NNSquad - Network Neutrality Squad
[ NNSquad ] Unintended Consequences: Google Instant and the "Suicide Search" (and a Solution)
Unintended Consequences: Google Instant and the
"Suicide Search" (and a Solution)
http://lauren.vortex.com/archive/000754.html
Greetings. I hadn't planned to be blogging anything for now regarding
the launch of "Google Instant" (GI) - ( http://bit.ly/c8xzFs [Official
Google Blog] ). It was all over the media yesterday, including a nice
spot on NBC Nightly News.
But a troubling note I received this morning has inspired this
posting -- the specifics in a moment.
So far I've primarily been viewing Google Instant (which isn't yet
available to all users in every country) as a superb engineering
achievement with the potential to significantly improve the user
search experience in various ways.
The scaling issues alone warrant hearty congratulations for all who
were involved in GI's development and deployment. In fact, most of my
discussions regarding GI yesterday in forums such as Google Buzz
( http://bit.ly/lauren-buzz ) were focused on how such a feature might
be implemented via browser toolbar searches (currently GI is tied to
Google Home Page searches, where the rapid display changes triggered
by entering a search don't conflict with an already displayed
non-Google page). My orientation shouldn't be too surprising -- keep
in mind that I'm basically a technical geek who morphed over the years
into heavy involvement in the policy arena.
I've seen some griping about GI in various venues -- almost all of
which strikes me as fairly silly. Even though Google asserts that
search rankings aren't affected by GI, there are concerns about how
the new search mechanism will affect SEO (Search Engine
Optimization) -- the business of trying to (often artificially)
boost search result rankings.
Personally, I promote the concept of creating quality "organic"
content and letting the chips fall where they may. While GI is now
the default mode where available, it's straightforward to disable --
both on the Google Home Page and via logged-in user preferences. As
usual with defaults, it's pretty safe to assume that the vast majority
of persons won't change this setting, and will use GI without a second
thought.
So what's the problem?
That note I received this morning was from an old friend who works in
suicide prevention. When GI was announced, he tried his standard
search of the word "suicide" and was very confused -- and then
concerned -- about what he saw.
When he typed "suicide," he was presented with page after page of
results for "suicidedoors.com" (a custom car parts firm), instead of
the usual pages of suicide-relevant results that prominently feature
suicide prevention discussions -- including a Google-inserted Suicide
prevention hotline number at the very top of the listings. I could
appreciate my friend's upset -- you want people searching on the topic
of suicide to find relevant factual listings as fast as possible
without confusion.
So what's going on?
The effect my friend noted, and that I verified with various other
searches, is a fundamental aspect of how Google Instant is
implemented. Since GI continuously updates pages of result listings
as the user types, they are presented through Google's estimation of
what the most appropriate results might be for partial search terms as
they are being entered in real-time.
The "problem" appears with words that are partial elements
(substrings) of other words and terms. Human nature being what it is,
the GI interface tends to suppress the previously standard behavior of
typing a newline after entering a search query.
So when you type the search term "man" -- there is a tendency to stop
typing at that point (after the letter "n") under GI. This currently
results in pages of search results for "Mandalay Bay" (the letters
m-a-n are bold, and "dalay bay" are in a dim gray text in the search
input box).
To get the actual desired search for the word "man" under GI you need
to enter a newline after the m-a-n.
In this example, the search results are obviously "wrong" for someone
who wants to search for the word "man" -- but will every searcher then
think to enter a newline at that point? After all, "man" is already
in bold. Reports I'm getting suggest that this is a confusing aspect
to potentially a considerable number of users.
Now it's clear how the "suicide search" problem occurs.
If a user simply enters "suicide" into GI and stops at that point, the
search results presented -- for "suicidedoors" -- are wrong and
potentially confusing from that user's point of view. The user
however sees "suicide" in bold text in the search box. At this
juncture, they must take extra actions to get the results that they
really wanted. That means thinking to enter newline in the search box
(there's no prompt to suggest this) or perhaps choose "suicide.com"
(which could be almost anything!) from the suggestions display. It
does not appear to be currently possible to use standard Windows
techniques to highlight and delete the unwanted, grayed out
("doors.com") section of the displayed search string.
It's possible to easily find significantly similar cases. A search
for "health" -- without a terminating newline after the word "health"
in bold text -- doesn't yield general health listings, but rather
seemingly endless results for "Health Net" insurance, which is the
first item on the suggested list. Good news for Health Net, perhaps
not so wonderful for most health searchers in a hurry.
What this all suggests is that while Google hasn't actually changed
search results rankings per se, the actions of the GI interface can
drastically alter the perceived searching and search results sequences
for users.
It appears that most of the time the result of these changes will
probably be neutral or positive overall. But I believe it's fair to
say that there are clearly some cases where negative effects could be
especially confusing to users -- and these situations should be worthy
of special handling.
Interestingly, a potential solution is already at hand, because Google
has already acknowledged the need to "special case" some search terms
under GI. As various observers have already noted, Google apparently
has established a "blacklist" of "not family nor work friendly" terms
that are not processed in real-time by GI, including a variety of
slang. The concern of course is that searches that matched those
terms as substrings would display pages of results during the typing
process that could prove embarrassing either to the searcher,
onlookers, or both.
When search strings on the blacklist are entered, the normal GI
real-time listings display does not appear, the user is specifically
prompted to enter a newline to receive their search results, and those
results are displayed in a manner very similar to that of the pre-GI
search procedure.
Given Google's laudable propensity to avoid manual intervention in the
search results process, they no doubt wish to minimize the length of
this blacklist.
However, I would suggest that strong consideration should be given to
using this same mechanism to avoid the sorts of effects noted above
involving searches related at least to basic health and safety
concerns.
Words like "suicide" and "health" could presumably be easily added to
the GI blacklist, along with a relatively few others of similar
import. The effect would be to return searches on such words back to
their pre-GI behavior, helping to avoid the risk of searcher confusion
on these often crucial queries.
If it's worthwhile to block some slang words to avoid unnecessarily
embarrassing Google searchers in various instances (and I agree that
this is true), then I believe that it should also be appropriate to
employ this same mechanism to help assure that a small set of critical
search terms reach the most relevant results with the least possible
confusion for potentially distraught searchers in a hurry.
This seems like common sense to me. I hope that Google agrees.
--Lauren--
Lauren Weinstein (lauren@vortex.com)
http://www.vortex.com/lauren
Tel: +1 (818) 225-2800
Co-Founder, PFIR (People For Internet Responsibility): http://www.pfir.org
Co-Founder, NNSquad (Network Neutrality Squad): http://www.nnsquad.org
Founder, GCTIP (Global Coalition for Transparent Internet Performance):
http://www.gctip.org
Founder, PRIVACY Forum: http://www.vortex.com
Member, ACM Committee on Computers and Public Policy
Lauren's Blog: http://lauren.vortex.com
Twitter: https://twitter.com/laurenweinstein
Google Buzz: http://bit.ly/lauren-buzz