For those who need to understand how the Web is changing the way businesses work
Meta Data
| Issue:
April 10, 1998
Author/Editor: David Weinberger Central Meme: Hyperlinks subvert hierarchy Favorite Beatle: John. Duh. Current Personal Crisis: Resentment of multi-lingual Swiss is still not enough to motivate me to learn French Home page: http://www.hyperorg.com Contact information: Click here. |
ContentsEmotional computing: Q&A with Dr. Roz Picard of MIT, author of Affective Computing
|
National Public Radio's "All Things Considered" last week ran my commentary about whether all human experience fits into a petabyte of memory, and the mystery of attention. You can listen to it here if you have the RealAudio plug-in, which you can get here.
The KMWorld web site has started a new commentary feature. The inaugural article is by, well, me on how the Web is killing office documents by (eventually) replacing them not with web pages but with web sites.
|
The model we have of the mind is so screwed up, it's no wonder that there are wars, sports events and and a successful Bruce Willis. What a species!
We consider rationality as the mind's norm and everything else is a deviation. So, we've designed computers to be rational also. Further, we've assumed "rational" means "non-emotional," as if the two things are separable.
In this maelstrom of cockamamie, metaphor-based, self-misunderstanding comes Rosalind Picard's Affective Computing. Professor Picard of the MIT Media Lab makes the case for building computers that can recognize and respond to human emotion. It is a sober, thoughtful disquisition.
JOHO met with Roz -- but she'll always be Prof. Picard to me -- in her office in Cambridge.
JOHO: Thanks for taking the time to talk with me. This must be a very hectic time for you.
RP: This past year I had a baby, wrote a book, and designed and built a house. So now it seems sort of calm.
JOHO: I ran into a friend a couple of days ago who said her company needs software that can identify the emotional tone of incoming email so the angry emails can float to the top ...
RP: One of the students in my seminar worked on an email editor like this. The idea is that it could analyze the text content and highlight the sections in red that might be negative or angry, and highlight other emotions in other colors. Even before I send mail, the software could let me know if I'm coming across one way or another. If there's a whole lot of red I can ask myself if I really want to send this angry-sounding message. Then, maybe the software should sit on that mail for an hour or a day depending on how much red is in it and ask you if you really want to send it. It could potentially suggest ways to tone it down or point out the lines that are most likely to be misconstrued. The same analysis could analyze incoming mail -- the tone and who is sending it. If a very important customer sends me angry mail, I want that at the top of my stack -- just like if a group of people were there in person, you'd deal with the squeaky wheel first.
JOHO: Given the state of grammar checkers that seem to attempt something simpler than gauging emotional states -- and do a lousy job of it -- when do you think such software might be real?
RP: It depends on how much you want from it. People don't read emotions perfectly. Neither do computers, and they probably never will. But for certain patterns of responses, if a human can read the emotion with some degree of confidence, so will the computer be able to. I expect it will be a mixture of recognizing vocabulary and textural patterns of the speech. But I think you'll always be able to fool the computer if you know how the algorithm works.
JOHO: Just as actors can fool humans.
RP: But, we can get to the point where these tools can be useful, but not perfect. We have to re-phrase the question to "Is this do-able enough to be useful?" When it will be depends on how many people work on it, of course. This isn't yet like speech recognition with 100 companies working on it.
JOHO: Why aren't there 100 companies working on it?
RP: Engineers have traditionally treated affective information as noise. It really hasn't been thought of as important. But this seems to be changing. We just had two papers accepted by the International Conference on Acoustics, Speech, and Signal Processing, a very traditional group that I was not at all sure would embrace this new work. I think they at least recognize that to detect affect is going to be hard.
JOHO: But it can't just be the difficulty of the problem that's holding back development of such systems. Recognizing affect has to be easier than doing continuous speech recognition. Besides, engineers like hard problems.
RP: Yes, we like hard problems, but in science and engineering the emphasis is on logic and rational thinking. We think of emotion as that which impedes rational thinking. We think it's like Data [from Star Trek: The Next Generation] with his emotion chip turned off. But the fact is that he wouldn't be intelligent without emotion.
JOHO: Why?
RP: Antonio Damasio and many others have studied people who, because of brain damage, are largely deprived of affect. These people are in many ways like computers without emotions. They're not very flexible. They can make rational, rule-based decisions but that turns out to be only a very small percentage of the decisions we need to make day to day. In making an appointment, they fumble around past the point where it's embarrassing, but they don't feel embarrassed. It can take them hours to decide between two open times. Computers also don't feel embarrassed, or have the internal emotional signals that assist in these day-to-day decisions that cannot be decided on pure logic. They don't know how to weigh importance in a continually changing environment.
JOHO: What are the business reasons for caring about emotional computers?
RP: Recently the human-computer interface people have become aware that the way people interact with computers is the same as the way they interact with everything else: naturally and socially. Reeves and Nass at Stanford call this "the media equation," that we treat computers like people. If your computer does something that annoys you, you'll probably let it know. Office rules for people apply. It doesn't matter if the computer isn't going to change its behavior ...
JOHO: Is this true for all inanimate objects?
RP: No, it's primarily true of media -- computers and televisions, according to Reeves and Nass's dozens of experiments. For example, if the computer gives you a lesson and asks you how it did, you might say it did great. But if another computer asks how that first computer did, you'll say that it did pretty well -- same as with humans. We praise people more to their face than when we report on them to others.
JOHO: Why this human-like behavior towards computers?
RP: When we encounter a new medium, we default to how we react with humans.
JOHO: Not with sponges we don't. Something about computers gets us to default to our human interaction methods.
RP: It is curious; the effect holds even for computer science students who know how computers work. I'm not sure why. Certainly the public is influenced by science fiction portrayals of humanoid machines and even non-humanoid machines that have emotions, like HAL [in 2001]. HAL also propagated the intimidating idea that computers could be foolproof and incapable of error despite the fact that HAL made errors. There is even a belief among some computer scientists that computers will become so much smarter than us, that as Minsky puts it, "we'll be lucky if they keep us around as household pets."
JOHO: Doesn't this assume (falsely, in my view) that being smart has something to do with being rule-based -- as computers ultimately are?
RP: This issue seems to distinguish man from machines: Man can recognize when rules are appropriate or adapt them on the fly. And machines don't have that. We sometimes call it common sense. I think it has to do with being able to switch priorities -- and this role is played in humans by the emotional system. If we want to build this into computers we might look at how humans do it. Your emotions help you in part by prioritizing and organizing. I suspect the blowing-up disruptive type of emotion is a very small part of it. Emotion lets us know what rules are important and when they should be bent and ignored.
JOHO: Talking about this as emotion perhaps leads people to think of the disruptive emotions. But there's in fact a range of affective qualities that includes moods and attitudes. In fact, the German phenomenologists have long pointed out that what's distinctive of our basic awareness of the world is that we don't just observe or cogitate but *care* about our world and our situation in it. Caring seems to subtend all these affective qualities. Maybe that's a way to talk about the emotional aspect of thought without having to raise the spectre of only the disruptive emotions.
RP: There certainly seems to be no caring, and maybe no feeling of meaningfulness without emotional feelings. Most of the time our feelings bias our thinking without our awareness. They influence just about everything we do, whether or not that thing is "emotional." The latter is part of what is so surprising about emotion, that it is so important even for "non-emotional" functioning. This is really what has led me to re-think the design of computers, to consider that giving them emotion is not merely some kind of luxury. When you bring up caring, however, this raises another issue: can machines really "care" about people? More importantly, how will people feel about such "artificial caring?" These are issues we will face before long.
AltaVista lists 399,372 pages that have the word "online" in their URL. For example, the Software Support Association's URL is http://www.sspa-online.com/, apparently because we might not otherwise figure out that we're browsing to an online address.
By the way, the telephone number 1-800-TELEPHONE (1-800-835-3746) just rings and rings and rings...
Stuart Hillston responded to our abbreviations contest with
IWGTOSYAABRTWDI-ATGAFOEACTEWMIGF.
When I replied in some confusion, he responded:
I was going to send you an abbreviation but realised this would do instead - actually this gives a form of encryption and compresses the email which means it goes faster.
Stuart also informs me that he's completing the Quake map that models the UK Open Text office where he works so that he can spend happy nights blowing his manager into bloody virtual bits. (Try substituting "co-workers" for "manager" in that sentence and it becomes oddly disturbing.)
BTW, if you want to try out a Quake2 map that Stuart just finished (it's a "test" map, he says), you can download it here. Install the file in your quake2/baseq2/maps directory and from within Quake2, at the console type "map tallone".
Jeff "Woof" Kassel -- desperately trying to live down his college-years nickname which I alone perpetuate -- responds to my recent commentary on NPR's "All Things Considered" (foreshadowed, shall we say, in these pages a few issues ago) that maintained that human experience cannot be "captured" in a petabyte of data:
What does it mean if the whole of my life's experiences can fit on a zip drive?
Ah, an entire new branch of psycho-computronics opens up! Tell me, Herr Voof, exactly what is your compression algorithm? Und perhaps you should stop und smell the flowers, ja?, so your life will be enriched to the point where it won't fit in a zip code, much less on a zip drive. Next!
Joe Brown -- not a JOHO reader -- heard the broadcast also and, with a gleam of urgency in his fevered eyes, tracked me down to inform me of the following:
...you said (approximately) "I don't care if it's off by a million times either way." Well, you can't BE off by a million times either way--only upward. When going downward, no matter what (thing) you are measuring, if you use "times" then your unit is that thing, and ONE times less than a thing is NOTHING. "x times less" is NOT (as is frequently and erroneously thought) a substitute for 1/x; "times" does not indicate division, but multiplication, and "less" indicates subtraction.
And just in case you hadn't thought about it, "x times MORE" than anything is that thing PLUS x times itself (as opposed to, and one times more than, "x times AS MANY").
People make these mistakes in languages other than English, too, simply for lack of analysis.
Ok, Joe, I accept your math but differ on your explanation of the motivation of those of us who make this mistake in many languages. There are multiple reasons for this error, not just the lack of analysis. For example, there's also the insistence on using English (and many other languages) as a way to communicate ideas, the ability to separate technical from non-technical communications, and, perhaps most galling of all, the error of not giving a rat's ass.
Thanks for writing.
Kevin Osborne points us to two -- count 'em, two -- April Fool's jokes from Sun Microsystems: http://www.sun.com/980330/javaapi/ and http://biz.yahoo.com/bw/980401/sun_micros_1.html.
Chris RageBoy Locke takes issue with part of last issue's essay on how web sites will replace business documents. I wrote:
I am assuming that in the future, it will be as easy to post a
page to a Web site as it is to attach it to email or run off
photocopies.
RageBoy leaps upon this statement like a terrier chasing a rabbit:
Wrong. Much easier! From the text editor in which I'm writing this,
I just type Alt-P and the current, uh... document gets published on
EGR -- assuming, that is, that it's an HTML document, which this is
not. Yet. But wait five seconds...
(Reminds me of a joke: It was so hot I saw a dog chasing a rabbit and both were walking.)
I'm sure Chris would agree that his ability to publish a text document from his weenie text editor does not exactly mean this is a mainstream capability that will be exercised by your typical idiot business person still trying to figure out which end of the pencil to shove up his nose. But we're getting there.
Adina Levin writes:
'Course, there's a cultural literary classic written the way you describe websites....contributions by groups, not individual authorship
...topics that are records of conversations and arguments
...non-linear structure based on cross-references
It's called the Talmud. Those guys had to work hard to use that style before the web.
Excellent point. Now, did you hear the one about a priest, a minister and a hyperlinked rabbi standing on a jetway...
Ron (hey, Ron, what's your last name?) provides a haiku for a previous bogus contest that comments eloquently on the nature of documents:
Which you can either rag as a late entry to the haiku non-comp, or apply to a dying document.
the underlying verb
is Uncovered.
I couldn't agree more and I wish I had put it as well. In fact, I used to maintain (in the way that speakers use provocativeness as an excuse for thoughtlessness) that documents are verbs.
Then Ron quotes from my article:
Places -- sites -- are not "done." They are
where processes occur, not the finished
result of processes. (When a site like this is
"done" is precisely when it becomes
uninteresting ... the opposite of the
traditional document publishing model.)
and comments:
..or would that be precisely when a site becomes a document of what has gone
before...
for as passe as that process/site may have become to its original
participants, its enduring value to those who were not present during its
lifetime, is difficult to determine. The best value we can offer to those
people is accurate document(ary) of the precedings, anything less is
destined for oblivion.
It is lack of this information that drives the wheel of re-search.
I agree again, although I like to think that I'd have spelled "information" as "in-formation." (By the way, if you have too much canned food for anti-abortionists, do you have a proliferation of a pro-life ration?)
Phil Murray writes in response to my assertion that better than thinking of knowledge management as a type of mining, we should view it as a process of finding stuff in a junkyard that has value because it meets some particular need of ours:
Having been to junkyards to buy parts for a worn out Buick Electra
and several other monstrosities, I can assure you that the junkyard
is a useful place *only* if the junkyard dealer has organized access
to that junk in a meaningful way. The wandering is useful if you like
constructing art out of carburetors and fenders.
Of course it has to be categorized in some way. But categorization can also hide objects of value because the categories are based on someone's assessment of useful categories for the *current* known situation. Note: this is *not* an argument against categorization.
Phil replies to this:
Take a look at faceted knowledge organization -- often
*combined* with hypertext linking -- as a better way of finding
information without imposing rigidity on a knowledge resource. The
Library Science community has been focusing on this issue for
decades, but the lessons they have learned are routinely ignored.
As in many other cases, the truth often lies between the extremes.
Top-down organization and bottom-up "associative" organization are
complements, not mutually exclusive. The assertion that they are
mutually exclusive is usually a straw man.
"Knowledge" is *not* an emergent property of collectively constructed
knowledgebases unless the resource is modeled to enable such
emergence. The Web itself is proof of that. The Web is an information
junkyard in the worst sense of the phrase.
Those of us who tout associative organizations (= messy contexts) at the expense of top-down, hierarchical structures frequently do so because overstatement is the only way to begin to redress the imbalance. Look for a boring-but-obvious article on "The Politics of Screeching: Nuance as Conservativism" in the never-to-be-published "Ideas We Sort of Had" issue of JOHO.
Adina Levin is on the mailing list for the Pepper and Rogers report on 1-to-1 Marketing which she gives the "Irony-Free Award" for an article on how doctors are using personalized marketing techniques:
Targeted direct mail increases demand for plastic surgery.I can't wait for the dinner-time phone calls selling liposuction.
RageBoy, as usual, one ups my recommendation of an atomic clock synchronizer. And then he goes on to join the JOHO Checker Board. Here's his message in its astute entirety:
check out Dimension 4 at:http://www.accessone.com/~thinkman/dimension4/index.htm
notice that I'm quoted there (if you wait for the stupid Jayvee scroll). very good tool and it's free.
oh yeah, and:
...for the Web incognoscenti, "wrt" does not mean "wart" or "who reads this?" but "with regards to."no it doesn't! It means "with *regard* to"! Regards are something you send to some person in mail even when you don't mean it, as in:
Regards,
RageBoy
There's nothing I like better than being humiliated in public. Keep it up, RageBoy and I'm going to have to TP your web site. Then we'll see who's laughing. (Herb Kelleher, for one.)
Ross Knights -- lately seen practicing his Jackie Chan moves outside a theatre in Allston, MA -- forwards a very amusing promotional offer for a managed friendship service. Here's an excerpt from the beginning, which, if it pleases you, you can see in its risible entirety by clicking here:
What's Wrong with my Current Friends?
If you're like most people, you are receiving friendship services from a network of providers haphazardly patched together from your old neighborhoods, jobs, and schools. The result is often costly duplication, inefficiency, and conflict. Many of your current friends may not meet national standards, responding to your needs with inappropriate, outmoded, or even experimental acts of friendship.
Under Managed Friendship, your friendship needs are coordinated by your designated Best Friend, who will ensure the quality and goodness of fit of all your friendly relationships.
Welcome to Managed Friendship, a whole new way of thinking about friends and relationships. The Managed Friendship Plan (MFP) combines all the advantages of a traditional friendship network with important cost-saving features.
interesting
all the way through except for the dull parts.
Um, thanks, Stan. And welcome aboard. (You might find kindred spirit in Gerry Murray who has provided a "testimonial" that rings a similar note, but this time about my speaking ability.)
Finally, JOHO has been reviewed by the editors of an ezines database site (http://www.DOMINIS.com/Zines/). Here's the review:
A serious yet light-hearted publication that outlines the implications of the rapidly evolving cyber world. JOHO focuses on the changes that the Internet is bringing upon our lives, opportunities that are bestowed upon us by this 'horizontal' world of information, as well as practical ways to understand and take advantage of it all. Fun to read and easy to understand.
I was pleased with this until I got to the final three words. Why do reviewers feel they have to get something negative into the review? I mean, I work and work to make JOHO serious yet light-hearted and fundamentally a pile of incomprehensible twaddle, and some punk-assed reviewer has to say it's "easy to understand"? Now no one will take me seriously.
As Hegel said to Teilhard de Chardin, "Absolute spirit wills becoming out of nothingness, but doesn't know how to charge for it." Serious yet light-hearted, yes. Easy to understand? I think not.
Allaire recently came out with an upgrade to version 3.0 of Homesite (the HTML editor I use to create JOHO). The release notes list about 60 fixes in the new version, along with ten "known issues" (Allaire, where our motto is "If it's still broken, don't fix it").
Now, I didn't just fall off the software truck. I've worked in software companies and I know that all software has bugs, and a lot of the 60 fixed bugs were rare and inconsequential. Nevertheless, you're pushing the marketing envelope when you make your users do a 9MB download (not a patch) that comes with a 15K release notes file and you call the new release "3.0.1"
Using this logic, HomeSite 3.1 -- with an order of magnitude more changes than went into 3.0.1 -- will be the equivalent of what we would otherwise have called HomeSite 4.0. And what Allaire calls "4.0" will be the equivalent of killing off HomeSite and launching a new product.
I can't wait.
Keith Dawson's TBTF recently ran the results of some "brute force" research by Roger Gonzalez on a topic of interest to JOHO. We had previously reported that all the triplet combos were taken for the COM domain names (e.g., "AAA.com"). The situation is much worse than that, according to Gonzalez's research which looks at both COM and NET domain names:
- All 2-letter domains are taken- All 2-alphanumeric domains in .com are taken; 104 are left in .net
- Of the remaining 3-character domains, only 55 are words
Gonzalez lists all the taken combos, including ones with punctuation and embedded numerals. (Is it just me or does this seem like the sort of project undertaken by someone trying to keep some peculiarly nasty demons at bay?)
Computerworld (March 16) ran an article on ethics in the IS world with scenarios like:
Oooh scary. Vendors making contributions to charities! This happens all the time. In fact, in restaurants, one of my favorite moves to get a really good table is to sidle up to the maitre d' and say, "Excuse me, , we'd like a ringside table so we can discuss the $20 contribution we'll be making to ... say, what's your favorite charity, Pierre?"
How about a little whiff of reality here? What are some of the real life ethical dilemmas we face, thanks to the Web? For example:
| Situation | Ethical Question |
| You come across a granola-and-Jesus site obviously composed by dear, gentle, genuine people who hope their words will make the world a better place. You rip off a sickly-sweet graphic and paste it into your site in an obscene context, making sneering fun of it. | When you put in a caption crediting the original site, do you in fact link it to www.satanrules.com? |
| You're flaming some jerk on comp.lang.c++ for correcting a bug causing a memory leak in a listing you've posted. You've exhausted your stock of imprecations and have done a search through Deja News to see if anyone has come up with suitable emotive language. | Are you morally required to change the subject line of the thread from "Dynamic memory arrays" to "FRED SMITH IS A FUCKING ASSHOLE WHO CARNALLY ASSAULTS LITTLE BOYS"? |
| You've received email from an old friend from college who has tracked you down via 411. She's immensely successful now, although she's quite modest about it. You, on the other hand, have succeeded only by repeatedly re-defining success downwards. | When you send her a message full of lies about how important you are, do you switch ISPs so she won't have to see that you're an AOL subscriber? |
JOHO welcomes your entries. And remember, only at JOHO is it true that To Enter To Win.™
The following information was found trapped at the top of my washing machine when I ran some issues of the JOHO through it.
JOHO is a free, independent newsletter written and produced by David Weinberger. He denies responsibility for any errors or problems. If you write him with corrections or criticisms, it will probably turn out to have been your fault.
Subscription information,
or requests to be removed from the JOHO mailing list, should be sent to
[email protected]. There is no need for harshness or recriminations. Sometimes things just don't work out between people. Dr. Weinberger is in a delicate nervous state, but if you want to send positive comments to him, his email address is [email protected].
Dr. Weinberger is represented by a fiercely aggressive legal team who responds to any provocation with massive litigatory procedures. This notice constitutes fair warning.
Any email sent to the JOHO may be published in the JOHO and snarkily commented on unless the email explicitly states that it's not for publication.
Note to distributors:
If you are interested in reselling the popular Hyperlinked Organization brand
line of memorabilia, please contact our manager of JOHO Channels, Divad
Regrebniew. (The JOHO corn dog attack vehicle with lifelike action figures
is no longer available, but will return once we fix the eject button and
pending the outcome of the lawsuit.)
The Journal of the Hyperlinked Organization is a publication of Evident Marketing, Inc.
"The Hyperlinked Organization" is trademarked by Open Text Corp. JOHO gratefully acknowledges Open Text's kind permission to use this felicitous phrase. A listing of all Preemptive Trademarks™™ can be found at http://www.hyperorg.com/misc/trademarks.html.