the struggle broke out over recounting Florida's votes in the 2000 presidential
election, many people fell quickly into predictable camps based on their
political leanings. One side tended to emphasize disenfranchisement in
Palm Beach due to the design of the now-infamous "butterfly ballot." The
other side tended to emphasize the rules: If you can't follow the voting
rules in plain English, you don't deserve to vote, let alone get another
shot at it.
My take on the
whole brouhaha comes more from a statistical perspective. I read a mathematician's
persuasive op-ed piece in The New York Times explaining that the
Florida vote was by any fair measure a tie — the margin of error in the
polling methods exceeded the margin of victory, whichever victor one might
choose through post-election remedies. One hopes that Florida and the nation
will make good on the promise to improve badly outdated equipment and procedures,
so that in future elections, the margin of error will almost always amount
to less than the vote differential.
The other lesson
from all this is that design matters. Design experts examined the
butterfly ballot and pointed out how the opposing-page-with-linear-choices
orientation flew in the face of common sense — and the way we read from
left to right, one column at a time, in English printing practice.
When art meets
design, the question always arises of whether form follows function. But
good designers care a great deal about whether a given design — for a ballot,
or the radio controls in a Lexus RX-300, or a Web form — misleads users.
It isn't a question of whether some users are stupid. It's a matter of
looking at a large population of users, and if the design misleads, trying
to improve it to lower the aggregate error rate. Even elegant design fails
if it causes confusion. Design matters for all kinds of human-machine interfaces
in general and all computer interfaces. We're still trying to perfect the
perfect mouse or "pointing device."
So what's all this
got to do with online searching?
We live in a paradoxical
age — a time of disintermediation, yet also a time when a national newspaperreports
employment of reference librarians in the United States at an all-time
high. But the Web revolution has created a new generation of searchers,
both those searching Web content and using Web interfaces on the open free
Web to those using traditional commercial databases, often on the Web.
To the extent that database searching continues to move to the desktops
of end users, design can be as important as training in how effective those
users' searches are.
One of the strengths
professional online searchers bring to the party is knowing which sources
to search — where are the best haystacks? But another equally important
asset is knowing how to conduct the search — how best to find the needles?
As searching moves
to the desktops of end users, it's vitally important that all information
professionals involved understand both aspects. Depending on the nature
of the audience, the stakes can rise very high. An engineer doing a poor
online search may fail to find the optimal choice in materials and may
design a structure that eventually fails. An attorney doing a poor online
search may fail to find the key relevant precedent and lose a high-profile
case. A physician doing a poor online search may miss a detail about drug
interactions and lose a patient.
support professions such as engineering, law, or medicine should care a
lot about this. They should — and do — take the lead in educating their
users in the right sources. Most research physicians will automatically
choose MEDLINE over a random Internet site. Most lawyers (especially those
who graduated law school in the last 15 years) know to use LEXIS-NEXIS
But do these professional
end users know the quirks of these databases — the little navigational
subtleties that searcher librarians learn? The more powerful the tool,
the greater the need for training. An untrained person holding a BB gun
is nowhere near as dangerous as one wielding an Uzi. Or have the Web designers
for these traditional databases incorporated these subtleties to make learning
them unnecessary for end-user searchers? The impact of design failures
can be greatly magnified in the hands of an unskilled user.
It's not sufficient
to train the end user in database idiosyncrasies. We must also keep pressure
on the vendors to get the interface right. We can't all become Jacob Nielsens,
beating the drum of effective (but conservative) design. But we can let
Flash animation or poorly chosen pick list or inadequate relevancy ranking
causes end users to make mistakes that affect lives. And in the most important
cases, librarians should still do the searching.
Now let's consider
the example of a Web-based service whose commitment to simplicity, coupled
with powerful, effective technology, has won millions of users.
An Example: Google and Design
new search engine Google has captured the eyeballs of millions of searchers
worldwide. How did they do it? Very simple. In fact, the Google interface
has always been a model of simplicity. While in the last 2 years most of
the major search engines have been morphing into portals and generating
ever more bewildering arrays of choices for their users, Google has arisen
from the research work of two Stanford University students and taken the
Web world by storm.
this engine appeals equally well to the neophyte as well as the experienced
searchers. Google has built a Web search service that works best with the
kinds of queries real users actually tend to enter: simple, non-Boolean
queries that represent a quick rendering of what a user wants to find.
When you type "Bill
Clinton" into Google, it assumes that you want pages that contain both
"bill" and "clinton." The search isn't treated as a phrase search, but,
by assuming that all search terms are to be ANDed, Google avoids the problem
of mixing into the hit list all the pages about dollar bills, duck bills,
bills in Congress, and DeWitt Clinton, the New York governor who built
the Erie Canal. And Google doesn't make any assumptions other than that
— for instance, the search will match terms regardless of case. AltaVista,
by contrast, tracks on the case you enter. If a hapless user searches for
"BILL CLINTON," he or she will miss most matching pages.
simple but powerful approach is to give extremely high credence to the
number of links it finds to a given URL as it builds the hit list. Google
is not the only search engine to use popularity of a page as part of the
relevancy ranking algorithm, but it may apply this methodology most aggressively.
Do a search for a common word or phrase — "Searcher" or "White House" or
"Jay Leno" — and Google will much more likely find you a page endorsed
by the links of the global community of Web users.
of these two simple design elements — requiring all search terms to be
present and ranking by link count — is extraordinarily powerful in coping
with an ever-expanding Web. Google helps expert and novice searchers find
popular, or official, or commercial sites much more effectively than any
It's hard to overstate
Google's emphasis on simplicity compared to the models used by its competitors.
To see what I mean, you need merely visit the Google home page alongside
the AltaVista page. The other major portals are no less guilty; Disney
has spilt considerable blood trying to make Go.com a coherent search service,
subject directory, and portal to the vast proprietary content of a media
giant. (Sam Donaldson and Mickey Mouse on the same page; what a concept!)
Despite the devotion
to simplicity, the folks at Google have been quietly adding features to
the service. Originally introduced sans advertising, the service now displays
ads — albeit discreetly. Interestingly, unlike all their competitors, Google
displays the ads only when the search terms seem to directly relate to
the ad in question. For instance, a search for "Internet Archive" yielded
no advertisement. A search for "long distance," however, brought up, to
the right of the hit list, a list of "Sponsored Links" (so labeled) that
offered discount long-distance telephone service. Other search portals
show ads all the time, from the home page to every page showing hits. (Of
course, when they can do so, all the search portals tie the ads displayed
to the user's search terms.)
addition is the new "Google Web Directory." This subject catalog leverages
the taxonomy developed by the Open Directory project. Back when there was
some idealism left in the Netscape corporation, Open Directory was launched
as a sort of open source Internet cataloging effort. The utopian theory
behind the project intended to apply the same model that allows individuals
worldwide to contribute to the development of Linux: An "army of editors"
would scale the growth of the Open Directory to the growth of the Web.
Yahoo!'s editorial staff could not grow at the same pace. (Of course, the
editors are unpaid volunteers.)
The Open Directory
project allows free use of the database. Google has taken the Open Directory
categories and entries and applied its technologies, such as link-count
page ranking, to subject-oriented browsing. Google says their rendition
of the Open Directory has 1.5 million URLs in the catalog — or about 1/1000th
the number of URLs in the main Google search index.
Google touts its
directory as a unique combination of searching and category browsing, and
in some ways it is. For instance I searched among the "Newspapers" looking
for "Decatur" — several states that have towns called Decatur — and got
a good list of Decatur newspapers. But in the "Magazines" category I did
a search for "DV," the name of a magazine devoted to digital video, and
Google delivered a hit list with many false matches. A structured bibliographic
database, this isn't. Somewhat surprisingly, a search in the main Google
index for that short search term found the DV Magazine Web site as the
first hit. (I suspect the combination of link popularity as well as the
fact that the domain is dv.com boosted the right item to the top of the
The most recent
addition to Google's features really surprised me: Google has finally added
an advanced search feature. See http://www.google.com/advanced_search
to try it out. With a simple Web form, Google delivers to end users a great
deal of power:
In addition, the Google
Advanced Search offers the ability to search for pages similar to one a
user has encountered and to generate a list of pages that link to a given
URL. (Oddly, the link search appears seriously deficient: It failed to
find links to whitehouse.gov; AltaVista finds 170,000.)
The ability to search
for phrases, to do AND or OR combinations of search terms, and to exclude
pages with unwanted terms.
The ability to limit
searches to titles, URLs, or full page text.
Filtering by language
of the source page.
Filtering by Internet
Filtering by "SafeSearch,"
a SurfWatch-based filter for objectionable material.
With all these
new features, is Google doomed to end up like the other search portals,
with a Christmas tree of hyperlinks and search options confronting the
poor user? I don't think so. The folks at Google seem to have a strong
instinct for minimalism. The new features have been added very unobtrusively.
This, to me, is the secret of a powerful interface: Give most users the
features they need most readily and make other features available to those
who need them — without burdening the casual user or novice with interface
choices they can't understand.
Google offers an
undaunting service with a gentle learning curve. I suspect many high-end
Web designers scoff at its simplicity; the Google home page is so bereft
of glitz and fancy graphics, it's almost a throwback to 1995. (But its
graphic artists do produce some mighty cute seasonal logos; see http://www.google.com/holidaylogos.html.)
My guess is that librarians teaching search engines in Internet classes
are flocking to it, as are their patrons. And I think it's a safe bet that
Google, through relentless pursuit of simplicity, will continue to beat
the established search portals at their game. I don't expect any Palm Beach
Ballots from the folks at Google.
e-mail address is email@example.com.