[Sigia-l] People thinking like computers (was ROI/Value of Search Engine...)

[Jay Linden]

Christopher Fahey [askrom] wrote:

[...]


>>(1) A user, looking for "a low-cost, yet reliable laptop 
>>computer that will last 4 years for his son who is going
>>away to school" typed "laptops" into Amazon's search 
>>and receives 210 results, for which he has no way of
>>discerning if any of these computers match his criteria.
>>
> 
> I'll confess that my first reaction to this was to roll my eyes at this
> person's deep misunderstanding of how search engines work. But then I
> wondered: why exactly did this person type in "laptops" instead of
> something more specific like "low cost reliable laptops" or even "I am
> looking for a low-cost, yet reliable laptop computer that will last 4
> years for my son who is going away to school"? I mean, s/he wouldn't
> walk up to a salesperson at CompUSA and simply say "laptops" and expect
> to find a useful recommendation either. Why did this person only type
> one word?


Because this person is accustomed to typing "low cost reliable laptops" 
and having search engines' first 20 pages of responses be about cattle 
lowing, storm pressure zones and cold temperatures, the price of housing 
in lower Upper Volta and whether the sanitation department in somebody 
else's city can be trusted to remove the garbage on the annointed days. 
But not a reliable and objective discussion about where to find a good 
and cost-efficient laptop computer.

We already "know" that search engines won't answer "real language" 
queries reliably (see above).  We can only hope (pray?) that if we pick 
the magic word, that somewhere in the first ten minutes of results will 
come a lucky guess.  Some of our "magic word" guesses are pretty silly 
in the hindsight of the results they produce.

And _that_ is why studies like Jared's keep finding that search engines 
are both reliable and trusted when looking for very specific things -- 
those are the things that users can confidently enter a query for and 
expect to get an acceptable response, like "Townes Van Zandt" at a CD 
store.  And are unreliable and untrusted with vague queries, after 
someone's tried and failed often enough at them.


> One way to remedy this is to make smarter and smarter search engines,
> programs that are designed to adapt to the way users actually think,
> programs that do not require users to deeply understand how to use
> search engines. This is clearly what Jared wants us to do, and I agree. 


But is probably a pipe dream.  (I too agree that it is clearly a 
worthwhile, if likely a futile, pursuit.)  We are a very fragmented 
society, and the more global, the more fragmented.  It is unreasonable 
to expect we can create a search engine that will capture and analyze 
the nuances of query from people of very diverse backgrounds (and native 
languages).


> But another way to make searches more effective is to educate the public
> on how to properly use search engines. While the latter is certainly not
> realistic on any single web site, particularly a site that is trying to
> sell things, it does point to a broad issue in public education and
> social theory: We are, in fact, slowly learning to adapt to the way
> computer programs work, instead of the other way around. Web sites no
> longer need to say "click on the underlined words to jump to other
> pages". 


Maybe.  Without wanting to seem too-too negative about this, I think 
Christopher is suggesting we train people in the untrainable.  Don't 
misunderstand me: I think his analysis is good, his ideas are noble and 
worth pursuing.  But while they'll improve things, they won't compensate 
for nature's shortfalls.

I'm going to stop writing (and quoting) here with one more caveat: 
Jared's conclusions are pretty much unavoidable.  We can improve the 
search engine's ability to "intuit", but only to a degree.  We can train 
people to better understand search engine nuances (especially if the 
search engines become more consistent, which is itself improbable) -- 
but only to a degree.

Ultimately, it will always be feasible for a good search engine to find 
specific information and it will always be infeasible for a good search 
engine to read the minds of end users who aren't sure of what they seek. 
  We can isolate known repeating trouble spots and take steps to deal 
with them (such as Christopher's good example of a DVD Player Buyer's 
Guide at the end of a futile search).  But we can't create a search 
engine that does what a whole physical store or library can do.

So long as we stop trying to prove otherwise, we will understand that 
the ultimate resolution for this gap is to have human assistance 
available wherever it is likely the end users aren't going to be sure of 
what they seek.  The more prompt and patient these humans are, the 
better the users' _Web_ experiences will also be.




-- 

Jay Linden                                    Cambridge, Ontario, Canada
e-business Advisor, IBM Global Services                 jeigh at rogers.com
These onions are MINE and not those of my EMPLOYER.
Wait, that should be OPINIONS. I have no idea whose onions these are.