expect to be taken as much of an authority. So, I did a little web research looking for studies of its efficiency. In doing so, I kept coming across an aggressively negative article about Dvorak by a pair of economists.

Right away I was nonplussed: what were economists doing studying typing efficiency in the first place? The answer is, these two take an interest in Dvorak because it is the poster child for an economic theory they refer to as “lock-in,” which is the (alleged) ability of inferior products to command inappropriate market share. The “lock-in” debate concerns the extent to which factors other than a product’s performance, such as an early foothold in the market or strong-arm market practices, can largely shape the competitive landscape.

Other classic examples of this economic theory are gradually losing their punch. For example, the Apple vs. Microsoft case study has become less relevant, as the obvious inferiority of MS-DOS to the Mac OS has given way to the arguable similarity of features between Mac and Windows. Or take the case of Betamax vs. VHS: once the source of lively debate, both video formats have become historical footnotes given the rise of the DVD. In contrast, the durable near-ubiquity of the QWERTY layout remains, justifiably, the perfect example of how performance and inefficiency can continue to take a backseat to sheer societal inertia.

The evil of two lessers

For philosophical, intellectual, and political reasons, Liebowitz and Margolis don’t like to accept that lock-in is a legitimate economic theory. I suppose they prefer to believe that the free market can solve all problems; in any case, it is essential to their position that Dvorak be discredited. So, despite the fact that they dwell in the abstract realm of economic theory instead of the literally hands-on physical realm of human/machine interfaces, they position themselves as authorities and assail the Dvorak layout with gusto in their article “The Fable of the Keys” in the Journal of Law & Economics, and (in shortened form) in their article “Typing Errors” in the June 1996 edition of Reason magazine.

“Typing Errors” article is pretty well written. It’s glib and polished, and the uncritical lay reader can be forgiven for being swayed by it. Even the normally unflappable Cecil Adams of The Straight Dope, after initially decrying the inefficiency of QWERTY, was taken in by “The Fable of the Keys”, falling on his sword and disavowing the entire content of his original article.

On closer inspection, however, Margolis and Liebowitz’s paper turns out to be poorly researched, and its arguments weakly constructed. The article has two fundamental problems. For one thing, the authors’ critique is targeted mainly at existing efficiency studies of the keyboard, rather than at the keyboard itself. Bad studies don’t mean a bad keyboard! Meanwhile, their article is far too focused on the feasibility of retraining QWERTY typists on Dvorak, not the ongoing benefits of the Dvorak layout once it’s been learned. The difficulty of retraining is beside the point: we don't need today's typists to unlearn QWERTY and learn Dvorak—we need tomorrow's typists to learn Dvorak to begin with.

A brief examination of these two fundamental flaws ought to be enough to thoroughly discredit the Margolis/Liebowitz thesis. But I won’t stop there. Because these hotshot economists have published a crappy article in a highly regarded journal, and because their quest to make an arcane point about economics has added to the inertia that prevents widespread adoption of a useful innovation, I aim to systematically dismember their argument to expose the full range of its flaws. Yes, your unsung blogger will take on the fancy eggheads using nothing more than logic and first-hand experience.

Red Herrings

The Margolis/Liebowitz argument begins with a section called “Tainted evidence for Dvorak.” They argue that Dvorak’s own study wasn’t subject to sufficient controls; they describe, for example, how Dvorak “compared students of different ages and abilities (for example, students learning Dvorak in grades 7 and 8 at the University of Chicago Lab School were compared with students learning QWERTY in conventional high schools).” After dispatching the large body of August Dvorak’s work in a few paragraphs, they proceed to a Navy study about the Dvorak layout. They find flaws with the Navy study as well: for example, “The participants’ IQs and dexterity skills are not reported for the QWERTY retraining group.” They cite differences in the study’s methods of measuring performance between the two layouts. They point out that August Dvorak, a Navy man, was the top expert in the Navy study. Margolis and Liebowitz present this last bit like it’s some sort of a conspiracy, and decry the fact of August Dvorak’s financial stake in the design.

Most of these grievances seem pretty nit-picky to me. The quality of the schools in the study wouldn’t seem to be a big deal, since plenty of people learn to type using simple software. (I learned Dvorak using a very basic website.) The lack of IQ benchmark strikes me as laughable; after all, Stephen Hawking, one of the foremost minds of our time, can’t type for beans. (Note to Margolis and Liebowitz: if you’re reading this, which I truly hope you are, please don’t build your rebuttal around that statement—it was a joke.) The financial interest Dvorak had is a bit more troubling, but don’t companies routinely fund their own performance studies? This is called marketing. (“Your mileage may vary.”)

But fine, let’s give these two the benefit of the doubt. Let’s assume that both the Dvorak study and the Navy study were complete frauds, conducted by deranged, dog-kicking sociopaths whacked out on coke and smack. Does that mean the Dvorak design isn’t more efficient? Of course not. Osama bin Laden could publish a pack of lies about the Dvorak design tomorrow, and it’s not going to make me type any slower.

Narrow interpretation

The next section of the Margolis/Liebowitz article is called “Evidence Against Dvorak,” and focuses on a government General Services Administration study. Margolis and Liebowitz describe the study as “a carefully controlled experiment designed to examine the costs and benefits of switching to Dvorak.” The ten subjects of the study took “well over 25 days” (whatever that means) to catch up to their old QWERTY speeds, after which their progress slowed. Meanwhile, a control group of QWERTY typists showed greater ongoing gains in speed than the Dvorak group. Based on these results, the director of the study concluded that “Dvorak training would never be able to amortize its costs.”

Is this really evidence against Dvorak? Note that the goal of the study wasn’t to assess the actual efficiency of Dvorak compared to QWERTY, but rather the merit of retraining QWERTY typists at government expense. The conclusions of the study itself, meanwhile, beg a lot of questions:

  • Is twenty-five days really an unreasonable amount of time to unlearn an automatic, unconscious skill and learn a new one? In other words, is it not possible that the study was ended too soon? (My own speed continued to increase for years after I learned Dvorak.)
  • Is speed the only measure of the validity of a keyboard design? Did anybody bother to ask the participants which keyboard was easier on their hands and wrists? Were the long-term costs of repetitive stress injuries even considered?
  • Is ten typists a large enough sample?

Margolis and Liebowitz go on to narrowly interpret the results of other studies. They cite a 1973 study of six typists that found that after 104 hours of Dvorak training, typists saw a 2.6% increase in speed. I take this as evidence for, not against, Dvorak given that, by my own rough calculations, I will spend around 45,000 more hours typing by the age of seventy. A mere 104 hours of training, after which my hands and wrists stop hurting, seems like a good investment to me. This 1973 study is only “evidence against Dvorak” because Margolis and Liebowitz label it as such.

Meanwhile, a fundamental difference between Dvorak’s own study and these others is that Dvorak assessed the ease with which children learn the Dvorak layout, and the children's ultimate results thereafter. He was looking to a future generation unhindered by the need to unlearn an obsolete keyboard layout. The “Typing Errors” authors don’t seem to appreciate this difference in approach, framing the debate only in terms of the difficulty of retraining.

More red herrings

From here, the Margolis and Liebowitz spend one brief paragraph referring in the most general terms to other works denying the validity of Dvorak. They cite “other studies” without naming them, and conclude “The consistent finding in ergonomic studies is that the results imply no clear advantage for Dvorak.” What ergonomic studies? Whose findings?

I wouldn’t dwell on this vagueness so much if Margolis and Liebowitz didn’t then proceed to blather on for eight long paragraphs, in their section titled “QWERTY’s competition,” about competing early typewriters that lost out to QWERTY in speed tests back in the 19th century, decades before the Dvorak layout came into existence. Margolis and Liebowitz conveniently neglect to mention that the only record that matters—the current speed record—was set on a Dvorak keyboard, with a words-per-minute rate of 212, significantly faster than anything anybody has done on a QWERTY. (See for yourself: http://www.answers.com/topic/typing; search within the page on the text “Blackburn.”) I find it absurd that these two economists see fit to so smugly discredit the validity of early Dvorak studies when their own research ignores any typing records set after the year 1889.

Hubris

The economists, apparently drunk on their own bathwater, go on to boast that “we published a more detailed version of this material in a Journal of Law and Economics article titled ‘The Fable of the Keys.’ This journal is well known and has published some of the most influential articles in economics. In the six years since we published that article there has been no attempt to refute any of our factual claims, to discredit the GSA study, or to resurrect the Navy study.”

These guys shouldn’t confuse widespread apathy on the part of their readers with tacit agreement. The fact is, their readership is almost entirely comprised of entrenched QWERTY users who aren't in a position to judge Dvorak for themselves. If Margolis and Liebowitz wrote an equally poor paper about, say, childbirth not actually being that painful, you can bet they’d meet with plenty of dissent. (Meanwhile, a month after Margolis and Liebowitz made this boast, Reason magazine—the publisher of “Typing Errors”—received, and printed, a scathing rebuttal to the original article.)

Speaking of dissenting opinions, where are the successful Dvorak converts in “Typing Errors”? Did Margolis and Liebowitz’s research not manage to find any? It’s actually not hard to do. I checked out an opinion piece in the New York Times and found fifteen comments (not counting my own) posted by happy Dvorak converts. Okay, not a huge number of people, but it’s just one website; besides, the GSA study—Margolis and Liebowitz’s centerpiece—had even fewer. And any one of us Dvorak converts has a legitimate real-world perspective on the merit of the Dvorak design, which strikes me as a lot more valid than an argument based solely on the available literature of others. (Would Margolis and Liebowitz refute the benefit of the two-button computer mouse just because reams of scientific performance data aren’t available to substantiate its utility?)

Conclusion

“Typing Errors” assumes that the merit of a product design can be creditably evaluated solely on the basis of existing literature. Its authors ignore obvious questions, such as how a layout like QWERTY, designed decades before the advent of touch-typing, could possibly be as efficient as one designed with touch-typing in mind. They apparently fail to notice the obvious failings of QWERTY, such as the scattering of indispensable vowels across the board with the lowly semicolon getting a prime spot on the home row. They’re looking at decades-old studies instead of at the keyboard they’re typing on.

This article should serve as a cautionary tale about the perils of making up your mind in advance of your research, and tailoring your interpretation to suit your thesis. It’s a real pity that the quest of a couple of academics to make an arcane point about economics has managed to mislead the public about something more important. Economic theory aside, Margolis and Liebowitz are hindering the adoption of a technology that can offer tangible benefits.

The Case for Dvorak


National Safety Month

As you may know, June is designated as “National Safety Month” in the U.S. by the National Safety Council. Perhaps it’s no coincidence, then, that I stumbled across two separate news stories the other day covering the same strange topic: acute computer-related injuries. An ABC news story warns readers that “computers are not play toys” and cites the risks of crushing, strangulation, and electrocution, though it also concedes that the percentage of ER visits due to PC injuries was just 0.008 percent. The other story, citing a different study, finds that more than 21 percent of injuries documented were from computer equipment falling on a person.

Slow news day, huh? But accidents aside, isn't there a much bigger story about the physical danger posed by PCs? I'm talking about repetitive stress injuries, the elephant in the room that these journalists evidently decided to ignore. And why? It's a serious problem. My wife recalls from her journalism days that half the copy editors in the news room wore wrist braces at least some of the time. Think of your own experience and talk to your friends and colleagues: who has had hand or wrist pain from typing, vs. who has had a PC monitor fall on his head?

Lawyers Cash In



Chances are good that you've had a laptop or keyboard with a little disclaimer sticker on it: “WARNING: To reduce risk of serious injury to hands, wrists, or other joints, read Safety & Comfort Guide.” (Never mind that in this photo it’s on my stapler. I moved it.) This disclaimer stems from a 1996 lawsuit in which Digital Equipment Corporation was successfully sued and forced to pay $6 million in compensatory damages to three office workers who developed repetitive stress injuries from using DEC keyboards. Why was DEC singled out, when any keyboard could have caused this injury? It’s because they'd trained their own employees on the dangers of typing but hadn’t warned their customers.

The lawyers on that case must have been pretty slick. But imagine that you're a lawyer pursuing similar litigation and could show the following:
  • This equipment corporation was using an ancient keyboard layout specifically designed to be inefficient, to keep 19th century typewriters from jamming;
  • Instead of thinking of customer safety, a silly marketing gimmick—the ability to spell the word “typewriter” using only the top row of keys—was a design goal of this layout;
  • The corporation declined to offer a more efficient keyboard layout even though one has been available since 1936.
All of these statements are true, but in practice would be of no use to a lawyer because no mainstream keyboard manufacturer offers the more efficient layout—that is, no single corporation is uniquely guilty here. Alas, the so-called “QWERTY” keyboard you’re sitting in front of right now is the very 19th century design I’ve described above. A lawsuit founded on the QWERTY’s lack of efficiency would have to be filed against an entire industry—no, actually, against an entire society—for tolerating this state of affairs. In short, I believe that QWERTY, not DEC or any corporation, is the real villain with regard to keyboard-related repetitive stress injuries.

Battling a legacy of lameness

People have been historically dense about typing efficiency. The typewriter wasn't originally introduced as a way to write faster—just to be more legible. In fact, the widespread practice of touch-typing came more than fifty years after the typewriter was invented. For the first fifty years, most people just hunt-n-pecked!

I first became aware of the existence of a superior keyboard layout in the mid-eighties when I saw the Guinness Book of World Records certified fastest typist go up against the “Late Night With David Letterman” secretary in head-to-head competition. Letterman turned the race into a farce by breaking whatever promise he’d made to the record-holder to provide her with the special typewriter she needed. Instead, just for laughs, he set her up on a QWERTY machine, and declared her a fraud on the basis of her having typed pure gibberish. The poor woman was almost in tears. I could tell something was up, because typing gibberish is actually no faster than typing actual text. (Try it!) I realized this record-holder must use a more efficient layout, and I was intrigued.

Somewhere along the line, I learned more about this alternate keyboard layout. It’s called Dvorak, was created in 1936, and is named after its creator, August Dvorak. He researched typing efficiency for years in developing his design, placing the most commonly typed letters in the most convenient places, and putting all the vowels on the home row under the fingers of the left hand, so that the two hands alternate as much as possible. The fact is, his layout wouldn’t need to be the most efficient possible to be a big improvement over QWERTY. If he’d done a merely passable job with his layout, it would be a big step up given the design intent of the original. Just knowing this more efficient layout existed, I decided I had to try it. Here’s what the keyboard looks like:


A False Start

In January through June of 1995 I tried to learn Dvorak and failed. Windows PC software didn’t support Dvorak yet so to use it I had to remap every key on my keyboard. I already had a macro-programmable keyboard that made this possible, and by sheer luck I had the same keyboard at work. But I had problems. For one thing, the keyboard would occasionally lose its mind, turning my typing into gibberish and requiring me to painstakingly remap the keys all over again. But more importantly, my approach to learning Dvorak, I think, was all wrong. I figured I’d learn it at home and then start using it at work. But all the QWERTY typing I did at work was undermining anything I could learn at home. Because touch-typing is an automated process—your fingers do what they’re supposed to without explicit, conscious instructions from the brain—you have to unlearn QWERTY before you can learn Dvorak. At least, I did.

My personal philosophy is that when you know what you ought to be doing, you should just do it—excuses are no good for the soul. In keeping with that philosophy, and because I was using a PC eight to ten hours a day and my hands and wrists ached every evening, I decided I had to get past QWERTY. I finally tackled the Dvorak project again in March of 2002, shortly after reading an inspirational article. By this time, Windows software had Dvorak support built in (Apple had supported it since the ‘80s). I decided to take the cold-turkey approach, converting my home and work PCs to Dvorak and vowing never to go back.

Success

Right after making this resolution I was tasked at work with writing a proposal that ran a couple hundred pages. I was of course tempted to postpone my Dvorak project, but decided any excuse would just breed others, and made up my mind to forge ahead on the alternative keyboard layout. I’ve heard that learning happens faster under duress, and perhaps it’s true: by the time that proposal was done, I was touch-typing on Dvorak. I wasn’t going very fast (maybe thirty or forty words per minute), but my speed has steadily increased ever since. According to an online speed test I just conducted, I’m typing on Dvorak at ninety words per minute, about five words per minute faster than I had after twenty years on QWERTY. More importantly: despite typing more now than I ever have in my life, I never have hand or wrist pain anymore. Never, ever. Dvorak is the real deal.

Why you will reject this innovation

I would be very surprised if anybody switched to Dvorak on the basis of this blog. First of all, I’m not so sure anybody reads this. Second, you are all weak. Okay, I’m kidding. Actually, I think there are a good many reasons why improvements in product efficiency fail to gain widespread adoption. In this essay I will explore some of these reasons, and their applicability to the Dvorak case in particular.

There are probably endless reasons why useful innovations fail, but I’m going to focus on four of them. Here are what I see as the most common low-adoption pitfalls:
  1. What’s in it for me? The precise gain in efficiency is difficult to predict or understand.
  2. The mixed bag. The gain in efficiency is offset by other decreases in overall functionality.
  3. Aesthetics & cultural signals. The revamped product is inferior aesthetically and/or is nerdy.
  4. Don’t go changing. The revamped product requires up-front work and/or behavioral change that scares people off.
(Yes, this is going to be a long article. I’m sure you have something better to do, like watching “American Idol.” Go right ahead. But if you do, all you’ll have to discuss tomorrow at the water cooler is “American Idol,” and all you’ll have to discuss during your retirement is how much your arthritis sucks. And I’ll be the smug guy in the nursing home making fun of your wrist braces.)

Low-Adoption Pitfall #1: What’s in it for me?

It’s hard to commit to a product innovation solely on the grounds of its supposed increase in efficiency. When the product’s performance is—or seems to be—measurable in simple numbers, people are easily persuaded to upgrade. If a 1 GHz PC processor is fast, why, 2 GHz must be twice as fast! If a 7.2 megapixel camera takes sharp pictures, 10 megapixels must be even sharper! Whether or not the numbers translate into actual benefit, people don’t need much convincing. But it’s harder when numbers aren’t involved, or when you don’t have a sense for how your existing product stacks up to the latest and greatest.

Take, for example, your refrigerator. According to the website p2pays.org, this appliance uses a sixth of all the electricity in your home. But how much of that electricity could you save by upgrading? Well, that depends on how old your existing refrigerator is, and what you’re replacing it with. P2pays.org tells me a new fridge could save me up to $94 a year—but what assumptions are they making about what I’m starting with? Stopglobalwarming.org suggests that a new fridge will save me $60 a year, but they don’t provide their assumptions either. A Cornell University study says a ten-year-old fridge uses twice as much electricity as a new Energy Star fridge, but they go on to list actual annual consumption numbers (690 kWh and 436 kWh, respectively) that belie their “twice as much” statement. After awhile my eyes glaze over, and through sheer inertia I stick with my old fridge.

Low-Adoption Pitfall #2: The mixed bag

Many innovative products solve one efficiency problem while introducing others. Perhaps the new problems struck the innovator as mere idiosyncrasies, or perhaps the new problems don’t bother everybody. I can think of several examples of this mixed bag scenario.

First, I give you the recumbent bicycle. I believe it is well established that these bikes are more aerodynamic than a traditional bike, especially if they have a full or partial fairing. The world record for human powered travel was set on a recumbent, and probably the next ninety-nine runners-up were recumbents as well. But the bikes are heavy, so they’re slower going uphill, and they have too long a wheelbase to corner quickly, and they’re not very stable on downhills. They’re also less visible to cars, and hard to mount on a car rack or take on the train. For most riders, the increase in efficiency on flat, straight roads isn’t enough to overcome the disadvantages.

I can think of other examples of mixed-bag innovations: the digital car speedometer (harder to read at a glance), the electric can opener (loud, takes up counter space), the car alarm (makes everybody in the vicinity rightfully wish for your death), and the Kindle (which, being an electronic device, denies its user the escape from electronic devices that is one of the great pleasures of books). I suppose these aren’t all classic cases of failed products, but they’re not runaway successes either.

Low-Adoption Pitfall #3: Aesthetics & cultural signals

We need look no further than the ongoing popularity of the stiletto heel to remind ourselves that efficiency and performance aren’t everything. Aesthetics can—and should—be a consideration when we decide whether or not to adopt a product innovation. Take the case of digital watches: they’re loaded with features—some of them actually useful, like an alarm or a backlight—but I for one am glad they haven’t replaced analog watches. That doesn’t mean I fault you if you prefer digital; I’m just glad I still have a choice. We humans have to look at consumer products all the time; they might as well look good.

In some cases I think we as a society have a real responsibility to reject innovations on aesthetic grounds even if the increase in efficiency is obvious. I give you the modern plastic squeeze bottle of ketchup. Just look at it, compared to its vastly superior ancestor:


It’s almost as though the newer bottle is designed to reflect the physique of the modern American: short, squat, and fat. And of course the new bottle dispenses the product much faster than the old one; instead of the subtle air-bubble-sliding technique, the consumer can now force the ketchup out as fast as he wants. The squeeze bottle even makes a fitting flatulent sound as it spews. Revolting. And yet the American consumer seems to have rolled over on this one: I haven’t seen a proper glass ketchup bottle in a store in years. (Here’s a tip: next time you’re in a decent restaurant or diner, all you have to do is use up all the ketchup in the glass bottle at your table. The label says “Not for resale” and “Do not refill.” On these grounds you can ask to be given the empty bottle, with reasonable expectation of success.)

Along with purely aesthetic considerations, we shouldn’t ignore the cultural signals that our product choices can send. Returning to the case of the recumbent bicyclist, it’s pretty obvious that—whether his turtle-on-its-back position strikes you as inelegant or not—he’s clearly an iconoclast, his odd choice of steed a tacit rebuke to the rest of us. To put it bluntly, the nerd factor of a recumbent is very high. Other high-nerd-factor products include the pocket protector (not in itself particularly ugly, at least no more so than a modern ketchup bottle) and its modern-day equivalent, the smartphone belt holster. I’m not real fond of Bluetooth earpieces, either.

Low-Adoption Pitfall #4: Don’t go changing

Consumer product upgrades are especially compelling when the only thing the consumer has to do is pay his money, following which the increase in performance is automatic. If you replace your 25-pound steel bike with a 16-pound carbon fiber one, you’re going to go faster as soon as you start pedaling. But upgrades are a harder sell when the better product requires work on the part of the consumer. I’m cheered by the huge success of compact fluorescent light bulbs, but at the same time I’m dumbfounded that so few of their adopters seem to have ever tried dimmer switches, which have been around for decades. The only explanation I can think of is that people are too intimidated by the prospect of working around live wires to install the dimmer switches, while anybody can change a light bulb.

The hardest sell of all is a product that requires the user to learn a new technique. The most dramatic example that comes to my mind is the 1989 Tour de France stage race, when Laurent Fignon lost the three-week, 2,000-mile race to Greg LeMond by only eight seconds, on a drastically less efficient bike. The big difference was the aerodynamic handlebar LeMond used. With it, he beat Fignon in all three time trials (races against the clock, where the rider must fight the wind alone).

The aero handlebar was nothing new, really; triathletes in the U.S. had been using them for years by this point, and the American 7-Eleven team had been using them in Europe all season. But Fignon, along with the other tradition-bound European racers, didn’t seem interested in this technology, even after losing time to LeMond in the time trials. Given that in head-to-head stages Fignon seemed the stronger rider, shouldn’t he have started to suspect that his traditional bike was holding him back, and switched to the aero bars for the final time trial? Perhaps Fignon just didn’t believe he could adapt quickly enough to the new position these bars required. (Needless to say, after LeMond’s triumph, the entire European peloton adopted the aero handlebar and with a few notable exceptions they’re ubiquitous in time trials to this day.)

Now that I’ve outlined some of the classic reasons an innovation can fail to gain widespread acceptance, I’ll evaluate the Dvorak keyboard with these reasons in mind.

Dvorak and Low-Adoption Pitfall #1: What’s in it for me?

Let’s face it, Dvorak has a pretty big image problem. For one thing, the vast majority of typists have never even heard of it. And those who have probably don’t know exactly what it is. The name doesn’t help; I’ve been typing on this thing for years and don’t even know how to pronounce “Dvorak.” Is it “De-VOR-ack,” or “De-VOR-zhock?” The word looks foreign and therefore suspicious, and it also summons the idea of wussy classical music. (I actually like the music of Dvorák the Czech composer, but then I’m a bit square to begin with.)

A further challenge: as status-quo-challenging innovations go, the Dvorak keyboard layout doesn’t have any heroes behind it. Its creator was an obscure educator who stayed obscure. To return to the handlebar anecdote: Greg LeMond was an American hero who won, three times, one of the biggest sporting events in history. Surely he more than anyone is responsible for the aero handlebar’s widespread success. Going back a bit further, let’s look at David and Goliath. When any of the rest of us would have cowered in fear, David went right out and fricking slew the evil giant. What could be more heroic than that? The closest thing Dvorak has to a hero is the Guinness Book of World Records fastest typist—whose moment in the sun, if you’ll recall, was spoiled by David Letterman. (Even if it hadn’t been, how heroic is typing fast, anyway?)

At least now you’re aware that the Dvorak alternative exists. And given that you probably type a whole lot, every day, you might even care that there’s a more efficient option to what you have today. At the same time, you’re entitled to be skeptical about the actual gains in efficiency this newer layout can offer you. Sure, it worked for Dana, but who the hell is he?

The theory behind Dvorak

Before I get to the body of evidence for (and, oddly, against) Dvorak, let’s take a moment to examine the gist of its design differences over QWERTY. There are two main principles at work with Dvorak.

For one, the vowel keys are located beneath the fingers of the left hand, with common consonants under the right hand, to maximize the extent to which the hands alternate when typing a word. (English words tend to alternate vowels with consonants.) Try typing (on your QWERTY keyboard) the following letter sequence: sf sf sf sf. Now try fl fl fl fl. Which was easier? Which was faster? Obviously fl fl fl fl. See? Alternating hands helps. (Of course, alternating consonants like this doesn’t actually help to type real words, so what you’re really seeing here is the lack of this trait on QWERTY.)

The second main principle in Dvorak is that the most commonly used letters in the English are located on the home row, right under where your fingers naturally rest. This decreases the amount of reaching you have to do with your fingers. Try typing a few letters on the home row: sldk sldk sldk. That’s really easy. Now try typing a few letters that aren’t: enoc enoc enoc. That’s a bit harder, isn’t it? To get to the upper row you have to straighten your finger out a bit and reach. To get to the lower row you have to curl your finger toward your palm a bit, which is even harder. In the process you occasionally miss the key you’re reaching or curling toward: a typo.

And yet, these trickier keys to reach—e, n, o, and c—are very common letters you shouldn’t have to reach for. With Dvorak, three of these letters are on the home row. Another example: type “the”—the most common word in the English language—on QWERTY. You start with a long diagonal reach to the upper row with the left index finger, a sideways reach with the right index, and a reach to the upper row with the left middle. Not very efficient. With Dvorak, all three letters are right under your fingertips on the home row. Much more efficient.

The key to QWERTY’s inefficiency

Out of top twenty words in the English language, only two can be typed on QWERTY without leaving the home row, whereas sixteen can be typed on the Dvorak home row. Of the ten most common letters in English, only three of them are on the QWERTY home row, whereas nine of them are on the Dvorak home row. (The ninth most popular letter, “r,” was evidently sacrificed to the cause of getting all the vowels on the Dvorak home row.)

Just stop for a second and stare at your QWERTY keyboard. It’s a mess! The indispensable vowels are scattered across the board, while the seldom-used semicolon gets a prominent spot on the home row. How did they get it so wrong? Simple. It’s a result of the state of the industry when QWERTY was created: the concept of the home row simply didn’t exist back then, because the QWERTY creators never envisioned that people could touch-type. Typing in the year 1872 was a two-finger operation where every keystroke was a reach. In 2009, doesn’t it just make more sense to type on a layout that was specifically designed for efficient touch-typing?

The buzz about Dvorak

What a silly section heading. There’s very little mainstream buzz about Dvorak. We have two very old studies from the 1930s and ‘40s establishing its merit: the original one funded by Dvorak himself around the time he created his layout, and another by the US Navy. Alas, neither report is to be easily found on the Internet. There’s another study, from the 1950s, by the General Services Administration (a government office) that followed the efforts of ten would-be Dvorak converts and concluded that the government should not bother retraining its employees. (I can’t find this online either.) Then there are individuals’ websites, like this one, set up by happy Dvorak users to promote the layout, simply for the benefit of society.

The only modern, formal, published paper you’ll find was written by a pair of economists, Stan Liebowitz and Stephen E. Margolis. Their names pop up every time you research the Dvorak keyboard. I read their paper, and was so annoyed by it that my discussion of its many failings has spawned its own essay: The Case Against Margolis & Liebowitz.”

In the meantime, suffice to say there isn’t much in the way of easily accessible, mainstream, authoritative testimony as to the actual efficiency of the Dvorak keyboard layout. There are testimonies, though, if you hunt for them. I found many among the comments from converts at this New York Times web page. Whether or not you accept up front that Dvorak really is more efficient, we’ve still got three more Low-Adoption Pitfalls to consider, starting with…

Dvorak and Low-Adoption Pitfall #2: The mixed bag

Assuming that Dvorak really is more efficient than QWERTY, we still have to decide if adopting something non-standard is actually worth the hassle. Clearly, there are some downsides to switching to a format that virtually nobody else uses. You may well be intimidated by the process of setting it up to begin with. Beyond that, we all type some of the time on other people’s computers, and there are things we do with a keyboard besides typing simple text, and then there’s the matter of our own computer having all the keys mislabeled.

Setting it up

Converting a Windows 2000 or XP PC to Dvorak the first time is admittedly difficult if you try to figure it out on your own. (With these operating systems Microsoft put keyboard layout options under “Regional and Language Options” instead of “Keyboard.”) But if you follow the explicit directions widely available on the web, this is a five-minute exercise and then you’re done. After that, switching between layouts is easy, with an icon Windows puts on your system tray. And if you employ user profiles on your home PC, you can pre-configure each one with its own layout, so you never need to change anything. (I’m not going to talk about Macs here, other than to predict that their Dvorak support is better than that of Windows, given Apple’s overall OS quality, the fact that they’ve supported Dvorak since the 1980s, and the fact that Steve Wozniak himself types on Dvorak.)

With Vista, Microsoft did a much better job on Dvorak support. (In fact, it’s about the only thing I can see that Vista does particularly well.) A keyboard icon is located, by default, on the taskbar (and a similar one is on the opening login screen):

Switching around

Okay, but what about typing on other computers? Well, you’ll still be able to do it, using QWERTY—the keys are labeled, after all. I can type at about 35-40 words per minute on the old layout, though I never, ever practice on it. (What slows me down is having to look at the keys.) I have to say, during my QWERTY test just now, I had to marvel at how my fingers were going all over the board, constantly reaching for far-flung keys. The difference is not subtle. It’s like the difference between bagging groceries and stocking shelves.
It’s actually very odd: when I first mastered Dvorak, I couldn’t type QWERTY at all, but it’s gradually coming back. Typing on my Blackberry—which I do no more slowly than anybody else—seems to have sharpened my QWERTY skills. Many Dvorak typists can switch back and forth with ease, and I’m gaining confidence that I, too, could develop this ability if I devoted any effort to it.
So ask yourself: how often do you really type in Internet cafés or on other people’s computers? Is protecting your speed on these other computers really worth typing more slowly, with more pain, all the time?

Beyond text

In trolling the web for Dvorak lore I have occasionally come across the argument that not all typing is of text, and that relearning control-character sequences represents a significant hurdle. To me, this sounds suspiciously like something an intransigent QWERTY typist would say. All I can tell you is, it’s no big deal. After all, aren’t those control-character sequences pretty ambiguous to begin with? I’ll grant you that Ctrl-C for “copy” makes sense, but why is Ctrl-V such an obvious assignment for “paste”?

Besides, I still type these same letters—I just do it with different keys, no differently than with text. I configure Cisco routers, using the non-GUI command-line interface, with no problem on Dvorak. (And when I had to take a PC-based Cisco certification exam on a QWERTY, that was no problem either. Well, actually, it was a big problem, but not because of the keyboard.)

What about PC gaming? Okay, now I’ll freely confess I’m out of my depth as I don’t ever play computer games. But, this being a full-service blog, I’ve done a cursory Google search and unearthed a small society of Dvorak gamers. Enjoy!

The little issue of labels

You may be wondering if it bothers me to type on a keyboard on which virtually all of the keys are mislabeled. Actually, this bothers me not one iota. In fact, I purchased my first Dvorak keyboard (for $20) just a few weeks ago, and I never use it. (It’s for my daughter, Alexa, so she never has to waste a single moment hunt-and-pecking on the QWERTY keyboard that she is forbidden to ever use.) The fact is, if you need to look at the keys, you really don’t know how to type! That’s actually good news for you, because it will make learning Dvorak that much easier—you have nothing to unlearn.

I first learned to type on the IBM Selectric typewriter, a gorgeous piece of American engineering. The only thing I didn’t like was that the keys were blank: my typing class in junior high used specially made typewriters. Having blank keys was the only way to ensure that students didn’t look at the keys while trying to learn. It is, I believe, well established that you cannot learn to touch type if you look at the keys. So don’t waste your money on a Dvorak-labeled keyboard or stickers for your laptop keys—you’re better off learning on a QWERTY-labeled board. (This is handy for when somebody borrows your computer, too.)

Dvorak and Low-Adoption Pitfall #3: Aesthetics & cultural signals

Five thousand words into this blog, it’s tempting to dispatch this low-adoption pitfall quickly with the simple argument that there is no aesthetic difference between QWERTY and Dvorak because you haven’t replaced any hardware. The act of typing might look slightly different to somebody who’s paying very close attention, but that’s about it. I do get comments about the sound of my typing, when I forget to mute my phone during conference calls. These comments are always some version of, “That must be Dana. Nobody else can type that fast.” An aesthetic demerit? I think not.

But of course there’s the cultural signaling issue to deal with. Anytime you reject the status quo in favor of something you feel is superior, you run the risk of seeming elitist. (Funny, though, how this doesn’t seem to worry people in the case of expensive cars and designer clothes.) Among those who know I use Dvorak, the responses have been benign, similar to people’s response to my good grammar and early morning workout regimen. That is, it’s treated like a generally harmless idiosyncrasy; I’m a nerd, and probably elitist, and probably no fun to have a beer with, but nothing to get up in arms about. (Actually, I like to think I’m a fine beer-drinking companion.)

It’s probably impossible to keep your Dvorak preference a secret from everybody, though some version of “don’t ask, don’t tell” would probably work just fine pretty well if you’re concerned about it. I wouldn’t put my Dvorak skill on my résumé, and I don’t generally talk about it, and that should be enough. Besides, as I’ll cover in this next section, I wouldn’t actually ask you, the reader, to adopt the Dvorak layout. This blog isn’t about you—it’s about your kids.

Low-Adoption Pitfall #4: Don’t go changing

Perhaps the strongest anti-Dvorak arguments pertain to the difficulty of retraining. Certainly this is the focus of the government General Services Administration study that provided much of the fodder for Margolis and Liebowitz’s polemic (click here for details). I’ll freely confess, unlearning QWERTY and learning Dvorak was really, really hard. There were moments when I’d get brain-freeze and for a split second be unable to type on either layout. It was in the same league, effort-wise, as learning not to cuss in front of my kids.

But the difficulty of retraining is really beside the point. Society has labored under the yoke of QWERTY for 137 years—another thirty or fifty years of it isn’t the end of the world. What’s important is to stop this hemorrhaging of efficiency for the next generation, by teaching the Dvorak layout to new typists. This isn’t a problem of aptitude—it’s attitude. If we can all agree that QWERTY is lame, can’t we take the next step and abandon the self-centered, narrow-minded idea that if it’s good enough for us, it’s good enough for our kids?

Generation gap

This wouldn’t be the first time that a better way of doing things was adopted first by the younger generation. Consider the long-awaited “paperless office.” After an unpromising start, this dream is finally beginning to approach reality. I know this because my colleagues, every time they want to print, are futzing around trying to get their PCs to talk to the LAN printer because they haven’t printing anything in so long. Myself, I inherited a local printer from a laid-off colleague about a decade ago and I’m still on that original print cartridge. The man who hired me had a credenza and a large file cabinet to store his papers; though I inherited these when he left, I’ve barely added to them in the last five years. The amount of work-related paper I’ve accumulated is a stack about two inches thick.

And yet, printing does go on in our office—a fair amount of it. Our printer and copier are in a room adjoining the kitchen, and every time I head in to fill my water or nuke my lunch, somebody is in there collecting his or her print job. And every time, it’s someone with gray hair. I’ve never seen a person my age in there printing—they’re at their desks, writing e-mail, many of them featuring footers like this one:


I’m not kidding: you could determine the age spread of our office simply by analyzing the LAN print queue. The older folk actually print e-mails. (I’m not sure I’ve ever printed an e-mail in my life.) Go up another ten or twenty years, and you’ve got people who print everything, as though the PC is just a really, really fancy typewriter. Continue further up the age scale and you find the people who don’t use computers at all. I’ll never forget the day I showed my first laptop to my grandpa, who was then ninety-six years old. “But it doesn’t do anything!” he cried, picking it up and dropping it on the table. My heart almost stopped, as I feared for my laptop’s hard drive.

Learning curve

As I approach my fortieth birthday this becomes painful to admit, but I’m heading, along with my peers, toward that state of brain ossification that makes any learning difficult. Asking us to unlearn QWERTY, after twenty or more years of using it, is daunting indeed. Naturally, we project this resistance to change onto our children—but we shouldn’t. For them to pick up Dvorak would be—forgive me—child’s play.

Consider this anecdote. The other day, I was chatting online with my brother Bryan, and when I stepped away, my daughter Alexa saw an opportunity to say hi to her uncle. She grasps that because my Windows profile is set to Dvorak, the labels on the keyboard are useless. Not yet able to touch type, she fetched her Dvorak-labeled keyboard, plugged it in, and typed away. She’s seven, and has not yet had a lesson about USB ports, but—being a kid—she evidently figured, “How hard could it be?” I have no concerns about the next generation’s ability to learn Dvorak, no matter how daunting it may seem to us. All we have to do is get out of their way—which means not exposing them to QWERTY.

My autocratic fantasy

I have this game I like to play: If I Were an Autocrat. For example, if I were an autocrat, the following would be illegal: SUVs, bottled water, car alarms, and teaching QWERTY to kids. (I realize that the first three items on this list, if not all four, may have assured the alienation of some of my blog readership. I guess I can live with this. Perhaps in my autocracy you’d be forced to continue reading anyway, and you’d be tested on the material.)

Being a benign autocrat, I would gradually phase in the anti-QWERTY laws. The first phase would require that Dvorak be mentioned, and operating system support guaranteed, to all students. Once the parenting public had been thus exposed to the technology, I would make Dvorak the standard, with a cumbersome opt-out policy whereby a parent would have to apply to a tribunal to explain why QWERTY made sense for his or her child. Eventually it would be a felony to teach QWERTY to kids.

Okay, I’m kidding!

Obviously, that wouldn’t be the most effective way to phase in Dvorak, nor is there much chance of me becoming an autocrat anytime soon. Actually, the best way to jump-start this evolution would be to provide the awareness and basic tools, and let the kids adopt it on their own as a way to simultaneously rebel and out-type their old, lame parents. I’ve even pondered the creation of video games that slyly promote Dvorak. Not lousy games like the Typing Tutor “typing lobster” game I used once, but cool, modern games. You could have one where the player’s weapon requires fast typing of a common word. For example, if you typed the word “the” too slowly, your avatar’s leg would kick back at the knee like when a girl throws a ball; if you typed it in milliseconds your avatar would thrust his hand beneath the rib cage of the opponent and rip out his heart. The kids would eventually figure out that switching to Dvorak is the way to win. (Yes, this is just a joke. Ultra-violent video games are actually on my autocrat-banned list.)

Am I high on drugs?

It might seem highly improbable that this technological problem has a political or legal solution. But things can change fast. I’m still pinching myself over the relatively recent anti-smoking laws. The day I heard they’d be outlawing smoking in San Francisco, I was sure it was a hoax. When the law was passed, and it gradually became apparent that it would actually be enforced, I really felt this was too good to be true. And then other cities across America started following suit, and even Dublin now bans smoking in pubs. Who could have predicted this?

Other laws show how quickly a societal improvement can be adopted. Ever the vanguard, San Francisco has now introduced a compulsory composting law. I’m in awe, frankly, of how seriously this city takes its civic duty. This new law sounds like the kind of thing I might introduce if I had political clout and/or Gavin Newsom's hair.

Is my lifestyle improved by composting? Not directly. All it personally means to me is emptying the kitchen compost bin into the big green waste bin every week, scooping the eerily warm, extraordinarily slimy, chunky, utterly disgusting compost goo off the bottom of the bin with my fingers, fighting the gag reflex and pondering the condensation on the lid of the bin. If the citizens of San Francisco can tolerate a law requiring them to do this, purely out of civic duty, surely we can tolerate a law requiring our kids to be offered Dvorak alongside QWERTY.

Remember the lawyers!

It’s not hard to see how a Dvorak law might take shape. Evidence of a public health issue gives prospective laws some serious teeth. Consider the DEC case: they were forced to pay $6 million in damages because their keyboard was shown to have caused repetitive stress injuries. If I had to guess at why Microsoft Windows supports Dvorak, I’d say it’s protection against lawsuits. After all, why else would they take the trouble, when the number of Dvorak users—though unknown—is likely very small?

Alas, we’re stuck in a chicken-and-egg situation. Before the lawyers get excited about punishing the lack of Dvorak adoption, somebody first needs to establish, in large, well-run scientific tests, the actual efficiency advantage of the Dvorak layout. Once this is established, this keyboard may finally start to gain some traction, and its merit will become widely acknowledged.

National Safety Month

Remember, June is national safety month. So in its honor, why not do something right now to mitigate the risk of a repetitive stress injury? I’m not asking you to switch to Dvorak (though you should). Here are some easy, free things you could do to promote this salutary product:
  • If you’re a parent, consider having your child learn to type on Dvorak
  • Talk to your child’s teacher about offering Dvorak to interested students
  • Read my companion blog post, “The Case Against Margolis & Liebowitz” to gain an appreciation for how a couple of ill-informed economists have helped to cement the ongoing use of QWERTY
  • If you’ve only skimmed this essay, bookmark it for later
If you’ve actually made it to the end of this blog post, congratulations: you have real staying power. In fact, you’re a perfect candidate for actually converting to the Dvorak layout! E-mail me at feedback@albertnet.us if you’d like any tips on how to do this. If not, I’d still like to hear from you. Send an e-mail, post a comment below, or at least take a second to click one of the “reaction” buttons. I’d really like to know if anybody read the whole thing. (The reaction buttons are at the very bottom of this post, past the “useful links.”)

Some useful links

--~--~--~--~--~--~--~---~--
For a complete index of albertnet posts, click here.